Can the Public Be Trusted?

In the age of MEGO (my eyes glazed over) and TMI (too much information), scientists who communicate with the public must tread a fine line between full disclosure and information overload. With current scientific output topping two million papers per year, nobody can keep up with everything. Of course, the public’s need to know does not extend to every corner of science. But for some policy-relevant fields—climate science is one—the public and policy-makers do need critical scientific information to inform important policy choices. We can’t all be climate experts, so the experts must decide what information is most useful to the public and how to frame what they share so the public can interpret it in the light of the decisions to be made.

The climate change debate has deservedly generated a public thirst for knowledge, and the climate science community has mounted an extraordinary effort through the United Nations Intergovernmental Panel on Climate Change (IPCC) to consolidate a tsunami of scientific and technical research and to try to capture the consensus of expert opinion when there is one—and to highlight areas of uncertainty when there is not. The IPCC’s Fifth Assessment Report, released in 2014, was the work of more than 800 experts divided into three topical working groups: the physical science basis; impacts, adaptation, and vulnerability; and mitigation of climate change. The physical science report alone involved more than 250 experts and ran to more than 2,000 pages in its unedited original release.

IPCC reports have long undergirded international efforts to craft a global response to the threat of climate disruption. National leaders and diplomats relied on the IPCC analyses—or at least the synthesis reports—in the United Nations Framework Convention on Climate Change (UNFCCC), which produced the Kyoto Protocol in 1997 and the Paris Agreement in 2015.

As comprehensive and authoritative as the IPCC reports are, they necessarily (and appropriately) simplify the science and cannot transfer the nuanced judgment and implicit knowledge that scientists acquire. For example, understanding the limitations of computer modeling, a core discipline in climate science, requires appreciating the importance of subjective assumptions about some future social and technological trends. Alluding to that unavoidable subjectivity, Andrea Saltelli, Philip B. Stark, William Becker, and Pawel Stano pointed out in “Climate Models as Economic Guides: Scientific Challenge or Quixotic Quest” (Issues, Spring 2015) that “models can be valuable guides to scientific inquiry, but they should not be used to guide climate policy decisions.”

The public’s inevitably limited scientific sophistication creates the potential for those who don’t like the implications of the IPCC consensus to selectively twist bits of the scientific literature to undermine the mainstream position. As science journalist Keith Kloor reported in “The Science Police” (Issues, Summer 2017), some scientists were so concerned that climate deniers were using studies that identified a short-term pause in the overall trend of global warming as evidence that the threat of climate change was exaggerated that they urged researchers to avoid addressing the phenomenon in their research. Kloor argued that such policing of the scientific literature revealed a lack of faith in the public and could in turn eventually undermine society’s belief in scientific transparency.

A few pages from here, Roger Pielke Jr. makes a case that the IPCC is exhibiting a lack of trust in the public by building into its computer models assumptions that by design advance a particular course of action in the UNFCCC negotiations. Pielke maintains that these underlying assumptions artificially narrow both the range of potential risk that we face and the variety of policy options that we should be considering. But here’s the rub: since experts need to frame scientific disputes in a way that facilitates public participation and decision-making, are the current IPCC modeling assumptions helping or misleading policy-makers and the public?

Climate science experts see an enormous potential danger in climatic changes. Their reasonable fear is that the problem is too slow moving to motivate action and yet too large and complex to manage. Pielke provides evidence that IPCC’s assumptions about future coal burning exaggerate the likelihood of catastrophic climate change in order to motivate policy-makers to act. At the same time, the IPCC’s assumptions about spontaneous decarbonization (economic changes that will reduce emissions without policy intervention), along with the feasibility of carbon capture and storage, reduce the climate risk enough to make the politically feasible actions being proposed by the UNFCCC adequate to prevent a catastrophe.

That’s not an irrational strategy. Many people who accept the reality of climate change do not think that it requires extensive, full-steam-ahead action. And many who do want to charge ahead are wary of promoting policies so ambitious that they will be tossed out as unrealistic. To motivate action, it makes sense to present the public with a serious but solvable problem. An ill-defined problem and an overly ambitious agenda could lead people to shrug or throw up their hands in despair.

Pielke’s worry is that by framing the climate issue as the IPCC does, it unwisely narrows the possible future scenarios we should be considering and the range of policy responses we should be exploring. As he rightly points out, the assumptions being built into the computer models could be wrong, and if they are, policy actions based on them may not work. He argues that the public needs to be presented with a wider field of vision of how the climate could change in the hope that it will consider a richer mix of policy options. Moreover, when important assumptions are not openly discussed by scientists, the public will rightly want to know why—especially if those assumptions turn out to be wrong.

Scientists in many disciplines face the same challenge of deciding how to frame the information they communicate to the public. Is it possible to convey the nuances of levels of certainty, the weight of the evidence, the state of development in a field? Small-sample studies versus meta-analysis? Clinical research versus big data mining? How much detail does the general public—or a congressional staffer—need? How might experts best characterize the points of controversy within a field? Is there a point of diminishing returns in public engagement? Is there anyone out there who is actually using scientific understanding to make decisions rather than to justify the predetermined outlook of the tribe?

The answers are neither simple nor obvious, and will vary with the role of science in the debate and the political environment. But the underlying principle is that if science is to play a constructive role in public policy, scientists must have the public’s trust. And for scientists to earn and keep that trust, they must trust the public. We live in a democracy, not a technocracy, and that’s a good thing. It’s tempting in these fact-challenged days to retreat to the comfort of our scientific tribe, but wisdom, unlike knowledge, is widely distributed. We should generously share knowledge—including the sources of uncertainty—so that wisdom can put it to work.

Keeping the Lights On

In September 2017, two hurricanes struck the US island of Puerto Rico, crippling its electric power grid. Because Puerto Rico is a major manufacturing site for medical supplies, the nation’s hospitals soon developed acute shortages of the intravenous bags used to administer medicines. By early 2018 the Food and Drug Administration was cautiously optimistic that the shortages would be alleviated. Even so, at that point more than half of the people in Puerto Rico still had no electricity. The role of electricity in modern life is one we take for granted—until the power goes out, with repercussions distant as well as local.

The electric power system is a subject of basic importance to Americans because universal instant access to electricity is both assumed and turns out to be unexpectedly at risk. Mason Willrich’s Modernizing America’s Electricity Infrastructure is a sophisticated policy statement by a longtime energy sector professional and analyst of high stature and deep experience. Willrich, a former executive at Pacific Gas and Electric, California’s largest utility, calls for a comprehensive response to the system’s risks, structured within the existing complex scheme of utility regulation and organization. This is a reasonable approach in principle but implausible in the existing governmental situation. Willrich is an intelligent visionary, yet he may not avoid the fate of Cassandra, whose foresight is remembered because it was not heeded.

Electric power systems have been shaped over time by three principal forces: technology, economics (including, importantly, finance), and politics (including regulation). These forces have acted in concert, though not coherently. Electrical supply began with a start-up phase, in the last years of the nineteenth century, in which alternating current technology and service monopolies at and above the metropolitan scale emerged as dominant; both endure today.

During the twentieth century, a growing grid enjoyed a long period of declining rates, lasting until the 1970s. Falling rates and rising demand reinforced one another, as technology, regulation, and the mechanisms of cost-recovery took shape in a nationwide—but not quite national—industry. There followed a period of increasing rates, in which we still find ourselves. Rising rates have in turn contributed to a sharply slowed increase in demand. This change in economic circumstances is reshaping (at times disruptively) a capital-intensive, highly regulated industry.

The electric power system now is operated by more than 3,000 entities. The retail utilities seen by consumers include nearly 50 investor-owned companies, serving more than two-thirds of the nation’s ratepayers. In addition, there are more than 2,000 publicly owned utilities controlled by a variety of public bodies, including major cities such as Seattle, cooperatives that began by serving rural areas, and regional agencies such as the Tennessee Valley Authority. Publicly owned utilities serve fewer than one-third of the customers. (The remainder of the power system includes a diverse set of owners: independent power providers, small-scale sources such as rooftop solar, and transmissions systems. These are regulated via a hodgepodge of rules and laws that vary by state.) These components of the national electric supply system are linked by transmission lines administered mainly by regional independent system operators overseen by the Federal Energy Regulatory Commission (FERC).

Customers spend, on average, somewhat under 11 cents per kilowatt-hour of electricity. This provides revenues of nearly $400 billion per year, on an asset base of over $1 trillion. Electricity accounts for about 5% of US economic output. Thanks to mobile phones charged from outlets, electricity is now a presence in every hour of most Americans’ existence. The reliability of the grid and the cost of power, as a result, matter much more than might be supposed from the quantitative contribution of the electric industry to gross domestic product.

Electric power has been a network phenomenon all along, shaped by independent sources of authority and economic and political power—never unified but coherent enough to allow interconnection of different geographic provinces and technological systems. The result today is a collection of electric utilities, technologically connected to one another, that is regulated in an intricate scheme of laws and policies administered by multiple state and national regulatory bodies. Whether owned by shareholders or governmental entities, an electric utility operates under economic and political forces unlike those facing conventional businesses or government agencies. As this thumbnail sketch indicates, Willrich has undertaken a formidable task in describing and analyzing this unusually complicated industry; a reader needs some fortitude to follow the author’s guidance through this labyrinth.

Electric power began as a vertically integrated industry, with a single firm owning the wires going into customers’ homes and businesses, as well as the distant power plants that generated the current flowing in the lines. Over the past generation a series of policy and financial changes—loosely grouped under the term “deregulation”—has shifted much of the ownership of generating resources onto independent power providers.

The generation of electricity was in the midst of a technological transition by 2015. One-third of power came from coal; this was a steep decline from nearly half less than a decade earlier. The change is driven principally by the availability of low-cost, relatively clean natural gas. This transition away from coal is unfolding with unprecedented speed, in an industry where major investments are planned to last many decades. Another significant component of electricity generation is the US nuclear fleet, which is still the largest in the world, as Willrich points out, even though there have been few additions since the 1980s.

Under policy mandates adopted with strong support from environmentalists, most jurisdictions are moving toward a system of mixed but coordinated power supply that includes efficiency and renewable sources, notably wind and solar. Largely unnoticed is the challenge of integrating the new, often decentralized sources of power into a transmission and distribution system designed around large central-station power plants. The policies that induce renewable resources frequently do not include provisions for rebuilding a grid that can distribute the power they provide. The retail utilities are left with the mandate of providing electricity whenever it is needed, but with revenue streams poorly aligned with the technological realities of this task.

More broadly, the cost of electricity is largely that of the equipment used to supply it, and much of that capital is funded through borrowing by utilities and power producers. The borrowed money is paid back over time—but more slowly than the competitive forces of natural gas or the policy mandates for renewables seem to allow.

Challenges loom accordingly. Stagnant demand restrains the revenues that power companies can collect. This is because end users can take steps to limit their own demand, but also because the regulatory authorities are reluctant to permit rate increases. Over the longer term, global climate change seems likely to require significant technological change: the provision of electricity currently accounts for 40% of the nation’s greenhouse gas emissions. How to pay for a large-scale rebuilding of the electricity supply and transmission system in order to reduce those emissions, in light of the modest returns on capital currently allowed, is not at all clear. California is wrestling with this problem now, learning lessons that may be illuminating (as was the case with the state’s unhappy experience with deregulating electricity markets nearly two decades ago). Closer to hand, perhaps, is the range of risks that accompany an essential resource of postindustrial life that remains so low in profile that difficulties gather out of sight.

In response to the real but underappreciated fragility of the electric power industry, Willrich argues that four goals are important national priorities:

• First, the grid must continue to be reliable, supplying power as it is needed. It should also be resilient in the face of the extreme weather events that are inevitable as climate change accelerates; resilience is what Puerto Rico lacks.
• Second, the component utility systems need to preserve interoperability while strengthening their security against cyberattack. That is, utilities must continue to respond to instantaneous changes in demand across their service territories while guarding against hackers.
• Third, the electric power system should continue to lower its emissions of greenhouse gases.
• And fourth, it is essential that electricity remain affordable. This needs to be done in such a way that the utilities can continue to finance their operations and the investment required to transition toward renewables and energy efficiency.

These goals are technical and intricate, even as they are important. Turning them into durable policy demands coordinated actions; so deep are the interdependencies that no simple strategy is plausible. To the discomfort of many environmentalists, addressing climate change may require long-term subsidies to nuclear power plants, which do not emit greenhouses gases in their operation. Wide-ranging changes to wholesale markets for electric power will be needed to facilitate the transition from fossil fuels to renewable sources and energy efficiency investments; consumers are unaware of the costs and institutional difficulties of doing so. Today’s populists continue to deny the economic reality that low-cost natural gas has doomed the coal industry. From all directions, the benign inattention that the public has lavished on electricity stands in the way of the changes needed to address the electricity we take for granted.

Willrich’s central message—although implicit—turns out to be that the institutional and technological complexity of the grid functions so smoothly on the surface, day to day, that consumers, voters, and nearly all leaders do not perceive its fragility. Neither do they understand the financial, political, and managerial burdens of conserving what is valuable. Unlike, say, national defense or health care, the energy system is a trillion-dollar nexus where public discussion is nearly impossible to marshal without a crisis to focus national attention. And it is sufficiently complex and delicate that a crisis is unlikely to be a good venue to make sensible choices: we need to rebuild the ship while it is underway, not head for the lifeboats as it sinks.

Willrich proposes policy reforms led by the federal government and involving the utility commissions of every state. Some states will continue to rely on deeply entrenched coal-fired generation, while others, such as Hawaii, aspire to a wholly renewable power supply. Many state commissions and regional operators chafe under the regulatory policies of FERC. The reform ideas that Willrich puts forth make sense conceptually, but require trust and careful choreography that currently appear to be in short supply. For instance, the decline of the coal industry has been blamed on attempts to lower greenhouse gas emissions, rather than on the technological innovations that made natural gas cheaper; this confusion has shifted debates about energy regulation in unproductive directions.

There are many, many more examples of the difficulties and confusions that must be surmounted. Doing so would be a formidable task even without the deep suspicions of government now embedded in US public life, and Willrich doesn’t acknowledge the scale of this challenge. But to citizens and leaders in business, civil society, and government who grasp the need for modernizing the nation’s electric infrastructure, Willrich’s book is a good way to begin learning about the tangle of wires and institutions behind the placid socket on the wall.

The erosion of the infrastructure of developed economies is a slow-gathering crisis whose scope and implications are hard to see until the costs of effective response become painfully high. In this electricity is not unlike other profound challenges facing the United States: streets and bridges, ever more congested, become structurally weakened and studded with potholes; public schooling is neglected until the workforce cannot keep up with economic change; central banks lower interest rates in order to prop up growth but then can no longer respond to downturns; and the electric power system weakens, invisibly, until a systemic outage exposes the limitations of its architecture.

The story of Puerto Rico is a worrying harbinger. The island’s recovery from the recent disastrous storms is hampered by an electric utility that was already bankrupt before the hurricanes struck. Puerto Rico’s government is moving to privatize the hapless public agency running the island grid. But federal assistance in rebuilding the electric power system has been limited by law to restoring what was there—which is quite different from a resilient system, in which solar and other renewables would play a significant role in a way that makes sense for users and the economy. The mismatch in Puerto Rico, between what is needed going forward and the economic and institutional structure that is responsible for doing so, resonates far beyond the island. More than 60 years ago, the musical West Side Story contained these bitter lines, sung by a Puerto Rican character: “Puerto Rico, you ugly island … always the hurricanes blowing.” The refrain that follows is “I like to be in America!” But today it is in America where the wind threatens to blow, hard.

The Natural Gas Grid Needs Better Monitoring

We are familiar with cascading electric grid outages such as the September 8, 2011, blackout that hit San Diego at rush hour, and the August 14, 2003, blackout that essentially shut down the Northeast. Less familiar are failures in the US natural gas pipeline system. But they occur.

Fuel-starvation outages at US gas power plants happened at an average rate of a thousand events per year and affected one in five plants between January 2012 and April 2016, according to the North American Electric Reliability Corporation (NERC). Sometimes, in very cold weather, many gas plants are starved of fuel at the same time.

Because data on the reliability of the natural gas pipeline system is almost impossible for anyone to find, our team spent a year combing through the reports filed by power plants—not pipelines—to count these outages. To our knowledge this is the first time anyone has done so.

Unlike electric power generator failures, gas pipeline outages are either not recorded or not available without a Freedom of Information Act request in most states. But disruptions in the natural gas system can have serious consequences, particularly for electric power generation.

For power system reliability, it is important to know how often, where, and why pipeline failures occur so that power plant operators can be better prepared for gas interruptions. Storing backup gas supplies at the generator site is impractical because the required tank farm to hold compressed gas for just one day’s power plant operation would increase the plant’s footprint by at least 10%, and that doesn’t even consider the ancillary equipment required to support the gas storage. Liquefied natural gas storage, even for a few hours’ worth of plant operation, is very expensive. And underground storage at the plant is equally impractical for most plants. Another option to protect against gas supply interruptions is to design in fuel-switching capability that can easily substitute oil for gas. But only one-quarter of gas power plants have the ability to switch to oil without halting operation, and about half of those plants can operate for only a short time with oil because of on-site oil storage limitations.

The remaining three-quarters of plants that do not have fuel-switching abilities are tied to the real-time reliability of the natural gas pipeline transportation network. When emergency situations arise on the natural gas grid, pipeline operators turn to a load-shedding protocol that outlines the order in which customers will have their gas supply turned off. The shedding of load restores operational stability to the gas grid in situations of high stress.

On the other side of the gas meter, however, as pipeline operators carry out their load-shedding procedure to restore stability to the gas grid, power plants might have to shut down, forcing other plants to increase their electric output. If the generation shifting creates a large enough stress on the electricity network, other power plants sometimes fail, creating further instability on the electric grid.

Under current reporting requirements it is possible to obtain only an incomplete picture of the frequency of these kinds of interdependent natural gas/electricity infrastructure failures. Recent lessons in interdependency between the gas and electric grids (see Box 1) are a call to action to better align data availability of both grids’ operational characteristics. This is not a new message. In 2013, NERC released phase II of its special reliability assessment report titled “Accommodating an Increased Dependence on Natural Gas for Electric Power.” It identified a lack of “compiled statistical data on gas system outages” that would be equivalent to the databases that NERC has complied in its Generating Availability Data System (GADS). NERC called on the natural gas transmission sector to work with it to establish a central pipeline outage database that would make it possible to conduct reliability analyses of the dual-grid system.

NERC’s message has been heard in the academic community. Currently, academic teams across the country, ours included, are exploring the issues presented in the special reliability assessment. But nothing has been done in the ensuing years to fix the data misalignment. We just don’t know how vulnerable the nation is, and we don’t know where to apply management attention to reduce the vulnerabilities.

To address this problem we explore the current federal reporting standards relevant to quantitative analysis of the reliability of the dual-grid system as they exist today and recommend a path of development for the central database recommended by NERC.

For electric generators, the GADS Data Reporting Instructions outline specific, numerical thresholds for mandatory reporting. Events are to be reported that cause any power plant with nameplate capacity of 20 megawatts (MW) or greater (the vast majority of all plants) to fail at start-up, to be completely unavailable unexpectedly, or to be unable to provide the full amount of power the plant promised to the grid. Power plant “derating” reports are mandatory for all events causing the equivalent of 2% or more of the power plant’s net maximum capacity to be unavailable for 30 minutes or more. A cause identification code is included with every power plant failure report. Between January 2012 and April 2016, more than 1,000 failure events per year were reported by power plant operators claiming lack of fuel from the gas pipeline network. The data from these reports are confidential, but aggregate data that are fine for measuring overall reliability have been published.

Reliability events for gas pipelines, on the other hand, are reported to various entities, but with reporting thresholds that vary by jurisdiction. The Federal Energy Regulatory Commission (FERC) has jurisdiction over operation of interstate pipelines; the Pipeline and Hazardous Materials Safety Administration (PHMSA) for interstate and intrastate pipeline safety; and the state Public Utility Commissions for intrastate pipeline networks—mostly for local distribution companies. According to mapping data provided by the Energy Information Administration, roughly 60% of natural gas power plants with capacity of 20 MW or larger are within five miles of an interstate pipeline. The remaining 40% are probably fueled by smaller, intrastate pipeline systems. Therefore, it is important that reliability data be available for both interstate and intrastate pipelines. Because the US natural gas grid does not have a central reliability organization, compiled data sources that are sufficient to model interdependencies between the gas and electric systems are hard to find.

One promising data source that could meet the needed criteria arises from a FERC rule that requires “emergency transaction” reports (Form 588) from pipeline operators. An emergency transaction occurs as a result of “any situation in which an actual or expected shortage of gas supply would require an interstate pipeline company, intrastate pipeline, local distribution company, or [pipeline that is not under FERC jurisdiction due to stipulations in the Natural Gas Act] to curtail deliveries of gas or provide less than the projected level of service to any customer.” The reporting requirements of the regulation could be read to require transaction records for partial as well as complete gas curtailment events.

But this is only one way to read the rule. By our interpretation of the definition of an emergency transaction, the FERC-588 reports should capture the data that are needed to study reliability, but they don’t. The filings under FERC-588 and other gas pipeline emergency reports are available on FERC’s eLibrary website. Searching the eLibrary for emergency filings using the keywords “interrupt,” “outage,” or “curtail” produces 32 results from 17 unique pipeline events between 2012 and 2015. Most of the events were for gas flow diversions to avoid pipe segments taken out of service for maintenance. In these cases, the emergency transactions were brokered to avoid gas interruptions to customers.

However, despite the fact that multiple delivery failures have occurred, only one report over the period details a service interruption that could have affected a power plant located on the pipeline. Thus, the FERC-588 data are no help in understanding the reliability of the natural gas system.

In March 2011, a gas gathering line in the Gulf of Mexico was struck by a dredging operation and knocked out of service for over 250 days. In January 2016, a 30-inch steel transmission pipeline in the Southwest ignited due to a rupture of the pipe material. The explosion caused service to be interrupted on the pipeline for 35 days as repairs were made. In July 2016, while crews at a western gas distribution utility worked to fix a leaky valve, they accidentally struck a 4-inch plastic main, causing the gas to ignite. Extensive system damage occurred, 30 people were evacuated, and gas service was shut down for a day.

Not one of those events is in the FERC data.

Since the FERC data are not very informative, the most comprehensive, easily accessible, centralized source that captures both inter- and intrastate pipeline data is the PHMSA Natural Gas Distribution, Transmission & Gathering Accident and Incident Database. The one service interruption in the FERC data is also captured by the PHMSA database. These data are filed by the pipeline operators and have been gathered since 1970. The data are compiled and catalogued with a description of each pipeline incident and its subsequent root-cause investigation. PHMSA makes these data available publicly on its website. The thresholds that trigger a mandatory report to PHMSA include an event that results in a release of gas or hazardous liquid from the pipeline as well as at least one of the following: a death, or personal injury necessitating in-patient hospitalization; estimated property damage of $50,000 or more … excluding the cost of gas lost; or unintentional estimated gas loss of three million cubic feet or more.

The regulatory language also calls for reporting any event that is “significant in the judgment of the operator, even though it did not meet the [previous] criteria … of this definition.” As PHMSA is a safety-centered organization, the thresholds focus on safety-related metrics; however, some of the fields on the forms that pipelines operators and investigators submit to PHMSA after an incident capture important reliability metrics such as the system component affected, shutdown time, and the primary cause.

An analysis of the 673 PHMSA accident and incident reports for distribution, gathering, and transmission pipelines between 2012 and 2015 shows that approximately 80% of reports met at least one of the automatic report conditions, and 20% did not. The 131 reports that did not meet the conditions can be viewed as those “judged significant” by the pipeline operator. But as mentioned in Box 1, the serious events at Aliso Canyon and in New Mexico are not found in the data available on PHMSA’s website. This leads us to wonder how many other significant events are missing from these data, or even what a significant event is judged to be.

The only way we can effectively study interdependent reliability is if the standards for reporting pipeline outages and power plant failures are sufficiently equivalent. In comparing the GADS and PHMSA reporting thresholds, it is evident that the language for reporting outage events at power plants is far more stringent than for gas pipeline outages. Again, this is probably because PHMSA’s mission is safety, and there is no central reliability organization for the gas network.

A 460 MW combined-cycle natural gas power plant (the median size of such plants) consumes the equivalent of almost three million cubic feet of natural gas per hour at normal atmospheric pressure. That means that an unintentional release of three million cubic feet of gas (the threshold for making a PHMSA report) represents just over one hour of the power plant’s full operation. For electricity-side reporting at a power plant of this size, a complete power plant outage of any duration or a derating event equivalent to just 2% of the plant’s capacity for 30 minutes or more must be reported. That is the equivalent of 30,000 cubic feet of natural gas at atmospheric pressure. The event would have to be 100 times as large to trigger a PHMSA report. But power plants are fueled by high-pressure natural gas supplies, which means that a 2% derating for 30 minutes represents roughly only 600 cubic feet of gas consumption at pressure, 5,000 times less than the PHMSA threshold.

Box 1. Gas-Electric Interdependence

In February 2011, an extreme weather event hit the Southwestern United States, chilling local temperatures to as low as 30 degrees below zero. The temperature dropped so low in places that water vapor at natural gas wellheads froze, restricting flow from production areas to the residents of the area. Simultaneously, regional electric power plants failed to keep up with demand due to inadequate planning for the unexpected cold weather. The Electric Reliability Council of Texas reported that over the first four days of February, 152 individual generator units at 60 power plants in the state didn’t provide the electricity they promised, triggering the initiation of rolling blackouts. More than 75% of the units reporting forced outages relied directly on natural gas as their primary fuel source. On the first night of the event, more than 8,000 megawatts of power generation unexpectedly dropped offline; that was 12% of the entire installed capacity of the electricity grid.

Further compounding the problem, a segment of the regional pipeline system that shipped natural gas from unfrozen production wells in Texas to markets in New Mexico and farther west relied on Texas grid electricity to power its compressor stations. When the rolling blackouts started, the electric compressor stations shut down, and the gas pressure in the regional pipeline system fell, starving customers in New Mexico of natural gas for heating. When all was said and done, 28,000 natural gas customers in New Mexico were forced to find other ways to protect themselves and their families from the bitter cold.

When large natural gas storage facilities fail, they wreak havoc on fuel supply stability for power generators. In October 2015, a 7-inch injection well casing at the Aliso Canyon natural gas storage field in Southern California failed, creating the largest natural gas leak in US history. Nearly four months passed as the operator and emergency responders worked to contain the leak.

This example helps illustrate why we think it is wrong that the only numerical, operational threshold for automatic gas pipeline incident reporting to the most comprehensive database is the volume of gas released. Gas volume released, although important for financial, environmental, and safety reasons, is inadequate for system reliability analysis. Fluctuations in system pressure, or volumetric flow rates, are the important system variables for gas system reliability as they characterize a pipeline company’s ability to serve loads. Furthermore, as the language specifies, the explicit thresholds trigger a mandatory report only for incidents that occur simultaneously with an unintentional release of gas or hazardous liquid. Important reliability events without releases of gas from pipelines, such as reductions in operating pressure of the gas system, are left out of these explicit definitions. In the absence of more encompassing data, reliability analysts working with the PHMSA data are left to depend on the events that the operator judges to be “significant.”

Perhaps more appropriate data are collected through other means and have been used internally for reliability assessments of the gas grid. We have not seen any reasons to believe this is the case, but even if it is, an internal assessment isn’t as good as having an open community reliability analysis, which would provide regulators and the many stakeholders of the gas grid with valuable information while also reducing the administrative burden of completing these analyses in-house. State agencies, academic institutions, trade organizations, businesses using gas for emergency backup generators, and large natural gas consumers such as power plants should be provided access to pipeline reliability data that are not deemed a threat to national security. For power plants, these data are crucial for both siting of new plants and contracting for gas supply. Access to data that can capture events on interstate and intrastate pipelines with the potential to affect the bulk power network should be provided outside the walls of government so experts across the country can analyze the reliability of the interdependent gas and electric grid systems on a level playing field.

In September 2013, the National Association of Pipeline Safety Representatives (NAPSR), an organization with ties to the National Association of Regulatory Utility Commissioners, released a document titled “Compendium of State Pipeline Safety Requirements & Initiatives Providing Increased Public Safety Levels compared to Code of Federal Regulations.” In the report, NAPSR noted that state regulators had 308 enhanced reporting initiatives in place that would require pipeline operators to report safety conditions above and beyond those required by federal standards. They also reported that 33 states had various types of enhanced reporting standards with specific reference to the regulation underlying the PHMSA reporting thresholds. These enhanced standards included lowered property damage thresholds, outpatient injury reports, and other modifications to the FERC regulatory language.

Some important initiatives identified by NAPSR require pipeline operators to report outages affecting a specific number of customers, outages of a specific duration, or complaints of gas delivery pressure issues. At the time that the compendium was released, 20 states had one of these categorical reporting standards in place.

The problem is that each of these 20 states has its own reporting thresholds with varying stringency. For instance, Pennsylvania requires reports of all gas outages affecting the lesser of 2,500 customers or 5% of total system customers. Florida requires reports of outages affecting the lesser of 500 customers or 10% of total gas meters on the pipeline network. Washington requires reports of outages affecting more than 25 customers. Wyoming requires reports of all service interruptions of any size.

The state reports appear to be a step toward solving one piece of the reliability puzzle. But only three states—New Hampshire, Rhode Island, and Washington—were listed by NAPSR as having a reporting requirement for system pressure issues. As discussed in Box 2, system pressure fluctuations without a complete gas outage can shut down gas turbines. One state, Maine, requires reports of all gas interruptions longer than a half hour that affect other utilities’ critical facilities.

Data accessibility is also state-specific. Some states, such as Wyoming and Pennsylvania, make the records they collected publicly available on their state information portal websites (if you know what search terms to use to find these data). In other states, the data from the records are referenced only as footnotes in annual pipeline safety reports or are simply unavailable, requiring a Freedom of Information Act request to access the records.

Box 2. Partial gas failures are also a problem

Complete natural gas outages are not as common as failures that drop the pressure in the pipeline. Power plant facilities are designed to receive natural gas from pipelines at a contracted pressure and volumetric flow rate based on pipeline capacity and their generator equipment specifications. For example, two common natural gas turbines built by General Electric (GE), the 50-megawatt model LM6000 and the 85-megawatt 7EA, require incoming natural gas pressures of 290 and 675 pounds per square inch (psi), respectively. The Natural Gas Supply Administration reports that natural gas is typically transported in interstate pipelines at pressures between 200 psi and 1,500 psi. The lowest-pressure interstate pipelines require power plant operators to maintain additional on-site compression equipment to run either model of the GE turbines. Pressure reductions in the lowest-pressure interstate pipelines add stress to these on-site compressors. Even for the highest-pressure pipelines, a 55% drop in pressure would put a generating unit using the 7EA at risk of operational failure. An event causing an 80% reduction would put the LM6000 at risk of operational failure.

To properly manage an increasingly interdependent gas and electricity system, the federal government should build on the states’ efforts in updating the reporting thresholds for natural gas pipeline incidents to better align with the power plant outage standards and create a national standard. We recommend that pipeline incidents of sufficient size to trigger a mandatory power plant outage report should be reported. This additional threshold should be a specific requirement of pipeline systems with active firm supply contracts with power plants. This recommendation is based on the agreement between the pipeline and the power plant that a firm contract is meant to imply: there will be no unplanned curtailment of natural gas service unless necessary in an emergency.

Construction of any new standards should be based on the average amount of natural gas heat input required to produce a unit of electricity (the power plant heat rate) and modified to correspond to the most stringent power plant outage standards. The new standard should also be periodically revisited or updated to account for technological advances.

For pipelines with firm gas service contracts to serve power plants of over 20 MW nameplate capacity, events that reduce the pipeline’s ability to deliver to a plant gas equivalent to 25,000 standard cubic feet per hour should be reported. Pipelines with firm service contracts in place to serve power plants with nameplate capacity of 20 MW or less should report events that reduce the pipeline’s ability to serve the plant by 900 standard cubic feet per hour. These thresholds are based on the average heat rates of an advanced combined-cycle power plant and a baseload distributed generation plant, respectively. They are scaled to represent 2% of the median plant’s net maximum capacity, the power plant reporting threshold.

During the development and implementation of this new standard, stakeholders of the electric and natural gas industries should be consulted. We recommend that representatives from the American Gas Association, the Gas Technology Institute, NAPSR, PHMSA, and NERC should be included. During meetings with these groups, a key topic of discussion should be to better define what “an event that is significant in the judgment of the [gas system] operator” should include for natural gas pipeline incident reporting and to whom certain types of significant events should be reported. Pipeline operators closely guard their data for internal use. The new standard should be crafted in a manner that preserves proprietary trade secrets while also identifying the information that must be collected to conduct reliability analysis of the whole pipeline network.

We also recommend that the government use the New Mexico and Aliso Canyon events as the impetus to follow the electricity sector’s example by designating a central entity to oversee the reliability of the natural gas delivery system. After the 2003 Northeast electric blackout, Congress passed the Electric Policy Act of 2005. The act authorized FERC to appoint an Electric Reliability Organization with authority to establish and enforce mandatory reliability and reporting standards for electricity utilities throughout the United States. In 2006, FERC appointed NERC to that role. Similarly, Congress and FERC could require the establishment of a national natural gas pipeline reliability organization.

The PHMSA data discussed earlier comes from an organization with the mission of “protect[ing] people and the environment by advancing the safe transportation of energy and other hazardous materials that are essential to our daily lives.” Because safety is PHMSA’s core mission, their data are unsuitable for conducting a thorough reliability analysis of the natural gas network. Instead, the effort to organize a central, NERC-like gas reliability organization could be spearheaded by a group such as NAPSR, with ties in both industry and government. Congress should replicate what it did for electric power.

Experts at NERC should provide guidance to the gas reliability organization. NERC’s involvement in the early stages of this effort could provide not only important lessons learned during its own establishment but also the foundation for a collaborative relationship between NERC and its gas counterpart. Given that the United States produces the largest share of its electricity from natural gas, it is critical to coordinate reliability issues between the two grids.

Philosopher’s Corner: What Is Science in the National Interest?

It all began in the middle of the fourth century BCE. Aristotle’s Nichomachean Ethics established a hierarchy of goods that laid the foundation for today’s science policy debates. A good pursued for its own sake would thenceforth lay claim to greater worth than goods pursued for the sake of something else. Although he recognized that applied research was also valuable, Aristotle concluded that laws should be passed to secure the pursuit of the highest good—the life of contemplation for its own sake. In today’s parlance, basic, curiosity-driven research is more valuable than applied, and society should serve science.

Scholars of science policy usually trace controversies over what research, if any, deserves public funding to the period between the publication of Vannevar Bush’s Science, The Endless Frontier in 1945 and the creation of the National Science Foundation (NSF) under the National Science Foundation Act of 1950. Keen to capitalize on the role that science and technology played in winning World War II, Bush argued that basic research—done for its own sake, without regard for possible applications, initiated and carried out by scientists on their own terms—was necessary to benefit society. President Truman had vetoed an initial version of this legislation, the National Science Foundation Act of 1947, and that seemed to quash Bush’s Aristotelian vision. Shortly thereafter, John R. Steelman, the first person to ever hold the title of Assistant to the President, issued another report, Science and Public Policy: A Program for the Nation, which recommended a different approach. Whereas Bush had strongly advocated scientific autonomy, Steelman argued for a much larger role for government in managing the research enterprise and linking it to national needs. Neither approach fully won the day, though the NSF created in 1950 more resembled the Bush than the Steelman proposal, preserving a great deal of autonomy for the scientific community to determine which research should receive public funds.

A February 16, 2016, New York Times editorial, “The Chirp Heard Across the Universe,” joined this battle over science policy. It opined that the discovery of gravitational waves by the Laser Interferometer Gravitational-Wave Observatory (LIGO) demonstrated that the enhancement of human knowledge needs no additional justification. LIGO was “simply cool” and clearly showed the value of the 40-year, $1.1 billion NSF investment.

The point of the editorial, however, was not primarily to celebrate LIGO’s discovery. Instead, it was to criticize a bill introduced by Rep. Lamar Smith (R-TX), chair of the House Committee on Science, Space, and Technology. Smith’s Scientific Research in the National Interest Act was supposed to make NSF and the researchers who receive its funding more accountable to the US taxpayer. Smith’s notion of accountability would be satisfied, in part, by requiring researchers to articulate in nontechnical language the benefits to society of their research.

Although such a requirement had been in place since October 1997, when NSF began making “broader impacts” one of the criteria applied in assessing research proposals, Smith’s bill would also have required researchers to explain how their grants were “in the national interest.” A list of activities that would satisfy the requirement was also provided. The Times editorial claimed that requiring NSF grants to be justified in terms of their being in the national interest would amount to “political meddling” of the sort that would have killed LIGO funding. Smith was quick to respond: “Contrary to your suggestion, the LIGO project would certainly fall under the legislation’s national interest definition to ‘promote the progress of science in the United States.’” Despite the fact that they disagreed about how to justify the claim, both the Times and Smith agreed that LIGO had been worth funding.

Though Smith’s bill was passed by the House, and despite the fact that it generated a great deal of discussion, it never became law. Instead, Congress passed the American Innovation and Competitiveness Act, which contains a modified version of Smith’s bill. Smith’s list of national interests survived largely intact. But the law replaces “promoting the progress of science in the United States,” which was Smith’s justification for LIGO, with “expanding participation of women and individuals from underrepresented groups” in science.

According to the American Innovation and Competitiveness Act, every grant by NSF is supposed to support research that satisfies one or more of these goals:

1. Increasing the economic competitiveness of the United States.
2. Advancing of the health and welfare of the American public.
3. Supporting the national defense of the United States.
4. Enhancing partnerships between academia and industry in the United States.
5. Developing an American STEM [science, technology, engineering, and mathematics] workforce that is globally competitive through improved pre-kindergarten through grade 12 STEM education and teacher development, and improved undergraduate STEM education and instruction.
6. Improving public scientific literacy and engagement with science and technology in the United States.
7. Expanding participation of women and individuals from underrepresented groups in STEM.

Let’s take for granted that everything on the list is in the national interest of the United States. Nevertheless, questions arise.

Are all seven interests of equal worth? Suppose NSF receives four grant proposals, each in the same area of research, each with equally qualified teams, each requesting the same amount of funding, yet each focused on meeting different needs. One will help the economy, one will support national defense, one will improve K-12 education, and one will expand participation of women in STEM fields. NSF has enough money to fund only two. Does the list provide any way of helping NSF decide? And if it did, wouldn’t that be politicizing peer review?

Should proposals meeting more of these needs receive more funding than those meeting fewer? Maybe the proposal that will help the economy will do so by forming a partnership between the researchers and industry. The K-12 education proposal might also improve public scientific literacy. Are these proposals now more worthy of funding than those supporting national defense or the participation of women in science?

Is the list exhaustive? “Promoting the progress of science” was left off the final list. Are we to presume that promoting scientific progress is no longer in the national interest? Why did Congress replace it with expanding participation of underrepresented groups? What about graduate education? The list explicitly mentions only pre-kindergarten through grade 12 and undergraduate education. And do we expect our national interests to remain the same over time?

Why suppose that it is NSF’s responsibility to pursue these interests? Is it not enough to have the Departments of Defense, Treasury, Commerce, and Education, as well as the National Institutes of Health, to promote the nation’s security, economic strength, student development, and public health? Are any of the listed national interests the special provenance of NSF?

Why suppose that Congress is especially in touch with national interests? Congress is made up of representatives of the citizens of all 50 states, plus Washington, DC, Puerto Rico, and four US territories. Although all swear an oath to support and defend the Constitution, members of the House represent the interests of their constituents, and senators represent the interests of their states. More cynically, members of Congress may represent special interests more than national interests. Perhaps the sort of horse trading that happens in Congress is meant to result in judgments that are, in the aggregate, in the national interest. But it may be worth considering other ways to approach the question.

In October 2017, Sen. Rand Paul (R-KY) introduced the BASIC Research Act (S. 1973). Among several intriguing sections contained in this bill is the suggestion that NSF should include two new sorts of “peers” on review panels: a nonacademic expert from a field different from that under consideration for funding and a “taxpayer advocate” focused on determining the value of the proposed research to the US taxpayer. Rather than relying on a congressionally predetermined list of national interests that tries to cover all the bases, the bill would rely on taxpayers to make judgments on a case-by-case basis. It would be possible, then, for a taxpayer advocate to assess the value to the nation of a wide variety of research proposals.

Of course, Paul’s bill raises its own set of questions. How would the taxpayer advocates be chosen? Does it matter whether they are Democrats or Republicans? Would they have veto power on funding decisions? Would scientific peers listen to the views of nonexperts? Would a taxpayer advocate be swayed by how cool the research seemed to be? These questions, and more, would have to be answered before we could answer the main question: What does it mean for science to be in the national interest?

Although the mechanism for answering the question is different in the existing law and Paul’s bill, the basic intuitions underlying both seem to be the same:

• If taxpayers are funding research, the research ought to benefit the society funding the research.
• If research benefits society, it ought to be either obvious or defensible on those grounds.
• Although the vast majority of scientists who propose, review, and receive funding for research from NSF are US citizens, scientists who have a scholarly stake in the sort of research that is funded cannot be trusted to put national interests over their professional interests—or perhaps even to consider societal benefits in their assessments.

NSF’s Broader Impacts Merit Review Criterion already satisfies the first two. But it depends on the good faith of the members of the scientific community to take benefits to society seriously as a criterion of funding. Members of Congress may still not fully agree on how to hold scientists receiving federal funding accountable to the public, but current laws and proposed legislation indicate agreement on one key point: Scientists cannot be trusted to pursue science in the national interest of their own accord.

In the end, we come back to Aristotle. Should society serve science, or should science serve society? Aristotle argued for the former, on the grounds that science (metaphysics, for Aristotle, which would reveal the truths for people to contemplate) pursues the highest good. From an Aristotelian perspective, basic scientific research itself is in the national interest, just because it is cool; society should be organized to foster its pursuit. To the extent that members of the scientific community share this belief, they will view legislation meant to guarantee accountability as political interference.

This also reveals a crucial difference between science in Aristotle’s day and our own. Aristotle did not have our assumption that science is supposed to enable technological development that will spur economic growth. This has become a truism—a truism that now needs questioning.

It is obvious in theory that some science is in the national interest, can benefit society, and could change the world for the better. To avoid a bad outcome for both science and society, we need to go beyond the Aristotelian fixation on the life of contemplation. Scientists may want society to give them money and leave them alone to pursue truth on their own terms, but society today seems no longer to like that deal (if it ever did). Scientists may hold the keys to unlock knowledge of eternal truths, but Congress holds the purse strings. As Sen. Paul put it, “There’s a lot of bizarre stuff that everybody agrees should not be going on. And if you don’t fix it, the danger is that people [in Congress] will get tired and there won’t be any more money for research.” Put differently, if scientists fail to answer the question “what is science in the national interest?” society will answer it for them. Or maybe society will start asking a more pointed question: Is science in the national interest at all?

Varieties of Technological Experience

For roughly two thousand years, from 500 BCE to 1500 CE, China was the most advanced scientific civilization in the world. During this period, the Chinese intellectual tradition was more knowledgeable about nature and more technically creative than any other in world history. Then around 1500, China began to lose its leadership. Why did China fail to maintain its dominance, ceding superiority in science and technology to Europe?

This question came to fascinate the great British biochemist Joseph Needham (1900-1995). No Westerner did more to celebrate the myriad Chinese contributions to science and technology than Needham. His multivolume Science and Civilization in China series, which he began publishing in 1954 and continues—27 books later—to this day, is described in the popular historian Simon Winchester’s The Man Who Loved China (2008) as a magnum opus comparable to Aristotle’s body of work.

Needham’s project originated in 1937 with his serendipitous exposure, through an affair with a postdoctoral student from Nanjing, to the richness of a civilization older, larger, and more continuous than any in the West. As an ardent disciple of Western science—and with a belief in its essentially universal character—Needham set out to identify hidden continuities between Chinese and European discovery and invention. Today, the Needham Research Institute is a leading center for continuing work on the history of science broadly construed (to include technology and medicine) in Asia.

In the same year as the Needham multivolume book project began in England, the German philosopher Martin Heidegger (1889-1976) published, from a series of talks given to German engineers, “Die Frage nach der Technik” (“The Question Concerning Technology”). This small essay, which took issue with the Western celebration of its technological and scientific superiority, has become a major influence on philosophical criticism of modern technology. For Heidegger, the truth of modern science is neither its rejection of superstition nor its development of the Big Bang cosmological theory.

Yes, science’s applied technological powers—from the steam engine through internal combustion to electricity and nuclear reactors—transformed the human condition by lightening the burdens of work, increasing material affluence, accelerating transport, and intensifying communications. At the same time, scientific and technological advances alienated workers from their industrial labor, undermined sociocultural stabilities, and left people feeling homeless in the cosmos. Social traditions of consensus about the good life and its meaning fractured, to be replaced by consumerism, pursuit of power, or both. The unmoored cosmopolitanism that a technologically sustained globalization seems to foster can give rise to reactionary political movements (see, e.g., Trumpism).

Heidegger fully recognized the enormous engineering achievements of the twentieth century. But what he considered more fundamental than the physical transformation of human experience was a metamorphosis in cultural assumptions and practices. For modern humans, basic beliefs about the nature of reality have changed. Whereas in premodern traditions humans understood nature as an ultimately stable order with which they aspired to live in harmony, the modern worldview sees nature as a plethora of resources available “for the use and convenience of man” (to quote the 1828 Royal Charter of the Institution of Civil Engineers). More than simply a given that might evoke homage and wonder, nature in the West today is something to be reconstructed by the multiple needs and wants of human makers. Insofar as innovation trumps contemplation, people find it increasingly difficult to appreciate what they have, to take grounded pleasure in the particularities and beauty of the world.

In The Question Concerning Technology in China, a young Chinese computer scientist and philosopher, Yuk Hui, makes a bold attempt to reassess the ideas of both Needham and Heidegger. There exists no more challenging work for anyone interested in trying to understand both the manifold philosophical challenges of Western scientific technology and the contemporary rise of China on the world-historical scene.

As a native of Hong Kong, Hui grew up fluent in Chinese and English. After earning a degree and working in computer science, he moved to Europe, where he became fluent in French and German (picking up a little Greek and Latin along the way). He earned a PhD under the contemporary French philosopher Bernard Stiegler, whose three-volume Technics and Time both deepened and challenged Heidegger. Drawing on his technical and philosophical expertise, Hui’s doctoral dissertation was an ontological examination of the historical development of computer programing, published in 2016 as On the Existence of Digital Objects, with a foreword by Stiegler. The Question Concerning Technology in China, dedicated to Stiegler, is an even deeper engagement with the fraught technoscientific mutation of our world.

The book is an effort to rethink technology, premised on and developing the idea that nature is not some one thing: that nature is co-constructed and therefore variable, and that its variability is reflected in historically and culturally diverse technologies. Though scientists posit a universe that is always the same underlying their theoretical and experimental discoveries, the discoveries themselves present an ever-shifting view of natural reality. Even Needham admitted that Chinese culture involved different cosmologies—ideas about the origins of the universe and humanity’s place in it—than the modern West. Against this background, Hui formulates a question: “If one admits that there are multiple natures, is it possible to think of multiple technics, which are different from each other not simply functionally and aesthetically, but also ontologically and cosmologically?”

As one illustration of the differences, Chinese immigrants to the United States during the California gold rush introduced the human-powered water wheel to pump water into dry streambeds for placer mining. This practice depended on a solidarity among workers that was difficult for individualist Americans to practice: someone had to be paddling while others did the panning, and all shared equally in the mining profits. A typically American, individualist social ontology has technological consequences.

To venture a more speculative instance, Chinese calligraphy is not just aesthetically different from the Latin alphabet. It enacts a distinctive way of understanding reality and what Heidegger called “being-in-the-world.” Chinese calligraphy, as a logographic system in which written characters represent things, privileges realities in the world over abstract human sounds. The Chinese art of writing can be understood as engaging the actuality of the world while it serves as well as a form of self-cultivation.

How did these very different conceptions of the world and human agency in it arise? Yuk Hui notes the different mythological accounts of the origin of “technics,” or the interplay of technologies with social circumstances. In both the Greek and Hebrew traditions of the West, technics is culturally conceived as a kind of opposition to the gods or God. Prometheus stole the technology of fire to give to deficient humanity, and was punished with eternal torment by Zeus. The Tower of Babel was a technical effort to reach heaven in defiance of God, and humanity was punished with a multitude of languages so that people could no longer understand one another.

As Hui points out, the Chinese mythopoeic account of technics is markedly different. There was no Promethean theft from nor human rebellion against the divine. Instead, there were three mythological leaders of ancient tribes: the half-human, half-snake female Nüwa; her half-dragon, half-human brother-husband Fuxi; and the divine farmer and later kitchen god Shennong. All three collaborated to create humans and to provide them with such tools as fire. Humans are seen as situated between, and natural combinations of, heaven and earth. There is no rebellion of humans against heaven; there is only working with earth and heaven to cultivate and take aesthetic, ritual, and common pleasure in the world.

Hui coins the term “cosmotechnics” to describe “the unification between the cosmic order and the moral order through technical activities” that is entailed by any mythology. Importantly, this concept connects cosmologies with cultural beliefs about what constitutes a good life—precisely what Heidegger thought modern technology had destroyed.

In his effort to valorize the Chinese inventive tradition, Needham asked what has become “Needham’s Question”: Why did China, which up until 1500 was more scientifically and technologically advanced than Europe, fail to develop modern technoscience? Yet he failed to notice the mythological difference that Hui identifies, instead attributing what Needham considered the failure of China to a set of historically contingent conditions: geographical, political, economic, and religious. Additionally, Hui might have added, Heidegger never considered the implications of a simple difference in Western engineering, which emerged out of the military, in contrast to the way Chinese qi (technics) and gong cheng (engineering) are more closely bound up with farming and a stabilized, sedentary life.

Following an extended introduction to his thought project, Yuk Hui divides his reflection into two parts. Part one, “In Search of Technological Thought in China,” explores the relationship between qi and dao (cosmic order) in the 3,000-year history of Chinese culture. This extended dialogue brings a new appreciation to Chinese philosophy in its many permutations across the centuries—in Daoism, Confucianism, and Buddhism—and promotes conversation with major philosophical traditions and thinkers of the West, from Plato and Aristotle on. It is, quite frankly, an original, provocative achievement that any future effort to consider technology in a global context will need to take into account.

Part two, “Modernity and Technological Consciousness,” draws on Hui’s presentation of traditional Chinese philosophy to reconsider both the philosophy of technology in the West and to offer alternatives to the contemporary tendency in China to follow the West too quickly. Hui’s challenge is not just to the West; it is also to China.

Repeatedly, Yuk Hui highlights mirror-image issues: In the West, the philosophical acidity of technoscience, which leaches away connections to particularities of place, tends to dilute social consensus about the good in favor of the pursuit of modern science itself (at least among the scientific few) or individualist freedoms (among the nonscientific many). Is it any accident that Ayn Rand’s extreme libertarianism thrives among engineering entrepreneurs financed by venture capitalists?

In China, a rich culture that became unable to defend itself against European imperialism—weaponized by advances in science and technology—has struggled since the Ming Dynasty to find a way to preserve “Chinese learning for essence, Western learning for application.” The Chinese effort deserves more consideration than it currently receives, Hui suggests, in either the West or China. Hui clearly wants to engage those who are trying to think about these issues, especially social critics of the contemporary world and philosophers of science and technology.

Other than philosophers, however, who might profitably place themselves in Yuk Hui’s audience? Anyone, I suggest, concerned with the relationship between science and politics or interested in the influence of China on the trajectory of globalization.

Most Western scientists and engineers—and even many Westerners tout court—will agree more with Needham’s view of science as devoted to understanding universal laws and the best account of nature that humans possess, than with Heidegger’s deep questioning of such a view. As a result, the majority of readers are likely to find Hui’s sustained engagement with Heidegger’s ideas, at best, strained. At the same time, scientists and engineers, along with those who study or formulate science policy, increasingly recognize problematic features in the relationship between science and society. Annual meetings of the American Association for the Advancement of Science, for example, are replete with discussions of the difficulties involved in science communication, distrust and rejection of scientific knowledge, and attacks on public support for scientific research.

As challenges such as climate change well illustrate, technoscientific prowess (allied with capitalist economic formations) can have unintended consequences that sponsor new visions of the natural. Some political criticisms of climate change research even enroll (unwittingly) science studies theories of the social construction of reality to marginalize evidence-based policy-making. In other words, insofar as scientific knowledge is constructed through both social and technical means—rather than simply “discovered”—powerful interests have claimed a right to construct things to their liking.

In response, Bruno Latour, a French science studies scholar who has radically challenged the self-image of scientists, has recently undertaken his own defense of science. His mission, as he outlined in a recent interview in Science, is “to regain some of the authority of science [which] is the complete opposite from where we started doing science studies.” As part of this attempt—and in response to the mutation of the human condition he explores most broadly in his 2017 book, Facing Gaia: Eight Lectures on the New Climatic Regime—Latour has been practicing what he calls a philosophically diplomatic engagement with the world’s plurality of modes of existence.

In 2017 this diplomatic engagement took Latour to Shanghai and Beijing, where he invited Chinese colleagues to help address the challenge of climate change. At one point, Yuk Hui was among the participants. Yet as an observer at some of these gatherings, I found them singularly unsatisfying. It was not clear that Latour had done sufficient homework to avoid the projections of Western orientalism or to parry Chinese occidentalism. Latour had trouble making clear what he wanted from China, but he seemed to be asking Chinese intellectuals to contribute to his own project of “re-describing” the world (to better investigate the connections between nature and society), without appreciating the deeper distinctiveness of traditional Chinese culture or the ways contemporary Chinese often want to imitate the West.

Too often when Latour asked interlocutors for Chinese perspectives that could contribute to what Hui might call cosmotechnical living amid ongoing mutations of nature and culture, there was either talk at cross purposes or dead air time. Latour wanted things from the Chinese scholars that they were not prepared to offer, while Chinese scholars were more interested in Latour’s actor-network theory—the academic work on complex social relationships for which he is most famous—than his concern for global climate change.

For anyone who wants to reflect deeply and seriously about a world experience that includes globalized social destabilization, biodiversity loss, nuclear proliferation, climate disruption, and expansion in artificial intelligence, there is more to the grand challenges of the twenty-first century than either Western or Eastern technical resources alone can easily address. Globalization demands thinking and discourse across historical, cultural, and technological categories. Reading Yuk Hui can be a stimulating start but is not enough. As Hui himself indicates, his aim is simply to open doors to ideas that we all need to take more seriously.

Self-Driving Cars: How Soon Is Soon Enough?

Some observers tout autonomous (self-driving) vehicles by claiming that a computer is “simply a better driver than a human,” as a recent Time magazine article put it. This is not true now and may never be true. Safe driving, the kind carried out on a daily basis by millions of people, is a highly skilled practice honed over time through years of practical experience. Perhaps nothing better illustrates this fact than the recent tragedy in Tempe, Arizona, where a pedestrian crossing a street outside of a marked crosswalk was struck and killed by an Uber vehicle operating in autonomous driving mode. An alert human driver would have had no problem in identifying the potential danger.

Autonomous vehicles are being developed to address unsafe practices—that is, driving under the influence of a host of impairments. The human toll from unsafe driving is enormous, resulting in roughly 37,000 deaths every year in the United States, so the massive high-tech effort to produce self-driving cars is justified. But it must be done in recognition of and compatible with safe driving practices and the people who value the freedom—and accept the responsibility—that driving entails.

Driving is a complex social process whereby individuals come to understand the driving environment through repetition and through experience with other drivers and conditions. As such, even when a number of unforeseen contingencies arise, drivers who are fully engaged are able to adjust and navigate the road safely. But not everyone is equally skilled at driving and avoiding crashes. The figure below illustrates differences based on age and gender. Young males and the older generation are most vulnerable to fatal automobile crashes.

All age groups and genders, however, are susceptible to a collection of impairments that can make driving less safe. In addition to testosterone-fueled teenage boys and cognitively impaired seniors, there are inebriated revelers, fatigued drivers, distracted teens and millennials texting, and individuals with mood swings, including rage. Better technology to counter the effects of unsafe driving is most welcome.

Efforts to improve the safety of the vehicle, of course, have been longstanding and continuous. Seatbelts, airbags, and vehicle design have all advanced safety over the decades, even if the automakers themselves haven’t always been at the forefront of the advances, as illustrated by Ralph Nader’s epic battle with General Motors, documented in his book Unsafe at Any Speed. Validation, perhaps, took place in 2013 when the US Centers for Disease Control and Prevention (CDC) declared motor vehicle safety one of the “ten great public health achievements of the 20th century.” To justify this award the CDC noted that the annual death rate attributable to motor-vehicle crashes had declined from 18 per 100 million vehicle miles traveled in 1925 to 1.7 per 100 million vehicle miles traveled in 1997—a 90% decrease. But this statistical jujitsu cannot protect society from the enormous carnage that still takes place on roads and the realization that society must do better.

The current level of bipartisan political support for autonomous vehicles in this age of political confrontation is remarkable, bridging political parties and administrations, legislators and agency officials. The downside of this enthusiasm is a kaleidoscope of state and local regulations covering autonomous vehicle testing and operation, creating confusion for the car industry. Establishing a federal standard would streamline commercialization and provide automakers with greater certainty. Toward this end, the US House of Representatives in September 2017 passed the Self-Drive Act, calling on the Department of Transportation to promulgate new regulations and safety standards under which self-driving cars could operate. The legislation had earlier been cleared by the House Committee on Energy and Commerce by a 54-0 vote. The bill would also permit up to 100,000 autonomous vehicles to be tested on public roads even if the testing violated state and local laws or regulations. The Senate has been holding hearings on the subject, but is reconsidering legislation in light of the Tempe accident and other Tesla accidents that occurred in vehicles operating in “autopilot” mode.

Independent of federal legislation, in September 2017 the National Highway Traffic Safety Administration published what it called a policy framework for development of automated driving systems, A Vison for Safety, which set forth its desired path toward putting self-driving vehicles on the roads. The document builds on an initial effort undertaken by the Obama administration and seeks to assure automakers that neither federal, state, nor local government would produce a plethora of regulations inhibiting testing and commercialization.

Although a strong case can and is being made for the public benefits derived from the move to autonomy, the more powerful impetus is the interests of high-tech Silicon Valley firms. As noted, driving is a repetitive experience, and thus can, at least in theory, be programmed to avoid human errors. The transformation of automobiles from mechanical devices to programmed machines run by software has been under way for some time as various computer-aided features have been added to cars. Autonomous vehicles are simply the logical extension of this trend.

Existing automakers are brought into this arena through the fear of disruption of their preferred business model. With the exception of Tesla, Silicon Valley firms do not seek to become independent automakers. Mutual interests, therefore, have led to the development of alliances between information technology firms and existing automakers.

And lest it be forgotten, the primary goal of computer and software companies has been to attract more eyeballs to the many screens sold to consumers. Most states have passed legislation to prohibit texting while driving, recognizing the dangers inherent in divided attention. Autonomous vehicles have the potential to turn this completely around by prohibiting human driving while texting—a result that would please Silicon Valley.

A recent study by the Brookings Institution listed 160 autonomous vehicle business deals struck between 2014 and 2017, with a total value of approximately $80 billion. It projected continued rapid growth. Presumably there are early-mover competitive advantages to bringing autonomy along quickly. The initial focus will be on creating autonomous vehicles for commercial ride-sharing fleets. Waymo, the autonomous vehicle program from Google, plans to have self-driving vehicles on Arizona roads in 2018. General Motors has announced its goal of having fully autonomous cars operating in major cities by 2019. Ford’s plan for introducing commercial fleets is scheduled for 2021. In fact, a host of global automobile companies and their affiliated high-tech partners are targeting the 2020-21 time frame for commercialization.

A case can be made that given the enormous number of casualties taking place on roads today, full vehicle automation should be supported and fostered as quickly as possible. A recent Rand Corporation report concluded that hundreds of thousands of lives could be saved if autonomous vehicles were quickly introduced, even though they are clearly not yet accident-proof. I believe a more compelling case can be made for taking a more measured and incremental approach to automation. The strongest argument for a go-slow—or at least go-slower—approach is that the public has yet to join the automation bandwagon. The dash to self-driving is fundamentally driven by competition between giant corporations, not by public safety advocates. Drivers are not demanding that they be allowed to take their hands off the wheel. Perhaps they should, but surveys have shown a public that is either skeptical or fearful of safety claims based on advanced automation. What this means is that the automobile industry faces the potential of a severe public backlash when autonomous vehicles come to market.

A 2017 survey by the Pew Research Center revealed that more than half of the US public is worried about the development of driverless cars. A recent poll conducted by the American Automobile Association revealed even more of the public fearful (nearly 75%) after the well-publicized accidents cited above. To be sure, there are some groups—men, the better educated, and millennials—who are less worried than others. But the differences are more of degree than kind. The majority of drivers share a fear of being in a self-driving car as well as a reluctance to relinquish the joy of driving.

Fear is not irrational. As evidenced by the Tempe tragedy, crashes will take place in the early days of automation. Fully autonomous vehicles come with a host of advanced sensing technologies, such as GPS, radar, and lasers. Yet the ability of the sensors to operate perfectly in all weather conditions remains problematical. Most significant, the software or algorithms on which driving will be based are still developing, and further testing is needed to build in responses to a wider variety of situations. A fully supportive public would respond to unforeseen accidents with a shrug and an expression of faith that things will improve in the future. The skeptical public of today, however, is likely to experience a strong case of “betrayal aversion,” a repudiation of the technology as oversold and underperforming. Such a reaction could set back commercialization of autonomous vehicles for decades. It could also feed into “techlash,” a reaction against the encroachment of large Silicon Valley firms and artificial intelligence into people’s lives. With this in mind, several public safety organizations, including Consumer Reports, the Consumer Federation of America, and Advocates for Highway Safety, have come out against congressional legislation that hastens the commercial development of autonomous vehicles in the absence of proven safety. And although politicians are now eager to extol the wonders of technological development, it will be interesting to see if they now leap to the industry’s defense in light of the recent crashes in Tempe and elsewhere.

Developing a more measured approach to commercialization of autonomous vehicles requires cognizance of the differing stages of autonomous driving. SAE International, a professional and standard-setting association of engineers and related technical experts in mobility industries, has produced a widely published chart setting forth six separate stages of autonomy: no automation (level 0), driver assistance (level 1), partial automation (level 2), conditional automation (level 3), high automation (level 4), and full automation (level 5). The levels are meant to convey a spectrum of human-technology interaction with full human control over vehicle safety at the beginning and full technology control at the end.

Surveys have shown that the public is supportive of technologies that aid drivers in the safe operation of a vehicle. Consumer Reports magazine recently found high satisfaction levels (65-80+%) with driver-assistance technologies such as adaptive cruise control, forward-collision warning, lane-keeping assistance, blind-spot warning, and emergency braking. Level 1 automation entails the introduction of one of these technologies into the vehicle. Level 2 entails the introduction of virtually all of these technologies and is available as an option in most new cars. But at level 2, control of the vehicle still remains the responsibility of the driver.

The automobile industry could make these technologies more foolproof and provide them as standard equipment in all makes and models in fairly short order, thereby reaping near-term safety benefits and, at the same time, building confidence among the public in a longer-term transfer of vehicle control. These active safety features can serve as logical stepping-stones to greater vehicle autonomy. A survey by the Alliance of Automobile Manufacturers has shown that drivers whose vehicles have at least two advanced driver-assistance technologies view the move to full autonomy more favorably than those whose vehicles have none. Expanded introduction of driver-assistance technologies could also have significant public safety implications, as several studies have forecast that their widespread use could yield a reduction of nearly 10,000 fatalities per year. At current and traditional rates of introduction and public acceptance of these safety features, however, it may take another 25 years before these technologies are present in virtually all registered vehicles.

Significant public safety benefits arise, therefore, with initial moves to automation long before reaching levels 3-5, where control over driving becomes largely the responsibility of the vehicle and where public concerns become manifest. In order for the public to gain more confidence in the movement to autonomous vehicles, it may be useful for public officials to focus on levels 1 and 2, with progress toward further development being contingent on greater proof of safety claims.

Current automotive plans, however, envision level 3 and level 4 driving in the near future. Level 3 would allow hands-off driving, with the proviso that humans remain fully attentive and prepared to take over in the event the vehicle sends an alert message. Unfortunately, we have already seen fatal crashes in level 2 when driver attention wanes over time, and in level 3 testing in Arizona. Moving directly to commercial level 3 would exacerbate this tendency and raise doubts as to whether drivers can reasonably be expected to perform smooth driving transitions under split responsibilities.

Level 4, taking place without humans in the driving seat, can conceivably work safely in carefully restricted settings. This is where companies eyeing commercial fleets, offering ride-sharing without a paid driver, are eager to begin. Dangers arise, however, when profit-seeking firms seek to expand the geographical boundaries of their operation, as they inevitably will.

Discussion thus far has revolved around the question of when drivers should be able to purchase and operate autonomous vehicles. Yet a more fundamental question is if these vehicles will ever be open to personal ownership. Some advocates of autonomous vehicles point out that personally owned automobiles sit idle 95% of the time. They envision a wholesale transfer of vehicles from private to communal or commercial ownership. On an efficiency basis alone, shared vehicles, used more intensively, would make sense. It may also be the only way to pay for what could be the high production costs of these vehicles. It could also work to reduce congestion and bring about desired energy savings and carbon dioxide reductions. And certainly it is consistent with what is perceived to be the younger generation’s reduced ardor for driving, compared with that of previous generations.

Still, the removal of both operational control and ownership of vehicles would constitute a momentous change in people’s daily lives—and would need public validation before proceeding. The sense of personal freedom and identity that accompanies private ownership of vehicles has not disappeared completely. A recent poll indicated that four out of five US residents believe that people should always have the option to drive themselves. If autonomous vehicles will serve only a limited market, such as older adults or the infirm, safety benefits would be diminished considerably. Any attempt by government to impose or even lead “transit as a service,” however, would face fierce opposition. Since level 5 driving is still many years, or even decades, away, society has time to debate the issue productively.

The movement toward autonomous vehicles, either privately or communally owned, is a hugely important development with the potential to save hundreds of thousands of lives in the future. But it should not be moved forward through technological determinism or in a hasty dash for commercial profit. A wary public, which largely engages in safe driving and may be reluctant to cede control to machines, will fixate on any technological failures to the detriment of the long-term viability of this movement to address unsafe driving. A slow-but-sure strategy to commercialization makes the most sense for now.

Indicators of R&D Tax Support

Investment in research and development (R&D) is a key driver of innovation and economic growth. In addition to funding R&D within public institutes or universities, governments within the Organisation for Economic Co-operation and Development (OECD) and beyond actively incentivize R&D by private companies, which conduct the largest share of R&D in the OECD area.

Evidence shows that tax incentives have become an increasingly popular instrument to promote business R&D investment, displacing direct funding such as grants and public procurement. In 2017, 30 out of 35 OECD countries gave preferential tax treatment to business R&D, up from 16 OECD countries in 2000. Other major governments such as Brazil, China, and the Russian Federation also provide R&D tax incentives.

OECD monitors how the design of R&D tax incentives may influence business decisions to increase R&D efforts or move R&D activities into a given country. Because each country has its own distinctive tax system and relief provisions, OECD analysis converts each tax structure into a comparable R&D tax subsidy indicator that can be analyzed alongside the resulting cost to the government.

The OECD R&D tax incentive indicators show the estimated cost of providing R&D tax support to businesses. This is a combination of foregone tax revenues and, in some cases, payable subsidies when companies do not have sufficiently large tax liabilities to benefit from earned allowances or credits.

Despite the growing literature on the impact of different forms of support for business R&D, there is no simple, widely applicable answer to the question of what are the right volume of total support and the appropriate mix of tax and direct support within countries. Tax incentives are not equally beneficial to all types of R&D performers. The impact of tax incentives may depend on the nature and structure of a country’s innovation system, as well as current conditions such as the business cycle. Indicators such as those presented in this report help provide an illustrative benchmark against which countries can compare themselves and bring about relevant dimensions that raise follow-on questions and avenues for analysis.

All of the data cited here are from the OECD R&D Tax Incentive database. Much more detailed data and analysis on the incidence and impact of tax incentives as business innovation policy tools are available on the OECD Science, Technology, and Innovation website on R&D tax incentives at http://oe.cd/rdtax.

By adding tax support to direct government research spending, it is possible to provide a more complete picture of the full extent of government support for business R&D across OECD, EU, and other major economies. In 2015, the Russian Federation, Belgium, France, Korea, and Hungary provided the most combined support for business R&D as a percentage of gross domestic product (GDP). The average rate of tax support in the OECD area—including countries that do not provide this type of support—is close to 0.09% of GDP. Tax support accounts for 5.4% of business R&D (BERD) in the OECD area.

Whereas direct support is by and large determined by government officials, the share of R&D tax support accounted for by small and medium-sized enterprises (SMEs) tends to be more closely aligned with the SME share in BERD, confirming that tax incentives are generally a more demand-driven instrument than direct government support. The SME share in tax support exceeds the share of direct funding in countries such as Austria, Canada, France, the Netherlands, Norway, and the United Kingdom, which offer refundable R&D tax incentives that target smaller R&D performers.

The relative importance of tax incentives is increasing among a majority of countries for which data are available. The United States makes relatively limited use of tax incentives, but use has increased slightly. Canada, which had been making extensive use of tax incentives, shifted its emphasis somewhat toward direct support. Mexico abolished its R&D tax credit during this period but reintroduced it in 2017. Germany and Switzerland, two countries with high business R&D intensity despite not offering tax incentives, are considering introducing tax relief for R&D.

Between 2000 and 2017, implied marginal tax subsidy rates for R&D increased significantly in the OECD area for both SMEs and large firms, regardless of their profit situation. Tax subsidy rates are measured as the difference between one unit of investment in R&D and the pre-tax income required to break even (B-index). They increased on average from approximately 0.06 to 0.17 in the case of profitable SMEs (0.04 to 0.15 for loss-making SMEs), and from approximately 0.04 to 0.14 in the case of large profitable firms (0.03 to 0.12 for large loss-making firms). Throughout this period, SMEs faced on average a higher marginal tax subsidy rate than large firms, comparing either profit-making or loss-making firms.

An indicator of the relationship between government support for R&D and business R&D performance shows that changes in total measured government support appear to explain around one tenth of the observed variation in BERD intensity. Changes in R&D tax support alone explain nearly 6% of the observed variation in the data.

These descriptive indicators raise research policy questions about how to measure the causal effects of support and improve the efficiency of existing instruments. They also call for better measurement of other forms of indirect support business R&D and innovation that still go unaccounted for.

Fernando Galindo-Rueda is a senior economist and Silvia Appelt is an economist in the Economic Analysis and Statistics Division of the OECD Directorate for Science, Technology, and Innovation. Ana González-Cabral is a junior economist at the OECD Directorate for Science, Technology, and Innovation and the OECD Centre for Tax Policy and Administration.

The Limits of Dual Use

Research and technologies designed to generate benefits for civilians that can also be used for military purposes are termed “dual use.” The concept of dual use frames and informs debates about how such research and technologies should be understood and regulated. But the emergence of neuroscience-based technologies, combined with the dissolution of any simple distinction between civilian and military domains, requires us to reconsider this binary concept.

Not only has neuroscience research contributed to the development and use of technology and weapons for national security, but a variety of factors have blurred the very issue of whether a technological application is military or civilian. These factors include the rise of asymmetric warfare, the erosion of clear differentiation between states of war abroad and defense against threats “at home,” and the use of military forces for homeland security. It is increasingly difficult to disentangle the relative contributions made by researchers undertaking basic studies in traditional universities from those made by researchers working in projects specifically organized or funded by military or defense sources. Amid such complexity, the binary world implied by “dual use” can often obscure rather than clarify which particular uses of science and technology are potentially problematic or objectionable.

To help in clarifying matters, we argue that policy makers and regulators need to identify and focus on specific harmful or undesirable uses in the following four domains: political, security, intelligence, and military (PSIM). We consider the ways that research justified in terms of socially constructive applications—in the European Human Brain Project, the US BRAIN initiative, and other brain projects and related areas of neuroscience—can also provide knowledge, information, products, or technologies that could be applied in these four domains. If those who fund, develop, or regulate research and development (R&D) in neuroscience, neurotechnology, and neurorobotics fail to move away from the dual-use framework, they may be unable to govern its diffusion.

Multilateral treaties and conventions most significant for developments in the neurosciences are the Biological Weapons Convention and the Chemical Weapons Convention. These conventions’ review conferences, held five time a year, fail to capture the rapidly evolving advances in the neurosciences, and they lack adequate implementation and oversight mechanisms to keep up with emerging capabilities. Beyond disarmament treaties, countries use various export-control instruments to regulate the movement of dual-use goods. However, this is constructed around moral and ethical interpretations of what constitutes good (dual use) and bad (misuse). Dual-use regulations typically allow for cross-border trade if technologies are intended for good uses, albeit often under substantial export controls, while prohibiting trade in technology if there is a significant potential for terrorist or criminal use.

But the idea of good and bad uses raises intractable issues of interpretation and definition concerning applications of research. Indeed, governments and regulatory institutions even differ in their approaches to defining and delineating dual use in research. Funding agencies sometimes prohibit research that is intended to be used for military applications. This is true for the Human Brain Project, for example. However, regulating intentions can hardly prevent research that is benign in intent from later being used for hostile purposes.

Regulators have sought to specify more precisely what kinds of R&D should be prohibited in which contexts. For example, in the context of trade, the European Commission, the politically independent executive arm of the European Union (EU), states that “‘dual-use items’ shall mean items, including software and technology, which can be used for both civil and military purposes.” Horizon 2020, the EU program that funds the Human Brain Project, is more specific, requiring applicants for funding to ensure that “research and innovation activities carried out under Horizon 2020 shall have an exclusive focus on civil applications,” and they are required to complete an ethics checklist to demonstrate that they comply with this requirement.

The Office of Science Policy of the US National Institutes of Health defines “dual use research of concern” to mean “life sciences research that, based on current understanding, can be reasonably anticipated to provide knowledge, information, products, or technologies that could be directly misapplied to pose a significant threat with broad potential consequences to public health and safety, agricultural crops and other plants, animals, the environment, material, or national security.” This definition is meant to be more specific and to help policy-makers pay attention to particular problematic research and technologies, yet science or technology that might be “misapplied” might also be legitimately justified for national security purposes. The criteria used to distinguish between legitimate use and misuse or misapplication are inevitably matters of context that depend on how the technologies will be applied across the four domains we have identified.

Political applications

By political applications we refer to the use of neuroscience or neurotechnologies by government authorities to govern or manage the conduct of individuals, groups, or populations—for example, by changing or manipulating attitudes, beliefs, opinions, emotions, or behavior. We will focus on one example from the developing field of “neuropolitics”: the uses of neuroscience research into explicit and implicit attitudes and decision-making. This is especially timely given the Cambridge Analytica scandal, where personal data collected by the political consulting company were used to build psychological profiles and target political advertising in the run-up to the US presidential election.

The political scientist Ingrid Haas has said that the focus of political science has traditionally been on understanding the behavior of groups, but new methods in the neurosciences can help to understand how individual political behavior relates to the behavior of a collective. Here, as elsewhere, the hope is that if one understands the neurobiology of decision-making, and especially the unconscious factors shaping decision-making, it will be possible to develop more effective interventions. Brain imaging technologies such as functional magnetic resonance imaging (fMRI) have been used to investigate the ways in which people form and revise attitudes and make evaluations. For example, in 2006 Kristine M. Knutson and colleagues used fMRI to measure in human subjects the activation of the part of the brain known as the amygdala—understood to be active in the processing of emotion—to study attitudes toward Democrat and Republican politicians. Further studies by Woo-Young Ahn and colleagues in 2014 claim that brain imaging can accurately predict political orientation, such as the differences between liberals and conservatives in the United States.

In contrast to imaging technologies, recent developments in “noninvasive” brain stimulation—such as transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS)—can be used to directly alter or disrupt brain activity. For example, in 2015 Roberta Sellaro and colleagues examined the neural correlates of implicit attitudes by applying tDCS to subjects’ medial prefrontal cortex while asking them to “categorize in-group and out-group names and positive and negative attributes.” The researchers found that “stimulation decreased implicit biased attitudes toward out-group members,” concluding that tDCS could be used to reduce prejudice “toward members of social out-groups.”

Despite the limitations of these imaging and stimulation technologies—for example, it is not known whether the effects occur in a particular region of the brain or whether other regions are also being stimulated—research investments have already been made into the further development of noninvasive brain stimulation technology, such as in the EU-funded Hyper Interactivity Viability Experiments (HIVE) and in the US BRAIN initiative. Although brain imaging and stimulation technologies were originally envisioned to aid in the diagnosis and treatment of conditions such as depression and schizophrenia, there are clearly potential political uses of brain imaging and brain stimulation technologies in devising more effective ways in which authorities can shape or manage the opinions and attitudes of citizens. Many of these will, of course, merely give a neurological basis to the ways in which authorities have long tried to shape the decisions of their citizens (for example, in antismoking or healthy eating campaigns), which may raise few ethical concerns. But one can imagine less benign uses, such as in shaping attitudes to particular minorities, or manipulating choices of political parties in elections, or in managing protests or challenges to political regimes.

Security applications

There are many ways in which neuroscience and neurotechnologies can be deployed in the name of security. We focus here on the use of calmatives for national security—biochemical agents that act on the central nervous system to incapacitate those exposed to them. These neurotoxic weapons have become possible only because of advances in neuroscience, pharmacology, and physiology.

International discussions on unconventional weapons have to date focused on chemical, biological, and nuclear weapons. But a number of government agencies, private organizations, and nongovernmental organizations have expressed concerns that calmatives and other new agents that act on the central nervous system seem to fall outside the scope of the Biological Weapons Convention or the Chemical Weapons Convention, and might be permitted under Article 11d of the latter convention, which excludes from its prohibition of toxic chemicals their use for “law enforcement including domestic riot control purposes.” Yet calmatives can be lethal, as tragically demonstrated by the Russian theater hostage crisis in 2002, where many of the rescued hostages died as a result of fentanyl overdose.

Israel has a history of using incapacitating chemical agents, such as capsicum pellets, in policing. And although Israel receives funding through the EU’s Horizon 2020 for its participation in the Human Brain Project and other neuroscience research, the country is not a party to either the Chemical Weapons Convention (signed, but not ratified) or the Biological Weapons Convention (not signed or ratified). Very limited public information exists about whether Israel continues to develop and test chemical and biological weapons, but Alastair Hay and colleagues demonstrate that because capsicum pellets used against protestors in 2005 had more severe symptoms than other reported injuries from similar chemical agents, it is plausible to assume that collaborations between the Israel Institute for Biological Research and the Ministry of Defense includes continued R&D on neurotoxin-based weapons.

The specific issue of the development of neurotoxic chemicals for domestic security might be addressed by the extension or modification of existing international treaties and conventions. As one manifestation of such concerns, a joint paper issued on behalf of 20 countries (including the United States and the United Kingdom) in 2015, titled “Aerosolisation of Central Nervous System-Acting Chemicals for Law Enforcement Purposes,” argued that the Chemical Weapons Convention should not offer an exemption for law enforcement for use of neurotoxins in security applications.

However, this approach would not address the issues raised if current research on “neural circuits”—one focus of the US BRAIN Initiative—opened the possibility of their modification for security purposes. It is clear that we need new ways of thinking about how these and similar security uses of neuroscience and neurotechnology can be evaluated and regulated.

Intelligence applications

Neuroscience research offers an array of potential applications in the intelligence domain—for example, in identifying criminal suspects, for lie detection, and in surveillance. US neuroscience research funded through the Defense Advanced Research Projects Agency (DARPA), for example, has focused on technologies that might accomplish “brain reading” or “mind reading” to identify malicious intent. A relatively recent technique, near infrared spectroscopy (NIRS), operates by using changes in blood oxygenation in particular regions as a proxy for brain activity. The technique fMRI operates on the same principle, but instead of having the subject lie immobilized in a large machine, NIRS involves the subject donning a kind of helmet with multiple sensors that use infrared light than can penetrate the intact skull to measure changes in blood oxygenation. The initial uses of NIRS were medical, but as early as 2006, Scott Bunce and colleagues, supported by funds from DARPA’s Augmented Cognition Program, the Office of Naval Research, and the Department of Homeland Security, pointed out its many practical advantages over other methods of brain imaging, because it enabled the subjects to move about and carry out tasks in a relatively normal environment. Bunce suggested that because of these advantages, NIRS had significant potential in the detection of deception and other investigations that needed to be done in clinical offices or environments other than laboratories.

Neuroscientists, such as John-Dylan Haynes and colleagues, also claim that machine learning coupled with fMRI is beginning to be able to identify specific thoughts or memories in human subjects. Although this work is currently in its infancy, developments in such brain reading technologies will have obvious appeal to intelligence and surveillance organizations—for example, in determining when an individual intends to commit a terrorist act or whether a suspect might have important security-related information that they are unwilling to divulge voluntarily or under interrogation. But such potential applications raise not only technical questions of reliability but also ethical questions of violation of “neural privacy” in cases where information derived from involuntary brain scans is used as the basis of government action against individuals.

More effective lie detection also has obvious application to the intelligence domain. Whereas the conventional polygraph monitors physiological indicators such as blood pressure and pulse, fMRI-based technologies developed by Cephos Corporation and NoLie MRI claim that specific patterns of brain activity can identify when a suspect is lying or has a memory of particular words or scenes linked to an actual or intended crime. However, the reliability of such technologies is not yet well established. It is difficult to extrapolate from laboratory-based studies in which individuals are instructed to lie or tell the truth to real-life situations where the very fact of being accused of malicious intent, or knowledge of guilt, generates unknown patterns of brain activity, and where those who are genuinely guilty are likely to employ multiple techniques to disguise their deception. Although the results of such technologies may ultimately be inadmissible in the courtroom—as with the polygraph—they may well find use in intelligence applications such as interrogation and investigatory procedures.

Another intelligence-related application of neurotechnology (which also may find security-related uses) is in facial recognition technologies, which are now being adopted by many nations for intelligence gathering—although ethical challenges of privacy and consent remain unaddressed. Leading companies in the field of deep learning, such as Google DeepMind, have developed algorithms that are effective on static images but not on moving ones. Large-scale brain mapping projects are expected to lead to advances in neuromorphic computer design and “smarter” algorithms that can improve intelligence gathering-capabilities, such as facial recognition in videos, thus greatly enhancing surveillance capacities and potentially enabling additional applications of enhancing the targeting capabilities of autonomous or semiautonomous weapon systems.

Military applications

There are many military applications of contemporary developments in neuroscience and neurotechnology, but we focus here on warfighter enhancement through cortical implants and on noninvasive brain-computer interfaces (BCIs). Drugs originally used to improve cognition in patients with schizophrenia and dementia have been used to increase the cognitive performance of soldiers and to combat sleep deprivation. However, much current neuroscience and technology research on warfighter enhancement relates not to drugs, but to prosthetics. DARPA has for more than a decade invested heavily in this area, in particular aiming to develop cortical implants that will control prosthetic limbs fitted to soldiers injured in combat. As described in a 2006 paper by Andrew B. Schwartz and colleagues, one goal is to develop “devices that capture brain transmissions involved in a subject’s intention to act, with the potential to restore communication and movement to those who are immobilized.”

New approaches are now being tested for transmitting signals wirelessly to enable a subject to move a prosthetic limb with thought alone. The technique, which uses a cerebral implant that records the electrical activity of populations of neurons in motor cortical areas, was used initially with monkeys and now is being tried in humans by Jennifer L. Collinger and colleagues. Although such advances address important clinical issues of restoring function to those who have lost it by military or other injury, they also have potential warfighting applications. In theory at least, a single individual, equipped with a cortical implant, could wirelessly control neuroprosthetic devices in one or many robots situated in distant locales. And although it currently requires invasive surgery to install the cortical implants, R&D funded by DARPA is under way to miniaturize the chips and simplify their insertion.

In addition, Michael Tennison and Jonathan Moreno in 2012 described how a project within DARPA’s Augmented Cognition Program that used noninvasive brain-computer interfaces “sought to find ways to use neurological information gathered from warfighters to modify their equipment.” For example, BCIs would be utilized in a so-called “cognitive cockpit” that would aim to avoid information overload by adjusting the ways information is displayed and conveyed to the pilot. DARPA has also funded research using noninvasive BCIs in its Accelerated Learning program, which aims to develop new training methods to accelerate improvements in human performance, and in its Narrative Networks program to quantify the effect of narratives on human cognition and behavior. And the agency’s Neurotechnology for Intelligence Analysts and Cognitive Technology Threat Warning System programs have utilized noninvasively recorded “target detection” brain signals to improve the efficiency of imagery analysis and real-time threat detection. The threat-warning system is intended to integrate a range of sensory data from military personnel on the ground, much of which might be below their level of awareness, and then to alert them about potential threats.

Noninvasive BCIs may find many civilian uses as well. For example, just as the devices might aid injured soldiers, they are showing increased potential in the broader population for using brain signals from motor cortex areas to control prosthetic limbs. BCIs have been developed to enable individuals to control the cursor of a computer so that, for example, individuals who have lost the use of their limbs may be able to communicate by writing and, more recently, to enable some individuals apparently unable to communicate at all (considered to be in a persistent vegetative state, or diagnosed with locked-in syndrome) to communicate by activating different brain areas to answer simple questions. Such brain-computer interfaces are also being developed in the gaming industry to enable hands-free play.

Such devices depend on complex algorithms to detect information flows on one hand and brain activity on the other. They operate below the level of human consciousness and depend on the calculative functions of algorithms that are not known to the operators, or even to anyone who purchases and uses the devices, thus raising difficult issues of volition and moral accountability that cannot be captured in standard discussions about intent.

Beyond dual use

Our message is that the dual-use lens is not adequate for scientists and policy-makers to anticipate the ways that neuroscience and neurotechnology may remake the regulatory and ethical landscape as the various technologies bleed into and emerge from the PSIM domains we have described. Current treaties and regulatory frameworks that focus on export controls are not designed to deal with the complex and far-reaching consequences that future brain research will have at the intersection of technological advance, or with the broad arenas where government and nongovernment actors seek to advance their interests. The Ethics and Society division of the Human Brain Project is preparing an opinion on “Responsible Dual Use” that includes more detailed recommendations. The opinion will be available at https://www.humanbrainproject.eu/en/social-ethical-reflective/.

Emerging alternatives to existing dual-use governance regimes focus on self-regulation by industry, universities, and philanthropic organizations, especially in relation to artificial intelligence, machine learning, and autonomous intelligent systems. Although these efforts to couple research and technology development with ethical and social considerations are undoubtedly important, their capacity to actually shape, let alone terminate, lines of research that raise fundamental ethical and social issues has not been demonstrated.

The University of Bradford’s Malcolm Dando, a leading advocate for stronger efforts at chemical and biological weapons disarmament, has consistently stressed the key role that can be played by the education of neuroscientists in the dual use potential and risks of their work. But he remains skeptical that such efforts can be sufficient to “match the scale of the security problems that will be thrown up by the continuing scope and pace of advances in the life and associated sciences in coming decades.” We urge regulators concerned with dual use in the United States, Europe, and elsewhere to think beyond the dangers of militarization and weaponization when considering how to regulate and monitor advances in neuroscience research today and to pay attention to the political, security, intelligence, and military domains of application.

Adapting to Global Warming: Four National Priorities

Global average annual temperatures have increased, and the rate of increase has accelerated, since the late nineteenth century. The extent of global warming in coming decades depends on future emissions of greenhouse gases, and even perfect mitigation to reduce emissions can only slow the pace of future warming. We cannot stop global warming in its tracks. Therefore, we need to adapt. But what are we adapting to?

Global average annual temperature is estimated to increase between 4.5 and 9 degrees Fahrenheit by the end of the century. This will be accompanied by average sea level rise of between 8 and 20 inches. These forecasts are for global average increases. Many locations will see peak temperatures increase even more, and during storms an increase of inches in the sea level can translate into floods measured in feet, not inches.

The likelihood of first order effects—extreme heat, drought, flooding, more severe storms, increased food insecurity, and the poleward spread of tropical diseases—is well known. We cannot, however, accurately forecast the magnitude and geographic distribution of effects over the next 50 years, because of both uncertainty about the future emissions of greenhouse gases and the challenge of determining how large-scale climate changes will affect local-scale climate.

An even greater challenge will be to predict the second order effects as first order effects tumble through a profoundly linked economy. What happens to food security when flooding regularly but unpredictably cuts off certain parts of a city? What happens to a tourist industry when a destination is no longer desirable because it is too hot or wet or dry? How much funding for local public services will have to be redirected to raise an airport runway by six feet or harden a seaport against future storm surge?

In sum, we do not know what is going to happen to the built environment, landscapes, and ecosystems that we depend on as the planet rapidly warms. If this were not bad enough, there are credible climate scenarios in which the planet reaches a tipping point that results in runaway changes in climate and sea level.

This unprecedented uncertainty about the future raises another set of issues regarding adaptation. Is it enough to prepare airports, farms, social safety nets, and emergency services to deal with creeping sea level and temperature rise? Or should we also prepare for large discontinuities in climate and worst-case events?

In the context of this uncertainty, magnified by the need to act despite meaningful precedents, there are four things we need to do now to address global warming:

1. Improve the quality, accessibility, and usability of climate hazard and risk information. Human adaptation to global warming is the sum of distributed decisions. Individuals, companies, and governments need much better and more locally relevant information than is currently available. The forecasts that are available are incomplete, not fine-scale enough, too infrequently updated, and disconnected from critical decision-making.

For the past several decades climate scientists have been focused on understanding the relationships among human activity, greenhouse gases, and global climate change. Their findings make a strong case for curtailing or changing certain human activities. In recent years, some researchers have shifted to understanding and modeling likely regional climate effects. This is critical for adaptation, and it needs to continue.

Although there are many well-meaning efforts to offer predicted climate data in support of local climate adaptation, the data that climate science can provide are inadequate. The information is too coarse in resolution, too uncertain, and too far in the future for cities, companies, and communities to use in their planning. Often, climate science can assess the magnitude of change, but cannot compare adaptation options. Significant gaps exist between the way scientists think about climate adaptation and the information citizens, governments, and company leaders need to make decisions or plan for the future.

Individuals buying homes or renting apartments need access to accurate and easy-to-understand presentations of location-specific flood risks; engineers, architects, and planners creating new facilities, or updating existing infrastructures, need to be able to anticipate climate hazards and risks far in the future; emergency services leaders need reliable estimates of the number and distribution of flooding or extreme heat days in their service area over the coming few years; and investors and insurers need reliable short- and longer-term predictions of climate change effects at a regional or even site-specific scale.

Governments can and should lead to remedy this situation, but the work will need to be done in collaboration with the variety of companies that have capacity in geoscience analytics, sensors, and map-based information presentation, as well as with an emerging cadre of adaptation professionals.

2. Increase investment in adaptation-focused research, development, and demonstration (RD&D). A substantial increase in breadth and depth of adaptation-oriented RD&D is needed. Solutions are needed in diverse areas, including human health and food security, infrastructure, urban systems, and natural resource management. And, of course, disaster preparedness and disaster response. This investment necessarily includes increasing the human capital for adaptation-related RD&D.

If research is exploration, applied research is problem solving. Applied research on human adaptation to global warming is woefully underdeveloped. Uncertainty about the scope and scale of the problem explains some of the current shortfall in applied research. Additional impediments are that critical applied research for adaptation is spread across literally dozens of disciplines and subdisciplines of engineering, science, social sciences, and medicine, and that critical expertise can be found in academia, governments, and the private sector. Mobilizing a dispersed and varied community of problem solvers to address a future threat of uncertain character and magnitude is inherently hard.

Some governments and corporations are beginning to integrate climate change adaptation into their strategic planning, provide resources to encourage others to do the same, and even organize research funding around adaptation challenges. For example, two international groups, the Adaptation Fund and the Special Climate Change Fund, both established under the United Nations Framework Convention on Climate Change, have mandates to fund adaptation efforts. In the United States, there are landmark federal efforts focused on adaptation, including the National Oceanic and Atmospheric Administration’s Regional Integrated Sciences and Assessments program and the Department of Interior’s Climate Adaptation Science Centers. Nonetheless, the bulk of federal funding for climate change work is focused on technology to reduce emissions. Private companies in mining, agriculture, and beverages have had to adapt to water scarcity through conservation and reuse, and consumer goods companies change their product profiles to react to weather patterns. Publicly traded companies in all sectors are under increasing pressure to predict and disclose climate-change-related liabilities as part of their financial reporting. These shifts are laudable and provide important background for action, but they are far from adequate to the challenges.

The nation’s RD&D enterprise—both government-funded and industrial—is decentralized but robust and can be responsive to well-articulated problems. It helps, of course, if funding accompanies the problem articulation. There are relevant RD&D programs at many federal agencies and in some states (e.g., California), but the scale is small relative to obvious needs, and the number of professionals with the necessary skills is small. A national effort to collate and update information on private, federal, and state government expenditures on adaptation-related RD&D, as distinct from overall climate-related RD&D, would be a good first step for linking existing adaptation efforts and identifying gaps.

With regard to global collaboration to reduce the national burden of adaptation-focused RD&D investment, the United States needs to be both strategic and tactical. It should make a fair contribution to the shared global RD&D enterprise without losing sight of problems that are critical for domestic needs, taking advantage of the value of shared investigations and global experience when possible.

Global networks of exchange and collaboration exist in almost every field of science and engineering, and this is the case for all the fields relevant to climate change adaptation. These need to be strengthened. The US government, universities, and companies need to be aggressively active on adaptation, including taking leadership roles in adaptation networks, so the nation can contribute as well as benefit. In today’s interconnected world, RD&D is a global enterprise, and it is foolish to ignore what is going on worldwide.

3. Plan for the public and private economic costs of adaptation. Individuals, companies, and governments have four approaches to preparing for uncertain future expenses: plan to reduce other expenditures, develop and then tap reserves, buy insurance, and borrow. The process of developing—and publishing for review—funding strategies for different levels of adaptation costs would be a good first step for both governments and companies.

There are a couple of reasons why we are unprepared for the costs of adaptation to global warming. First, there is genuine uncertainty about the timing and severity of effects. We do not know what adaptation will cost or when the bill will come due. That is true for individuals, state and local governments, and companies, as well as for the United States as a whole. Second, despite the uncertainty about just how bad it will be and when it will occur, we know global warming is bad news; any reasonable estimate of the funds required to adapt successfully is a large and scary number.

The American Society of Civil Engineers (ASCE) estimates a $2 trillion shortfall between planned funding and funding needed to maintain the US infrastructure between 2017 and 2025. Taking a single infrastructure system that has obvious connections to climate adaptation, the ASCE has estimated the near-term cost of fixing the nation’s 30,000 miles of levees at $80 billon. As global warming challenges our infrastructure those costs will certainly increase, and the United States is not prepared even to estimate those costs. The $1.5 trillion infrastructure plan released by the Trump administration on February 12, 2018, drew immediate criticism as flawed because it did not adequately address the additional infrastructure improvement needed to increase US resilience (adaptation) to climate change.

If $2 trillion over the next eight to ten years, as estimated by ASCE, is a realistic picture of the shortfall, that means our near-term needs for annual infrastructure funding—not seriously considering the costs of global warming—are already between 5% and 6% of the total US federal budget. And infrastructure upgrading is by no means the only public cost associated with global warming adaptation. Public health protection, population relocations, and disaster relief are just a few of the functions where we can expect increased public expense.

It is not unreasonable to believe that we’ll be fighting a war of adaptation to climate change for the next 20 to 50 years. How do we estimate the cost of adaptation and raise the necessary federal, state, and local funds? And how much do we need to spend internationally to protect US overseas interests and prevent conflicts arising from, or being exacerbated by, climate change effects?

There are no easy answers. Companies and individuals sometimes have the foresight and resources to reserve funds or buy insurance, but they need better information to be able to determine how much of either or both. The US government has not typically built reserves to deal with future needs (anticipated or unanticipated), but rather has adopted a pay-as-you-go and borrow-as-you-must approach to everything from building infrastructure to social security to the costs of wars. State and local governments operate in much the same way, often with much tighter constraints on borrowing.

The best thing we can do now may be to develop, and continuously improve, several public and private cost scenarios to help us recognize the questions we’ll face. In particular, what new trade-offs will come up in the application of public funds? Is the demand likely to be for obvious adaptation expenses, such as increases in disaster relief expenditures, or it is likely to be cooked into other demands, such as increases in the costs for infrastructure maintenance and improvement?

An explicit public discussion of different cost scenarios would help build public understanding and flesh out possible responses. None of these discussions is complete, of course, without also considering the cost of not adapting. It is already politically difficult to invest in disaster prevention despite the saving that comes from such investment. This difficulty will almost certainly increase as global warming raises the stakes.

4. Strengthen policies and plans related to adaptation. There is a real risk that the nation will look back from 2040 and realize that its responses to the challenges of adaptation were too little, too late. Leaders need to increase the level of attention and resources dedicated to implementing and evaluating national, state, and local levels of preparedness. Policies and plans for risk management, crisis avoidance, and protecting vulnerable populations should be the focus of that effort.

For some federal government activities (e.g., flood insurance, disaster relief, and coastal military bases) the issues raised by global warming, and the requirements of adaptation, will be front and center in the coming decades. Public officials responsible for these activities are already planning for adaptation. For other federal activities (e.g., interstate transportation, border security, and international diplomacy) the requirements of adaptation to global warming are likely to be significant but, perhaps, not the most immediate concern in coming decades. Planning and preparation for adaptation might be more challenging in these agencies.

At the level of state and local public services—fire departments, policing, building codes and zoning, schools, water services and sanitation—exposures to the effects of global warming are very substantial. In response, some cities have established positions or offices focused on resilience. A few have started aggressively planning for adaptation to extreme climate change effects including heat events, droughts, and increased flooding. Concerns about flooding seem to be driving many of the adaptation leaders; coastal cities such as New York, Los Angeles, New Orleans, and Miami are in the vanguard, as are states with substantial coastlines such as California and Florida. This would be a good time to launch a serious national evaluation of US preparedness for the effects of global warming.

Although there is considerable uncertainty about specific effects, there is certainty that global warming is significant, accelerating, and seriously threatening. Doing nothing is not an option, as the cost of not adapting is many times greater than the cost of the most aggressive adaptation efforts. And even as we start now, we have to prepare for an extended effort. This is likely to be a marathon, not a sprint.

Human ingenuity has rescued our species from other dire situations. The green revolution saved us from widespread famine, and human ingenuity has conquered some diseases, such as polio, and increased life spans around the globe. Human ingenuity will also help us deal with climate change. As first steps, we need to be ingenious enough to develop more useful forecasts; invest in applied research for adaptation; plan for the costs and economic consequences of global warming; and strengthen policies for risk mitigation, crisis avoidance, and protection of vulnerable populations.

Nuclear Power Needs Leadership, but Not from the Military

For much of the atomic age, the United States led the world in developing and deploying nuclear technologies. Despite building the world’s largest fleet of reactors—99 of which remain operational—and seeding most of the designs built worldwide, US commercial nuclear development has dramatically slowed. Indeed, the nuclear power industry now faces unprecedented—arguably existential—challenges. The nation’s demand for electricity has decreased, and the power distribution grid is rapidly becoming decentralized. Nuclear power is having trouble competing in current deregulated energy markets dominated by low-cost natural gas and renewable energy sources. The industry hasn’t been able to build new power plants within budget and in a timely manner, as recent efforts in South Carolina and Georgia illustrate. There are concerns about safety, waste management, and nuclear proliferation. And efforts to develop advanced reactors that might meet these challenges have lagged. The industry can’t afford major research and development, and efforts by the Department of Energy, once a prime mover in reactor development, have been moribund as a result of inadequate funding and leadership.

The decline of nuclear power creates huge challenges for important US policy goals on two fronts. First, national security experts are debating the consequences of a world where China and Russia become preeminent in nuclear science and technology, thus diminishing the ability of the United States and other western countries to set the rules for controlling access to nuclear material and technology. In addition to deploying existing designs domestically and exporting them, China and Russia have extensive and well-financed programs to develop the next generation of advanced, non-light water reactor concepts, many of which were first demonstrated decades ago in US national laboratories. Unfortunately, neither of these countries has displayed a commitment to maintaining and strengthening the international control regime.

Second, despite the challenges it faces, nuclear power remains a proven low-carbon technology. Many energy and climate experts think that without a portfolio of low-carbon energy options that includes at least some nuclear power, the nation and the world will be hard pressed in coming decades to sufficiently cut emissions of the warming greenhouse gases that drive climate change. But those assessments do not reflect nuclear power’s grim realities on the ground.

Often when US industries have found themselves in especially dire straits, they have suggested intimate cooperation between the Department of Defense (DoD) and technology vendors to restore their vitality. In the case of nuclear power, national security arguments are being offered to justify diverting large sums of public money through DoD to catalyze the development and deployment of small modular reactor (SMR) technologies. Such arguments have flowed from a variety of sources, including the director of research and development at Southern Company; defense-focused organizations such as the Center for Naval Analyses, Defense Science Board, and National Defense University; clean energy experts at Sandia National Lab; and organizations such as Third Way that include nuclear advocacy as part of their clean energy program.

Two decades ago, some in the industry recognized that continuing to build large, complex reactors made little sense. Companies proposed shifting to a class of smaller systems: small modular reactors with a nominal power output of less than 300 megawatts of electricity. Their costs would be both more affordable and more predictable—assuming they managed to exploit factory fabrication and modular construction, the way airliners and gas turbines have—somewhat ameliorating nuclear power’s steep economic cost. In the United States, more than 30 companies are pursuing SMR variants. Most face monumental challenges, lacking as they do the resources or experience to proceed. None has a guaranteed order book that could underpin the commercialization of either a reactor or its manufacturing infrastructure. The Department of Energy is helping one vendor develop a light water SMR plant in Idaho and may consider signing a power purchase agreement for some of the electricity it generates. It may also support another project in Tennessee by the same company. Despite these efforts, some observers see DoD as the only institution capable of providing a large enough market to justify commercial production, at least in the short term.

We also see SMRs as theoretically attractive, and we believe that DoD can play a role—though a supporting role—in strengthening US fission innovation. However, resorting to national security arguments and placing DoD in the lead is neither a sustainable nor wise way to revitalize the nation’s brittle nuclear enterprise. Indeed, a counterargument could be made that this proposed fix could actually weaken national security by diverting DoD’s focus and dollars to a lower priority mission. As to the role that DoD might usefully play, we offer three suggestions.

DoD and nuclear development

The introduction of commercial nuclear power in the United States was intimately tied to the military’s quest for rapid nuclear development. After all, construction of the first commercial demonstration plant near Shippingport, Pennsylvania, was overseen by Admiral Hyman Rickover himself. Building Shippingport was in Rickover’s interest at the time, before the Navy’s nuclear future was secured and before nuclear power was known well enough to engender strong feelings, one way or another. This approach generated both political momentum for nuclear development and the industrial supply chain for the light water reactor designs his naval program pioneered.

Resorting to national security arguments and placing DoD in the lead is neither a sustainable nor wise way to revitalize the nation’s brittle nuclear enterprise.

Once Shippingport was completed, the government established a firebreak—imprecise and imperfect—between the civilian and military nuclear programs. Political leaders saw much value in distinguishing the sunny side of the atom from its more sinister one. The civilian nuclear industry welcomed this separation. Acutely aware of the association between nuclear power and the bomb, it stressed in its literature, for example, that nuclear power plants cannot explode.

The separation was further promoted by the fact that the Navy had achieved its goal of developing a mature nuclear propulsion system and a fairly limited number of carefully managed plants that had proven themselves remarkably safe. Civilian industry, on the other hand, was projecting thousands of reactors. Because the Navy would not develop and deploy these, it had no way of vouching for their safety and security, and it worried that accidents at nuclear power plants might rebound on its program. Separating the two enterprises helped shield a successful naval nuclear program from any accidents that might occur at commercial facilities. Ostensibly, the civilian and military programs have since maintained that separation, though there remains much cross-pollination.

Should DoD assume a key role in developing SMRs, there are two types of reactors that it might conceivably pursue. The first is a very small reactor (vSMR) that the military could use on “forward operating bases” or perhaps some main operating bases in support of combat operations. The second is a larger reactor that the military could use to supply power for its domestic (and perhaps international) bases in order to isolate them from potentially vulnerable public electric grids.

There are also a variety of program models that DoD could adopt in developing and acquiring SMRs, though we see two as most likely. First, the military could initiate a tightly controlled development process, akin to that used in major defense acquisition programs such as the Joint Strike Fighter or the Littoral Combat Ship. Second, it could “bid for service,” offering commercial vendors the opportunity to deliver power to military installations without tightly controlling development and deployment. The economic and political implications of the two models differ greatly. It is highly likely that DoD would choose the first model for acquiring vSMRs, whereas either model might conceivably work for larger SMRs that power bases. Regardless of model, the practical and normative challenges of such an undertaking would be enormous.

Practical arguments

If DoD were to consider acquiring a fleet of SMRs, it would start by assessing its need for them and identifying the key mission performance parameters they would need to satisfy. DoD manages the development of new programs through its Joint Capabilities Integration and Development System, which starts by identifying either new threats or gaps in current capabilities. Assuming a valid gap is identified, the process then moves through material solutions analysis—essentially looking at all possible alternatives to fill that gap. There is little evidence that the SMR designs that would emerge from this analytic process would be appropriate for wide commercial deployment. In fact, most of the evidence suggests quite the opposite.

The operational requirements and performance standards that would be critical in a vSMR slated for use in forward operating bases would most likely yield a very small reactor (nominally less than 10 megawatts of electricity), blunting its broad commercial desirability. In addition to their small size, these reactors would likely be developed with other characteristics that may prove troublesome for commercial industry or regulatory bodies. First, they would need to have a long core life, which means they would require fuel “enriched” with uranium-235 to a higher level than used in commercial reactors. Second, they would have to be transportable and thus rugged. Third, for variants that are small enough to be deployed in remote or forward operating bases, the design might have to employ very modest containment structures. Fourth, they would be designed to absorb shocks and to continue operating in spite of these. Fifth, control systems would most likely have to be automated to a great degree, because it costs money to train personnel, and DoD would not look kindly on SMRs that turn forward operating bases into superbly guarded power plants that have more plant operators than warfighters.

Though SMRs designed to provide power to US military bases could be larger, it is likely that DoD would still want substantially different technical specifications from those of commercial reactors. As with vSMR variants, sound arguments can be made for more highly enriched cores to lengthen refueling intervals, greater ruggedness to support defense and security resilience, and robust standalone control systems to limit vulnerability to cyber and physical risks.

The characteristics outlined above would most likely be endorsed by many reactor designers, and some experts have called for DoD to be an early leader to ensure technical specifications such as these are tailored to meet unique DoD requirements. But it is difficult to overstate how politically unpalatable some of them would be to the Nuclear Regulatory Commission. This is especially true of any move toward higher fuel enrichment. Should they be incorporated into one system, we have great difficulty envisioning a license being granted without institutional reforms so deep that they would be sufficient to enhance nuclear power’s prospects on their own, weakening the case for the military’s involvement. Of course, since DoD regulates its own reactors, it would not have to deal with these institutional constraints. However, the argument that its role as first mover could accelerate the commercial licensing process is not very compelling.

Moreover, any SMR fit for DoD’s needs would likely be too expensive for a commercial utility to deploy. As noted above, the same requirements and standards on which DoD would insist might render commercial variants economically uncompetitive in most of today’s markets. In fact, the military variant itself might not be deployed widely: even if the political will did materialize, and substantial appropriations were forthcoming, and industry eagerly participated, the likelihood of successful project execution is still modest. Given the long history of cost overruns and waste in defense acquisition, the empirical record is now so robust that this statement is almost axiomatic.

SMRs designed to serve a military base would face an even greater challenge since they could be attacked on the same economic grounds as commercial reactors. The arguments in favor of vSMR deployment—that they could minimize the need for long fuel supply lines and thus fill a warfighting capability gap, for example—evaporate when considering large, domestic bases. Even assuming that such a need could be identified and that it justifies new program development, defense programs must be competitive when possible. Unless resilient, independent supply were given far more weight than it now receives, there is little guarantee that a nuclear design would win the day for base power supply any more than it is dominating deregulated energy markets. Other generating sources such as combined cycle natural gas constitute viable alternatives if sound dual-fuel management practices are adopted. In short, SMRs would remain victims of their poor economic competitiveness, absent a hard, statutory mandate or an overly rigid performance parameter that forces a nuclear solution. Regardless, the existence of ready substitutes makes this program a recurring target for the DoD budgeter’s ax: it would consistently be the last program in and the first one out of any submission to Congress.

Any SMR fit for DoD’s needs would likely be too expensive for a commercial utility to deploy.

One often-cited advantage of having the military act as first mover is its looser restrictions on siting large infrastructure. But even this is not guaranteed to help nuclear power. The military is now the only government institution trusted by a plurality of US citizens. It is certainly sensitive to the fact that this well of trust is not bottomless and that goodwill needs to be cultivated among communities in which it operates. One way of cultivating this goodwill is to follow state environmental guidelines and processes when they do not compromise the defense mission. This is to say that the siting of small reactors would likely still become an issue for DoD in a range of locations, not just those with some of the largest concentrations of military installations, such as Hawaii or California, that reject nuclear power outright.

Should a fleet of SMRs developed by DoD somehow succeed in fulfilling its (poorly defined) mission, and should it succeed in spawning a following within Congress, another prosaic question—that of intellectual property ownership and program management—would emerge. Once DoD had deployed several dozen SMRs, and once the infrastructure that supports these SMRs was developed, even following a “fee for service” model, the military might consider bringing the entire enterprise in-house. This would effectively create another Naval Reactors program and would surely engender strong opposition. Given the contentious nature of budget battles across the services, the military does not need additional fiefdoms. The creation of Naval Reactors was initially highly controversial, and any SMR fleet that relied extensively on DoD development and deployment would undoubtedly create another such mini-empire. For such a program not to devolve into a fiasco, a similarly vigorous level of project management, situational awareness, and political lobbying would be required. Military organizations that have been created to address more critical warfighting needs, such as the Joint Improvised Explosive Device Defeat Organization, grew far beyond their original mandate and expected scope. If commercial development lags or fails to materialize, there is also the very real possibility of developing another industrial base component that is entirely reliant on the military for support, much like shipbuilding. Parts of the military-industrial complex already resemble a monopsony, with programs maintained to support the nation’s industrial base despite often-limited warfighting impact: the controversy surrounding the USS Zumwalt-class guided-missile destroyer, which many critics have called a boondoggle, serves as a prime example.

The organizational troubles do not end once the overall management structure is defined. Whether managed in-house or by a contracted vendor (as with the Navy nuclear labs), another fiefdom would almost certainly be created in the form of a waste-management organization. Given the absence of a national geological repository, this would effectively force the military into the political imbroglio that is nuclear waste management and storage. Even a small fleet of a few dozen SMRs would quickly saddle the military with an unenviable job. Complicating things further, this waste would likely not be under International Atomic Energy Agency safeguards. It would be logical for other nuclear powers to assume that it is potentially weaponizable and alter their own strategic doctrines accordingly. Alternatively, DoD could lobby for a transfer of responsibility to the Department of Energy, which would simply add to that organization’s already challenging waste management task.

Finally, should a DoD sponsor tightly control reactor development, as the Navy does, it is likely that the details of these SMR designs would become classified due to the nature of their end use. Even if DoD refrains from exercising strict control over the program and instead relies extensively on commercial industry, creating a commercial variant would be difficult, and exporting to other nations would be virtually impossible outside a handful of allies. Despite efforts to reform the US export control system, it is still notoriously difficult for US companies or researchers to engage with foreign commercial entities and researchers on nuclear power. A reactor attractive enough for DoD to buy in numbers large enough to exploit learning economies would likely have enough classified equipment, materials, technology, and expertise to greatly limit its export potential.

Normative and policy arguments

In addition to the host of practical challenges outlined above, there are compelling normative arguments to be made against relying on the military to revivify the nuclear enterprise, regardless of the development model. First, this job rests far outside the military’s remit. The military develops new technologies when they are the only available solution to a problem. If a competitive and suitable commercial option is available, the military can and often does acquire and adapt that technology to meet its needs, supporting the national economy in the process. Examples include the purchase of low-emission vehicles for official government fleets and the Boeing 737-derived P8 Poseidon Maritime Patrol Aircraft. Such cases are far removed from the scenarios proposed for military leadership in SMR development. In addition to being overcommitted, the military is facing acute and evolving threats. There are national imperatives to which it could attend, but it is not realistic to suggest that SMRs rank (or should rank) high on this list. Even conceding that a security case could be made for vSMRs and SMRs—namely on forward operating bases and for independent, secure, and reliable power on US bases, respectively—these would not rank near the top of a list of new priorities when compared with existing or emerging needs and threats.

In our view, those who support a leadership role for the military in this venture misrepresent or misunderstand the role that it plays in sustaining the nation’s industrial supply chain. The military certainly relies on the country’s civilian industrial base and its human capital in developing new systems. It often adjusts its plans to sustain troubled companies if their loss would render critical capabilities extinct and thus threaten national defense. These actions require careful judgment and create political controversy even when costs are limited and the technologies are actual combat systems.

Perhaps the closest analog is the shipbuilding industry. To maintain US capabilities in this field, the military provides extensive support to Electric Boat and Huntington-Ingalls Shipbuilding, and a number of other shipyards that are almost exclusively tied to government projects. Years ago, when the shipbuilding industry was in danger of collapsing, threatening the supply of critical components, the military stepped in to secure those supplies, paying a premium for some components in the process. Though this approach may not be ideal, it at least acknowledges the structural realities of the US and global economies, instead of exploiting the military as an instrument for changing them. The situation for nuclear power is markedly different. The future of the nuclear Navy is not in doubt, and the argument that the collapse of civilian nuclear power will endanger that future is unconvincing. The Navy relies on a more limited reactor supply chain, has its own design and research laboratories, supports its own extensive computing capabilities, and trains its own operators. If anything, the civilian industry is enriched by that infrastructure and the infusion of former nuclear navy operators, not the reverse. Were a core capability to degrade, the military could rescue that capability at far less cost than developing an entire complex commercial venture.

Often, both advocates and critics of nuclear power note that the firebreak between the civilian and military nuclear programs is artificial. But no matter how imprecise or imperfect, this firebreak has substantial normative value. It ensures that power and weapons remain separate enough for evolutions in the latter not to compromise or cloud the future of the former. Maintaining this separation is perhaps the easiest way to guarantee that US-made SMRs are exportable while ensuring that qualitative US military superiority in reactor development is maintained. Also, if the boundary were eroded, it would make it far more difficult for the United States to restore its standing with respect to international controls on the development and use of civilian nuclear power and fuel cycles.

There are fundamental civic values at stake too. At a time when the nation’s civic and political norms rest on precarious ground, using the military to rescue a commercial industry—especially one that has so consistently failed to assuage public concerns about its products—might further degrade the social fabric from which the military derives legitimacy. Should the proposed SMR rollout by DoD succeed, that might prove even more problematic, given that the Republic was designed to prevent large concentrations of political power. It would also undercut the Department of Energy by underscoring its failure to develop advanced nuclear designs, much as the Department of State’s leadership and effectiveness in international relations have been undercut by DoD’s expansion into its bailiwick in Iraq and Afghanistan. This undertaking would substitute a US military role in technocratic planning for a national industrial strategy and a functioning political system. The United States is certainly in need of nation-building at home, with an emphasis on rebuilding large swaths of public infrastructure, but the military should complement this process and not lead it.

Finally, adopting this model would amount to an admission of failure on the nuclear industry’s part. Defaulting to the national security argument in an effort to salvage the commercial nuclear industry concedes the failure of the technical and economic arguments in favor of the technology. The nuclear energy enterprise is at a critical juncture, its problems are unlikely to abate in the near future, and most of these are fundamentally economic and political. The deregulation of the electricity markets has yielded utilities that focus on short-term profits and neglect long-term, structural challenges. Every lender with the assets to finance generation expansion is similarly preoccupied with short-term profits. Moreover, US energy markets remain warped: where policies favor low-carbon power, they often exclude nuclear technologies. These are profound problems that deserve attention, and fixing them will benefit nuclear power, the nation, and the climate.

Ways for DoD to help

If pressing DoD to play a leading role in developing and deploying commercial nuclear technologies is unwise and potentially counterproductive, what role can it play? We suggest three modest policy changes that might help DoD buttress the enterprise without burdening it with the responsibility for SMR development.

First, federal policies governing the length and budgeting parameters of federal power purchase agreements (PPAs) with commercial utilities ought to be amended or clarified to ensure that long-term agreements are recognized as a viable option. Modifications to budgeting rules that would enable funding to be spread over many years (not over a single year, as is typically required) would give DoD greater flexibility to enter into these agreements. These could then provide commercial SMR developers with the stable source of income necessary to raise capital for first-of-kind reactor deployment. Much as the Department of Energy is considering long-term PPAs for two of its national laboratory facilities, utilities serving large military installations could sign long-term agreements with these installations to hedge against large upfront development costs. Key to this would be clear policy guidance from DoD that supports 30-year PPAs for nuclear, which are already allowed by the General Services Administration for DoD. This option could help ensure long-term commitments to nuclear power from within the commercial nuclear industry and venture capital providers.

Second, the Navy ought to explore the possibility of supporting development of human capital in the nuclear field by expanded teaming with academia and industry. Formal collaboration in areas such as operator training, student fellowships at the Bettis and Knolls nuclear power laboratories, and advanced degree options for nuclear-trained officers at select academic institutions would benefit both military and civilian nuclear programs. This may help offset reductions in enrollment at some nuclear engineering programs and keep them viable while the industry works to find its footing.

Third, the military has great interest and experience in enhancing the resilience of electric power grids. Given the growing interest in distributed generation, microgrids, and SMRs, modest collaborations would help civilian groups learn about critical and potential vulnerabilities and gain confidence in deploying strategies and technologies that can increase the resilience of these systems.

The Navy ought to explore the possibility of supporting development of human capital in the nuclear field by expanded teaming with academia and industry.

To date, the military has expressed only modest interest in small reactors, and as we note earlier, may opt to back away from them if budget cuts are required. However, a high-tech SMR program that deploys a civilian-operated fleet of reactors around the country could sustain skilled, high-paying jobs winning broad congressional support. This would make them hard to kill once they are embedded in the budget—even if military support wanes.

There is a way forward that could help the nuclear industry, enhance international control regimes for nuclear power and fuels, and aid in addressing climate change. But the way is not to treat DoD as a venture capital firm. Advocates of DoD leadership in the SMR business must acknowledge the enormous challenges that face their proposal. Building a “great SMR fleet” with DoD as its core customer may shore up the leaky nuclear ship for a time, but it might also bring the industry very near the type of shoal water that made the US shipbuilding industry a monopsony. Committing the military to another too-big-to-fail development effort would not rescue the commercial nuclear enterprise. In fact, it may weaken national security, forcing the Department of Defense to support an entire new industry instead of focusing on a host of emerging threats. Supporters of this strategy must think carefully about what the nation may gain and lose over the long term should their proposed remedy succeed.

Is Innovation China’s Next Great Leap Forward?

Innovation refers to the process of implementing new or improved technology and management practices that offer products and services with desirable performance at affordable cost. Innovation encompasses the entire pathway from early-stage idea creation through technology development and demonstration, and finally to late-stage production and deployment. Innovation aims at increasing consumer satisfaction, economic productivity, growth, exports, and jobs.

A number of significant indicators suggest that China’s innovative capability is increasing more rapidly than that of the United States, portending a weakening of the United States’ relative global competitive position:

• China’s total research and development (R&D) expenditures are growing faster than those of the United States, but China is at a lower absolute level.
• Many attribute China’s technical advances to illicit transfer of technology, theft of intellectual property (IP), and requiring transfer of technology from foreign firms operating in China, in violation of World Trade Organization rules
• China has adopted aggressive technology targets in key areas such as robotics, artificial intelligence, and digital learning, with the announced goal of surpassing US capabilities.
• China is increasing its investment in small and large high-technology US enterprises without allowing US firms similar investment opportunity in China.
• China’s intelligence agencies are expanding efforts to collect information from Chinese students doing research on US university campuses in areas relevant to China’s key technology

Do these developments add up to a real threat to US innovation leadership? And if so, what is an appropriate policy response. The sanguine view is that China’s innovative progress is the inevitable result of its economic growth and technological maturity and that market forces will cause the United States and other members of the Organisation for Economic Co-operation and Development (OECD) to adjust the balance of goods and services they offer in response. The harsher view is that the United States must adopt tailored protectionist measures to compensate for unfair trade practices until the two countries reach a negotiated agreement on stable and mutually beneficial market access, cross-border investment, and technology transfer.

This is not the first time the United States has appeared to lose global technical and innovative leadership. The Soviet Union’s 1957 launch of Sputnik, the world’s first artificial satellite, raised the specter that the United States was losing the “space race.” In the 1980s there was widespread concern that the United States had fallen irrevocably behind Japan in manufacturing, in areas such as autos, flat panel displays, and consumer electronics.

China has adopted aggressive technology targets in key areas such as robotics, artificial intelligence, and digital learning, with the announced goal of surpassing US capabilities.

In the case of Sputnik, the US government took prompt actions to build a significant civilian and military space capability. As a result, the United States has a competitive but not overwhelmingly dominant global position in space systems and, more important, in use of space systems for communication, precise geospatial location, navigation, and remote earth sensing.

In the case of Japanese manufacturing, the United States, Korea, and Taiwan quickly developed a broad array of competitive products such as Apple computers, CISCO communications hardware, and Microsoft software. In addition, Japan’s aging work force limits its ability to grow and adapt.

However, the present situation vis à vis China is more serious for two reasons. First, China’s rate of increase in economic activity and technical capability is greater than the rate of the United States and other OECD countries. Second, policy uncertainty, especially with regard to trade and climate, has slowed US private investment. The US’ vulnerability is due not to significant weakening of its innovation capability, but rather to China’s growing relative economic strength and emphasis on innovation in its economy.

Comparison of innovation trends must be seen in the broader context of US-China economic and political relations. Xi Jinping is now serving simultaneously as general secretary of the Communist Party of China, president of the People’s Republic of China, and chair of the Central Military Commission, without term. China seems to be reversing its slow but steady evolution to more democratic governance and reverting to a centralized party system with a single leader making all key decisions. Washington and Beijing are divided on many issues: tariff barriers, China’s actions in the South China Sea, its modernization of its military forces, difference over the status of Taiwan and Tibet, the importance of North Korean sanctions, and human rights violations. In this climate, differences related to innovation are amplified and more difficult to resolve.

What do the data say?

There are three sources for comparative country data that bear on innovation: the US National Science Foundation’s biannual Science and Technologies Indicators, the OECD’s series on R&D statistics, and the World Bank’s indicator databank that covers science and technology. These data share the shortcomings of uncertain data quality and reliance on purchasing power parity rather than market exchange rates to compare national efforts; add to that the caution of the late Lester Thurow, the influential economist and professor at the Massachusetts Institute of Technology: “Never believe a number coming out of China.”

Total R&D expenditure is the indicator most frequently cited. China’s total R&D expenditure is increasing more rapidly than that of the United States but has not yet reached the US level. As illustrated in Figure 1, if the United States has reason to be concerned with this trend, certainly the European Union (EU) and Japan should have greater concern.

Other indicators of inputs to the innovation process are the size of the science, technology, engineering, and mathematics (STEM) workforce; expenditures on R&D plant and equipment; and the availability of venture capital. Figure 2 compares the size of the STEM workforce in several countries over time. China’s workforce exceeds that of the United States, and its rate of increase is greater; however, the EU has a larger workforce than either country. In 2014 China awarded about 1.65 million bachelor of science and engineering degrees, whereas the US total was about 742,000.

In 2015 the percentage of foreign workers in the US STEM workforce was approximately 30%, with India providing 20% and China 10% of that total. Labor market experts project that the US economy will need to rely on increasing numbers of foreign STEM workers because the US education pipeline is not large enough to meet anticipated demand. But one cannot predict with any certainty how many foreign-born scientists and engineers will want to study and work in the United States or how many that spend time studying and working in the United States will leave and take with them valuable knowledge that will then benefit other countries.

Regarding the second factor, relatively little data or analysis is available to compare the inventory and flow of expenditures on R&D plant and equipment of different countries. The R&D facilities inventory is probably higher in the United States and the EU than in China, but China’s annual expenditure on R&D facilities is rising and may now exceed the annual US expenditures. The balance of payment flows for intellectual property favors the United States by a factor of four, but payments to China are increasing.

Although measuring inputs is one way of understanding the innovation process, the most important consideration is outputs; in particular, identifying the factors that influence process efficiency in different countries. Technical micro output indicators such as patents, publication, citations, start-ups, and licensing agreements do not tie directly to the innovation capacity of a firm, industry, or nation, and they are even less useful for assessing the contribution that new technology or business practices makes to profitability or competitiveness.

Direct measurement of innovation capacity is difficult, but there have been several attempts to develop macro indicators, which have important qualitative features. The French international economics research center, CEPII, presents a thorough discussion of China’s 13th Five-Year Plan (2016-2020) and concludes: “The first and most important objective is the shift from capital accumulation-led growth to innovation led-growth in order to enhance total factor productivity (TFP) and release the huge potential of consumer spending.” Also, the McKinsey Global Institute has analyzed four different aspects of innovation—customer focused, efficiency driven, engineering based, and science based—concluding that “China has the potential to meet its ‘innovation imperative’ and to emerge as a driving force in innovation globally.”

The Asia Society’s Policy Institute in cooperation with the Rhodium Group tracks progress in China’s economic “reform” program in 10 clusters, one of which is innovation. The groups estimate innovation progress by the ratio of value-added output from “innovative” industries to total value-added industrial output over time and compare the trend in China to the trend in United States, the EU, and Japan. Over the past five years, China’s position has increased from 30% to 32%, while the United States, EU, and Japan have remained relatively constant at 34%, 37%, and 46%, respectively. The indicator is a crude measure because the portfolio of innovative industries is arbitrary, consisting of seven sectors, with equipment and communications, computers, and electronics comprising about 50% of the total. The share of innovative industry compared with conventional industry in Chinese exports has not changed dramatically during this period.

Innovation infrastructure

A country’s innovation performance depends strongly on its underlying innovation infrastructure, including:

• Education of scientists and engineers who will enter the technical workforce
• Industry/university/government partnerships
• Standards for materials, products, safety, and subsystem interfaces
• Established patent, publication, and intellectual property rights
• Tax treatment of R&D activities
• Export controls on technology transfer and on participation of foreign scientists and engineers in the R&D enterprise
• Access to venture capital

The US innovation infrastructure is the envy of the world, especially for early-stage R&D and for giving foreign students the opportunity to learn both the technical and entrepreneurial aspects of innovation. China has taken a number of steps to improve its R&D infrastructure, but it still lags the United States, the EU, and Japan. The United States also has a massive lead in private venture capital spending (see Figure 3).

In March 2016 China’s National People’s Congress ratified its 13th Five-Year Plan, which includes increased R&D spending by the Ministry of Science and Technology, National Natural Science Foundation, and Academy of Science to improve the science and engineering base as a means to strengthen the infrastructure that fosters innovation. It also provides guidance for a number of industry sectors and draws heavily on the July 2015 Made in China 2025 strategic blueprint to achieve global manufacturing leadership through innovation. The goal is to upgrade industry writ large, but the plan highlights 10 priority sectors: new advanced information technology, automated machine tools and robotics, aerospace and aeronautical equipment, maritime equipment and high-tech shipping, modern rail transport equipment, new-energy vehicles and equipment, power equipment, agricultural equipment, new materials, and biopharma and advanced medical products. China 2025 is supported by other documents, such as a very detailed blueprint for artificial intelligence announced by the State Council.

One cannot help but admire China 2025. It sets clear goals based on explicit strategic priorities and support mechanisms, and it lists 12 specific quantitative key performance indicators for measuring improvement between 2015 and 2025. Technology strategy documents issued by the United States do not compare. For example, the White House Impact Report: 100 Examples of Obama’s Leadership in Science, Technology, and Innovation, released in June 2016, recounts many valuable actions but no overall strategy and no comprehensive plan to strengthen federally supported innovation activities across the board.

Expert commentators have recognized the threat implicit in China 2025. The Center for Strategic and International Studies notes the challenge presented to multinational corporations in the priority sectors. The American Institute of Physics warns the United States is at risk of losing global leadership to China. A Council of Foreign Relations blog is titled “Why Does Everyone Hate Made in China 2025”? The US Chamber of Commerce subtitles its review of Made in China 2025 as “Global Ambitions Built on Local Protections,” asserting that it “aims to leverage the power of the state to alter competitive dynamics in global markets in industries core to economic competitiveness.” The US government’s US-China Economic and Security Review Commission’s analysis has the provocative title “China’s Technonationalism Tool Box: A Primer.”

The unanimous message is that China is adopting a predatory economic policy that presents threats to the US economy and national security, including:

• IP cyber theft and illegal technology transfer
• Restricted access of foreign firms to China and requiring foreign firms operating in China to transfer technical know-how to China, violating World Trade Organization rules
• Unfair trade practices, including providing subsidies to key high-technology priority industry and dumping key products in export markets at prices below domestic Chinese prices and costs
• Nontraditional collection of advanced technology that targets professors and students in US universities and research centers

A number of official US reports address these threats. The Office of the US Trade Representative issued a report in March 2018: Findings of the Investigation into China’s Acts, Policies, and Practices Related to Technology Transfer, Intellectual property, and Innovation. The 180-page report catalogs a wide range of Chinese actions that are “unreasonable or discriminatory and burden or restrict US commerce” and justify further investigation and response. China has made some efforts to address these concerns. In 2017, Chinese Premier Li Kequiang announced, “We will fully open up the manufacturing sector, with no mandatory technology transfers allowed, and we will protect intellectual property.” Also in 2017 China’s State Council released Several Measures for Promoting Foreign Investment Growth, which contains 22 measures that are aimed at increasing foreign investment and optimizing the utilization of foreign capital. As Martin Feldstein, a former chair of the Council of Economic Advisers and now a professor at Harvard University, has noted, these expressions of intent have not eased concern.

In another illustration of concern, the National Bureau of Asian Research established the Commission on the Theft of American Intellectual Property (IP Commission), chaired by former Director of National Intelligence Admiral Dennis Blair and Ambassador Jon Huntsman Jr., which issued a report in 2013 and an update in 2017 on the extent of US loss of IP as the result of a range of China’s illicit activities, including cyber-enabled means. The IP Commission’s latest report estimates that the annual cost to the US economy from IP theft to be between $225 billion and $600 billion, with “China being the world’s principal infringer.” Whereas other nations commit industrial espionage against the United States, China’s activity is more extensive and better organized at exploiting stolen information and data in its domestic economy.

The IP Commission’s latest report estimates that the annual cost to the US economy from IP theft to be between $225 billion and $600 billion, with “China being the world’s principal infringer.”

On February 13, 2018, Director of National Intelligence Daniel R. Coates released at a Senate Intelligence hearing a statement titled “World Wide Threat Assessment of the U.S. Intelligence Community.” Among its observations, it states: “China, for example, has acquired proprietary technology and early-stage ideas through cyber-enabled means. At the same time, some actors use largely legitimate, legal transfers and relationships to gain access to research fields, experts, and key enabling industrial processes that could, over time, erode America’s long-term competitive advantages.”

At the same hearing, FBI Director Christopher A. Wray was much more explicit about China’s threat to US universities and from the more than 50 Chinese Ministry of Education Confucius Institutes on US university campuses. Asked by Senator Marco Rubio (R-FL) to comment on “the counterintelligence risk posed to U.S. national security from Chinese students, particularly those in advanced programs in the sciences and mathematics,” Wray responded: “I think in this setting I would just say that the use of nontraditional collectors, especially in the academic setting, whether it’s professors, scientists, students, we see in almost every field office that the FBI has around the country. It’s not just in major cities. It’s in small ones as well. It’s across basically every discipline. And I think the level of naïveté on the part of the academic sector about this creates its own issues. They’re exploiting the very open research and development environment that we have, which we all revere, but they’re taking advantage of it. So one of the things we’re trying to do is view the China threat as not just a whole-of-government threat but a whole-of-society threat on their end, and I think it’s going to take a whole-of-society response by us. So it’s not just the intelligence community, but it’s raising awareness within our academic sector, within our private sector, as part of the defense.”

The record clearly establishes extensive Chinese illicit technology transfer behavior. Anyone with US national security experience does not need to be convinced. What is striking is the implied judgment that this illicit behavior has been and will continue to be decisive to the advance of China’s innovative capability. There are few, if any, voices raised to say that significant improvement in Chinese innovation should be expected with the growth of China’s economy and the increased maturity of its indigenous science and technology infrastructure without any illicit behavior.

Summing up, I believe the comparative advantage of China will continue to be in build-to-print manufacturing and its efficient supply chain. The US strength will continue to be its customer focus and developing new technologies for widespread application. The story of China’s production of photovoltaic (PV) modules is illustrative. China has global leadership in low-cost manufacturing of PV modules based on conventional silicon solar cells. The impressive drop in average sale price and accompanying demand growth is due to manufacturing overcapacity stimulated by Chinese provincial, not central, government subsidies. Up until 2017 Chinese PV firms had not enjoyed profitability. In contrast, the United States maintains its lead in creating advanced PV technologies and developing production equipment and technology, which Chinese firms import and on which they rely. In a joint study, the Department of Energy National Renewable Energy Laboratory and the Massachusetts Institute of Technology examined China’s advantage in low-cost PV manufacturing, concluding that the “price advantage of a China-based factory relative to a US-based factory is not driven by country-specific advantages, but instead by scale and supply-chain development.” On the other hand, it is very unlikely that China will dominate the United States in key innovation areas such as artificial intelligence, robotics, machine learning, and genetic editing using a technology called CRISPR, at least for the next several decades. The pattern of China’s manufacturing strength and US strength in creating new technology opportunities is likely to continue in the future, at least for the next decade or so, because of the relative strength and maturing of each country’s innovation infrastructure.

Implications for US policy

The current relative strength of US innovation should be appreciated but not taken for granted. If the United States is to maintain its innovation leadership relative to China, it cannot stand still. The United States must adopt policies that both respond to China 2025 and strengthen the US innovation capability. Actions required include:

First, the United States’ bilateral trade discussions with China should go beyond tariff negotiations to matters that directly affect innovation: market access, cross-border investment, and technology transfer.

Second, the United State must continue to monitor and expose illicit Chinese activities such as IP theft, especially by cyber penetration, patent infringement, export of counterfeit goods, widespread use of unlicensed software, and forced transfer of technology from foreign firms operating in China. The US government needs to track each documented incident and establish a process for confronting China in each case and to enforce US law by assessing penalties on violating firms. Present and former administrations have taken steps to protect the country from these illegal efforts, but much more needs to be done.

Third, the United States must develop a new policy for engaging China on cross-border investments in high-technology firms and activities. This recommendation is motivated by two realities. First, the United States and China have very different motivations for cross-border investment. US firms are primarily interested in offering goods and services to China’s large and growing domestic market. These firms fear China’s practice of high-jacking technology in order to establish competitive indigenous capability. Chinese investments in the United States are primarily in firms with capabilities in the advanced technologies outlined in China 2025. Increasingly these investments will be in start-up companies and in joint ventures that are creating key technologies for future innovation.

The second reality is that many of the key technologies, notably artificial intelligence and robotics, are inherently dual use, with important applications in both the commercial and national security sectors. The US multiagency Committee on Foreign Investment in the United States (CFIUS) has a long history of reviewing transactions that could result in the control by a foreign entity in order to determine the effect of such transactions on US national security. The original CFIUS mandate required it to focus “solely on any genuine national security concerns by a covered transaction, not on other national interests.” But over time CFIUS has been pressed to examine transactions that are perceived to affect US economic competitiveness, including foreign transactions that involve “critical technologies” or that might entail industrial espionage.

The United States must develop a new policy for engaging China on cross-border investments in high technology firms and activities.

CFIUS is ill-equipped to assess the implications of start-ups and joint ventures in rapidly evolving key dual-use technologies of uncertain future application or success. If the United States wishes to reduce the ease with which China (and possibly other countries) can acquire US advanced technologies to fuel its innovation initiative, it is necessary to adopt controls that restrict access.

These controls should begin by requiring that all investment by Chinese entities in US enterprises—start-up companies, joint ventures, and venture capital funds—in a defined set of key advanced technologies be registered with one of the CFIUS agencies; prohibiting China from holding a 100% interest in US advanced technology enterprises; and mandating that any enterprise in which China holds a majority create an independent security supervisory board similar to those sometimes required by CFIUS to monitor and report all offshore technology transfer. Federal agencies that fund technology development would be permitted to award contracts or grants only to Chinese enterprises that had operating subsidiaries in the United States.

China should be allowed to support research activities in US universities, provided that research results are made publicly available. Chinese firms supporting research on US campuses would not be permitted preferential or exclusive licenses to any intellectual property produced.

Much effort is required to define precisely each of these suggested measures and a supporting administrative structure. The justification for such protectionist measures is to respond to China’s innovation initiative whose announced intention is to dominate world markets and whose progress depends to a significant degree on illicit technology transfer.

Fourth, the United States should not place restrictions on US universities and research centers. Restrictive proposals include preclearing publication of research results supported by the Department of Defense, applying the classification category of “sensitive but unclassified” on some government-sponsored research, and placing restrictions on foreign graduate students joining “sensitive” research projects and on presentations of research at international meetings. Each of these measures conflicts with the open structure of admission, research, and publication that keeps the US innovative ecosystem fresh, exciting, and agile.

Efforts by the federal government to control university research would be ineffective because its agencies are unlikely to balance properly the benefit of a proposed restrictive measure with the adverse impact on the quality of research; university faculty and administrators are ill-equipped to administer such restriction; and a restrictive environment will inevitably slow the flow of students from China and other countries who are needed by US industry. If the government imposes restrictions on research it sponsors, it will weaken its link with the universities that have been so central to US innovation. The risk of technology leakage is minor compared with the losses that will be incurred by restricting inquiry on university campuses.

It would be futile for the United States to try to maintain its competitiveness and its lead in early-stage innovation by trying to keep others out of US universities or to keep ideas in. The only effective response to China’s growing capability is to strive to master the new intellectual frontiers and to continue to recruit the most talented workforce able to translate new ideas into practice. In this regard, the United States should continue to welcome Chinese and other science and engineering graduates to US universities and liberalize immigration green card requirements to assure adequate supply for US industry.

Fifth, the United States must dramatically increase its innovation effort, especially in manufacturing, to bring the key future technologies to market. The United States should not adopt, as China has, a single national strategy. Nor should the United States rely simply on increasing federal R&D support from traditional agencies. US innovative activity is tremendously dynamic, with individual entities applying new technical applications in unique ways. This dynamic process, distinctive to the United States, is possible because of its formidable innovation infrastructure and strong tradition of customer and application focus. The approach depends on the open and free character of US society; it is difficult to imagine such productive vitality existing in communist China, where freedom of expression and association is so restricted.

The three prongs of US innovation—the federal government, industry, and academic research centers—need to follow distinct pathways to achieve a higher level of national innovation.

Individual industry sectors have the most important role in bringing new technology and business practice to market. Industry associations need to convince members of the urgency of increasing the pace of innovation and to provide them with case studies of successful unconventional new innovation.

The United States should continue to welcome Chinese and other science and engineering graduates to US universities and liberalize immigration green card requirements to assure adequate supply for US industry.

Universities and research centers have two important roles: to increase the flow of researchers with the motivation and experience to achieve innovations, and to expand work on key technologies. Both roles serve to enhance the dominant US innovation infrastructure.

The federal government’s role is to enable enhanced innovation in several ways. R&D support should place priority on proposed work that stresses innovation (without endangering early-stage fundamental research) and should explore different mechanisms for providing support, such as the Department of Energy’s ARPA-E program. Management of technology demonstration projects is a critical aspect of federal support for innovation. The Office of Management and Budget should undertake a thorough review to identify regulations that slow innovation in the areas of patents, security registration, tax provisions, and federal acquisition regulations, and then recommend changes that streamline the innovation process. The president should form an interagency council charged with overseeing the innovation efforts of different agencies and share best practices; among its efforts, the council should study and track the implications of an increasingly digital-based economy on the future of work and changing educational needs. The federal government must also continue to combat illegal theft and hacking of technology and know-how from China and other countries, as outlined above.

Finally, Congress should establish a national commission comprising political, industry, university, and public interest groups to communicate the nation’s need for advancing innovation and to report progress. The Unites States’ approach to improving its innovation capability is based on the core strength of its innovation infrastructure and the diverse and dynamic entrepreneurial enterprises that are incentivized to succeed. All elements of the public have a role to play in the process but should share a high-level vision of the importance of the task. If the United States attains its potential improvements in innovation performance, China’s great leap forward will likely be, at best, just a few steps toward closing the innovation leadership gap that the United States currently enjoys.

Beyond Patents

Over the past 20 years, patent analysis has become a dominant method of studying innovation. The number of papers containing “patent citation” listed in Google Scholar rose from fewer than 10 in 1998 to more than 900 in 2014. For papers published in seven journals of the American Economic Association since 2010, 9 of the 20 judged most relevant to innovation by the association’s search engine used patent analyses. If papers that analyze sociolinguistic, behavioral, consumer sentiment, and measurement innovations are excluded, to focus exclusively on product, process, and service innovation, the proportion rises to 9 of 16. For Management Science, the leading management journal on innovation, 11 of its 15 articles most relevant to innovation and published since 2010 used patent analyses.

For many of these papers, the focus is on the relationship between innovation and advances in science. Patents themselves are used as a measure for innovation, whereas papers cited in the patent applications are used as a measure for advances in science. For example, Science, the world’s leading generalist science journal and one that publishes few social science papers, did so in August 2017 with an article, “The dual frontier: Patented inventions and prior scientific advance.” This study concluded that “most patents (61%) link backward to a prior research article” and “most cited research articles (80%) link forward to a future patent.”

Questions about these types of analyses have been asked since long before they became the dominant method of analyzing innovation. Strengths and limitations of patent analyses as indicators of innovativeness and scientific advance—such as the fact that not all inventions are patentable and not all innovations are patented—have been thoroughly reviewed over the past few decades by scholars of innovation such as Zvi Griliches and Martin Meyer. But the mixed value of patent analysis seems to have had little influence on the growing use of patent data by those studying innovation and science. In my view this growing influence of patent analysis reflects a fundamental bias in the social sciences toward large-scale empirical analyses of quantifiable databases aimed at uncovering allegedly general economic principles, and away from studies that can reveal the complexity, context, and processes of technological innovation in the economy.

I argue that the growing dominance of patent analysis not only fails to provide valuable and reliable insight into innovation processes, but is a smoke screen that prevents social scientists and policy-makers from understanding real problems and processes of innovation. Further, excessive focus by innovation scholars on studying patents, and by academic scientists on pursuing patents, together suggest the possibility that increased patenting activity may contribute to a long-term slowdown in productivity growth. We need an alternative research agenda for studying the relations between science and innovation, and I propose the outline of such an agenda with a focus on productivity.

What good is patent analysis, and what good are patents?

It has long been recognized that most innovations are not patented and that many patents don’t represent important innovations. There are many reasons for this. One is that most innovations represent a combination of ideas, some of which represent novel patentable designs and some of which do not. Another reason is that the benefits from patenting do not exceed the costs for many types of innovations, and thus firms do not apply for patents. Patent law offers patent recipients protection in return for disclosure, but not all innovators benefit from this trade-off. The ones who do not apply for patents do not have their innovative activities counted as innovations in patent analyses, even though many of their activities are highly beneficial to the users.

For example, consider the Wall Street Journal’s “billion-dollar startup club,” new firms that are valued at $1 billion or more. These start-ups can certainly be defined as innovative, even if their values are probably inflated. Table 1 shows the percentage of start-ups by numbers of patents (as of February 1, 2018). Only 41% of the 170 start-ups had at least one patent and only 20% and 9.4% had at least 10 and 50 patents, respectively. The percentages are particularly low for e-commerce, financial services (fintech), and consumer internet, categories that include highly valued services such as ridesharing and room sharing, e-commerce for fashion, social networking, peer-to-peer loans, and mobile payments. Patent analyses miss these types of innovations and thus represent a highly skewed view of innovation in the US economy.

What economists really want to understand about innovation is its impact on productivity, because in the long run productivity growth is the most important issue for economies. Economists have long been aware that there is little correlation between total factor (or labor) productivity and total patenting numbers. This led early researchers on innovation such as Jakob Schmookler to recognize as far back as the 1960s that patents were a better index of innovative “activity” than of the actual economic output from this activity. He was concerned with what patents can measure rather than what we would want them to measure, a lesson that many innovation scholars have forgotten.

Figure 1 plots the number of patent applications, patent awards, and average productivity growth (five-year averages) over time. The numbers of US patent applications and awards were flat until the mid-1980s, when patenting activity began its current period of explosive growth. First noted by the economist Brownyn Hall in 2004, patent applications increased by more than three times between 1984 and 2003 and more than six times by 2015. Patent awards also rose by almost six times between 1984 and 2015. Reports from the World Intellectual Property Organization show similar trends; global patent applications quadrupled between 1980 and 2015 while global licensing income from patents grew even faster, rising by more than five times between 1980 and 2009 and by three times in terms of percent of gross national product (GNP).

Figure 1 also shows that US productivity growth has not increased along with patent applications and awards, suggesting that patents are not a good measure of innovation. If patenting activities were a good measure for innovation, productivity growth should have also increased, probably with some type of time lag. However, growth in US productivity has slowed since 1970, and that slowing is in addition to the slowing that has occurred since 1940. Other than short-term decreases (1980) and increases (2000), the average productivity growth after 1984 is not higher than before 1984. Furthermore, Anne-Marie Knott of Washington University has found that corporate research and development (R&D) productivity has been falling as well, with corporate revenues generated per dollar of corporate R&D spending having dropped 65% in the past 30 years, even as patent applications and awards have exploded. Understanding the reasons for a slowdown in economic productivity growth and a fall in R&D productivity should be major goals of the economics discipline, but don’t expect much from patent analyses.

Looking at Figure 1, it is tempting to advance a hypothesis that patent applications could be one reason for the lack of productivity growth. In addition to patents slowing the diffusion of information, could patent applications be a distraction for engineers, scientists, and their managers because they require large amounts of administrative work in applications and infringement cases? Since the cost of applying for a patent is about $10,000, it is not hard to imagine that if the cost of engineers, internal lawyers, and other personnel are included, the cost could easily reach $50,000 per patent. With 630,000 patent applications in the United States in 2015, this suggests the US market for patent activities that year was about $30 billion, equivalent to about half of the entire federal investment in nondefense R&D. Since economists have noted the negative impact of environmental, health, and safety regulations on productivity growth, why wouldn’t they take seriously the possibility that the administrative work of patent applications distracts the nation’s engineers and scientists, including those at universities, from the real work of innovation?

Evidence for the wastefulness of patenting can be seen in the lack of correlation between patenting and market capitalization (i.e., profitability) in Table 2. Researchers have long recognized that patents have little impact on firm profitability, and this problem still exists. The World Intellectual Property Organization reported in 2016 the top 100 organizations (both firms and universities) in terms of patent applications for four recent years (2010-2013), and PwC listed the top 100 in terms of market capitalization as of March 30, 2013. Of the top 10 patent applicants, only three were in the top 100 for market capitalization: IBM was ranked 9th, Samsung was 19th, and Toyota was 27th. Expanding the leading patent applicants to the top 100 adds only six more from the top 100 market-capitalized companies, for a total of nine companies. Looking the other way, only four of the top 10 companies for market capitalization were among the top 100 patent applicants, and they were ranked 9th (IBM), 31st (GE), 45th (Microsoft), and 65th (Google). Apple, Exxon Mobil, PetroChina, Walmart, and Nestle are not even in the top 100 even though many of them are in industries that do apply for many patents.

Proponents of patents would argue that patent citations are a better predictor of profitability because they show which patents are viewed as important by other innovators, and thus Table 2 should rank firms by the number of citations to their patents, and not the number of their patent applications. But such logic calls to mind the old adage (variously attributed to Mark Twain, Niels Bohr, and Yogi Berra) that “prediction is hard, especially about the future.” Patent citations are the same. In retrospect one can identify the important patents, but beforehand most of us have great difficulty knowing which patents will end up being heavily cited. So using patent citation to show that patents are important indicators of profitability is circular. If top companies aren’t patenting much, they must not think that patenting is important for profitability.

These observations suggest that scholars of innovation need to take seriously not only the possibility that patents are not important predictors of innovativeness, but that the increasingly intense focus by firms and universities on patenting may be distracting them from more important work.

What can patents tell us about science?

As in the recent Science paper cited above, many patent analyses treat citations of science papers in patent applications as a measure of knowledge flows. Yet scholars have long known that citations of science and engineering papers in patents are not a very good measure of knowledge flows, or of advances in science, and that informal interactions between and among university researchers and corporate engineers may be more important. Patent citations do not reflect the actual activities or thought processes of the engineers and scientists who devise novel designs and submit patents. Instead, they reflect the efforts of many participants in the patent application process to distinguish new ideas from prior ideas, including prior patents and public information. Such efforts amount to a complex sociopolitical dance among applicants and patent examiners to determine the scope of patents. Specialist patent advisers, not the scientists and engineers themselves, are necessary because expanding the scope of a submitted patent while narrowing the scope of competing patents requires a special set of skills. The patent examiner then tries to understand the novelty and limits of the claimed versus competing inventions. The result is that both patent examiners and advisers add many of the papers eventually included in patent applications, whereas the engineers and scientists who did the work may not have read many of the papers, or even have been aware of them.

This process was well illustrated in a detailed study of patent applications by Martin Meyer, who used interviews with scientists and engineers to better understand where their ideas came from. The study looked at patents for nanoscale technologies, an area of innovation that probably benefited from advances in science more than most technologies, yet the study concluded that few of the ideas came from academic papers, but rather from the independent work of the engineers and scientists. Many of the papers added by the patent examiners could not be recognized by the scientists and engineers who devised the novel ideas for the patent applications.

Nor are counts of citations of papers in patents a good measure for advances in science. For example, the types of journals cited in patents include not only scientific journals but also engineering and even management journals. Patent citation studies typically use the Science Citation Index (SCI) or the Web of Science (which includes the SCI) to identify science journals, but the SCI includes not only science journals but also engineering and business journals; even Harvard Business Review is included in the index. Including engineering journals in the list of journals used to count scientific papers overestimates the contribution from scientific journals. In a previous article in Issues (Spring 2017), I looked at 143 members of the billion-dollar startup club and found that most of the papers cited in their patents were from engineering journals published by groups such as the Institute of Electrical and Electronics Engineers or the Association for Computing Machinery, and not basic science journals such as Nature and Science, nor standard disciplinary journals. More specifically, although 18% of the 143 start-ups had patents that cited at least one paper in the SCI, only 6% of them had patents that cited a scientific as opposed to an engineering journal. Lumping engineering papers together with basic scientific papers as a surrogate for advances in science so generalizes the phenomenon of “knowledge flows” that it thoroughly undermines the value of that concept for assessing the relations between scientific advance and innovation.

I am arguing that research on patent databases may not help us understand innovation. But I am also suggesting that such research may rather be a smoke screen, hiding the true issues, problems, and dynamics of innovation behind an illusion that innovation is booming and that rising scientific and patenting activities are the reason for the boom. Many innovation scholars seem to believe that patents are more important indicators of innovation than are new products, services, and processes and their contribution to productivity improvements and company profits. Although some highly cited patents are of course extremely important and valuable, and some scientific papers are very influential, the patent obsession makes us think that such cases are everywhere, and are what drive an innovative economy, when in fact “home runs” such as transistors are rare.

This type of research is also self-serving, placing academics at the center of the economy while ignoring the other sources and forces behind new products, services, processes, and the entrepreneurs that introduce them. It enables university professors to claim that they are doing what they are supposed to do: writing papers that are cited in patents. It enables Science magazine to claim that it is the driver of innovation and productivity growth. More generally speaking, the emphasis on patents and papers reflects a much larger problem in the social sciences: an increasing obsession with large databases, sophisticated statistics, and elegant mathematics. This quantification arms race has pushed and continues to push the social science research on innovation toward more analyses of large patent databases, yet it may also be pushing it away from true understanding of how innovation works in today’s complex and varied economic settings. In part this is a classic problem of looking for one’s lost keys under the streetlight. In this case, if the keys are the complex dynamics of science, innovation, and productivity, the streetlight is patent data, and what keeps researchers looking in the same place is the academic incentive system, where publications are the sources of promotions, job security, social and intellectual status, and pay increases.

A new agenda for innovation research

If patent analysis is a smoke screen that has prevented scholars from improving their understanding of innovative processes, how should they address innovation instead? Although there are many possible alternative avenues, here I focus on a new agenda for productivity that is being advocated by a small number of scholars such as Robert Gordon and Tyler Cowen, but is largely neglected by most innovation scholars. A first set of questions involves the extent to which new products, services, and processes have emerged and are currently emerging. GNP data tell us some of this, but not anywhere to the extent that is needed. We need databases that better highlight these changes, particularly in the early years of new products, services, and processes. This includes science-based technologies such as superconductors, quantum computers, nanotechnology, synthetic food, glycomics, and tissue engineering; new forms of digital products such as augmented reality and drones; and new forms of internet services including those that are free, such as music and user-generated content.

Without data on recent products, services, and processes, it is difficult to understand what has emerged in the past 20 to 40 years, the period that is the most important to analyze. Fine-grained sales data from this and previous periods can help us understand the extent to which new products, services, and processes have been emerging and help us analyze the impact of changes in regulatory policy and in university and government R&D policy on the emergence of new products, services, and processes, particularly recent ones. A possible source of data is the growing number of market research organizations that provide regularly updated data on new technologies. The challenge is to integrate these data with traditional GNP data because this integration requires a different set of skills. Scholars must understand the technologies that form the basis for the new products, services, and processes, a research activity that is very different from statistically analyzing large databases.

Second, once we know what is coming out, we can address where these new products, services, and processes come from. Do they come from existing technologies or new technologies? If they come from existing technologies, what enabled them to emerge when they did? Was it changes in regulation, consumer demand, or cost and performance of existing technologies? If they are from new technologies, how did these new technologies emerge? What were the changes in cost and performance that enabled them to emerge, and were there advances in science that formed the basis for the new products or that helped the improvements to occur? Addressing these questions will help us understand the overall processes that lead to the emergence of new products, services, and processes, along with the factors that influence the processes. Currently, there is little agreement about where new products, services, and processes come from, a seemingly basic issue.

Third, researchers should be looking the opposite way. How are the new types of products, services, and processes that are economically important today connected to the many types of scientific research that have been funded over the past 20 to 40 years by the National Science Foundation (NSF), the National Institutes of Health, and other government agencies? Most funding agencies and analyses of this research merely focus on academic papers as an output, but what matters is the research that eventually leads to new products, services, and processes. This analysis should go beyond the standard litany of anecdotes (by now everyone knows that the Google algorithm was developed by NSF-funded researchers at Stanford University) and be able to trace the linkages between specific advances in science and the eventual products, services, and processes, including the intermediate steps of new product concepts and improvements in the cost and performance of the resulting technologies.

In a blog that accompanied the above-mentioned Science paper, the authors illustrated the importance of indirect linkages between science and technology by invoking the necessity of Einstein’s theory of relatively for the Global Positioning System and thus Uber. But such high-level, general linkages to paradigm-busting geniuses such as Einstein tell us almost nothing useful about knowledge flows, innovation, or science policy options. Einstein’s theory of relativity was published in 1916. What we need to know is the extent to which recent advances in science affect new products and services. We already know that most of the world’s products and services depend on past advances in science and that these advances are cumulative. We want to know how many advances made since 1980, 1990, or even 2000 have had a major influence on new products and services. This type of data is needed to understand the bottlenecks for innovation.

Patent analyses suggest that advances in science have been making direct contributions to every new product and service, but this conclusion seems unlikely given the successive productivity slowdowns in the US economy since 1970, even as government support for academic basic research increased more than tenfold (after inflation) between 1950 and 1980. These increases in basic science should have led to a productivity boom in the late twentieth century, one that rivals the late nineteenth century. Techno-optimists might argue that the boom is coming soon, but we need better data and analyses to test such a hypothesis. Among other things, we need such data to analyze the effects of various policies (such as the Bayh-Dole Act of 1980 that incentivized university patenting) on the linkages between advances in science and new products, services, and processes. Linking advances in science with real products and services requires a completely different form of research than is currently done in patent analyses. Rather than do large-scale empirical analyses, one must understand many intermediate linkages through detailed case studies. What types of new explanations did these advances entail? What types of new concepts or what types of performance and cost improvements in the resulting technologies did the explanations enable? What types of products, services, and processes emerged from these technologies, and what were the time lags? This type of research changes the focus from statistical analysis that searches for perhaps nonexistent general trends, to a real-world, case-based understanding of science and the emergence of new products and processes, such as the work pioneered decades ago by innovation scholars such as Kenneth Flamm on computers, Nathan Rosenberg and David Mowery on aircraft, Yujior Hiyami and Vernon Ruttan on agriculture, and Richard Nelson on transistors.

We also need to understand from which scientific disciplines the new products, services, and processes have emerged if we are to make better funding decisions and if we are to help engineering and science students make better career decisions in an ocean of hype about the value of science, technology, engineering, and mathematics (STEM) education. The current system makes little attempt to help students understand what types of innovations are occurring, what fields of science are making the most useful contributions to innovation, and thus which courses of study offer the most opportunities.

A fourth set of questions revolve around why some sectors have faster productivity growth than do others. Because most economic and management research focuses on organizational factors such as employee performance measures, incentives, and skills, and uses patents as a surrogate for innovation, the reasons for the differences between sectors is largely being missed. For example, the mechanics of Moore’s Law, the role of smaller scale, and the impact of this smaller scale on rapid improvements in cost and performance of information processing technologies are well documented, but little systematic effort has been made to look across industrial sectors and classes of technology to understand the overall impact of Moore’s Law and new materials on differences in productivity, or to identify other specific pathways of innovation on productivity. Patent analyses gloss over these details and leave us with a vague feeling that innovation is occurring, science supports this innovation, and as long as we have more of both, everything will be okay.

Innovation scholars should be trying to better understand the reasons for the productivity slowdown and how it can be fixed. Identifying and analyzing these reasons will require scholars to consider multiple types of data and information, much of which cannot be placed in a spreadsheet and analyzed with sophisticated statistics, and will not likely be found in academic journals. Scholars will have to get their hands dirty, understanding the specifics of new advances in science, new technologies, and their resulting new products and services. They will have to make judgments, create new definitions, identify new linkages, and begin building new bodies of data. Just as Charles Darwin left home to understand the real world, innovation scholars need to do the same, leaving the safety of existing databases and theories so that they can start collecting new data and generating new hypotheses. Ultimately, such research can help inform a more constructive discussion about fostering economic opportunities across all levels of society.

Forum – Summer 2018

There’s a strange pattern to popular handwringing about the skills gap. Whether one reviews pre-Great Recession headlines from 2008, when unemployment hovered around 4.5%, or those printed at the height of joblessness in 2010, the refrain is the same: workers’ lack of proper skills left them in the lurch. Today, big business sings the same song, calling on workers to “skill up!” before the robots overtake them.

Thankfully, John Alic’s scholarship has keenly kept track of this pattern, rigorously linking it to something more structural. There’s no empirical evidence that “skill”—distinct from social status and the economic means to access higher education—works as a silver bullet. That means that it’s too simple to pin low wages and lack of advancement on individual workers making bad educational choices. Alic’s early work was some of the first scholarship to trace the disconnect between skill attainment and the United States’ transition from a stable manufacturing economy to the staccato pace of service-driven economies. It illustrated that specific skills don’t lead to economic security. Sound economic policies written with workers’ best interests and rights to organize in mind do. Today, three-quarters of the nation’s workforce is employed in low-paying, dead-end service jobs, not because workers took the wrong classes at school but because labor laws do nothing to push businesses to fully value everyday workers’ economic contributions. In other words, as Alic presciently noted nearly two decades ago, if service work leads to low pay and little advancement for workers, it’s an absence of policy interventions that hold businesses accountable, not an individual skill mismatch, to blame.

Alic’s “What We Mean When We Talk about Workforce Skills” (Issues, Spring 2018) could not be timelier, as the conversation around automation and the future of work heats up for the 2018 mid-term elections. It is not uncommon for technologists, policy-makers, and economists across the political spectrum to trot out the old saw that workers’ lack of high-tech skills is the reason US wages and prosperity have stalled. Alic offers a sweeping and compelling argument challenging such unexamined conventional wisdom. As he argues, skills “overlap, combine, and melt together invisibly, defying measurement,” making explicit adjustment to them an ineffective lever for economic change. But, importantly, the attention on skills education “deflects attention from broader and deeper institutional problems in the labor market.” For Alic, placing debt and speculation about the “right skills” on the shoulders of young people (or parents helping them with schooling) harms individuals and businesses alike. Human capital, as Alic poignantly puts it, is “a national resource.” As Alic argues, one of the most important shifts we must make is recognizing this fact. This is the logical conclusion when one takes in Alic’s case for the immeasurability and unpredictable nature and value of skill as society-at-large stands at the precipice of the future of work.

Mary L. Gray
Senior Researcher, Microsoft Research
Professor, School of Informatics, Computing, and Engineering, Indiana University
Fellow, Harvard University Berkman Klein Center for Internet & Society

In “Reauthorizing the National Flood Insurance Program” (Issues, Winter 2018), Howard C. Kunreuther raises several points that are germane to the Insurance Information Institute’s mandate to “improve public understanding of insurance—what it does and how it works.” One point in particular goes to the heart of why my organization believes that the pending renewal of the National Flood Insurance Program (NFIP) presents an ideal opportunity to modernize the program. After 50 years of existence, the NFIP has become financially unsustainable (per a report from the US Government Accountability Office). Yet even with the NFIP as a backstop against a total loss, an average of only 14% of the 3.3 million households in counties declared federal disaster areas after Hurricane Irma ravaged Florida in September 2017 had NFIP coverage. Equally frustrating are the reasons why many people decline to purchase coverage, including a lack of awareness that flood-caused damage is not covered under their homeowners, renters, or business policies; underestimating flooding risk; and the cost of coverage.

Proposals in Kunreuther’s article include behavioral risk audits and other tools to better understand how homeowners, renters, and business owners view and rationalize risk; a call for Congress to provide increased investment in producing more accurate flood maps; and premising NFIP reauthorization on the adherence to “the guiding principles for insurance.” These also happen to be principles long embraced as core values of private insurance carriers, which are stepping up to expand insurance options available to residential and business customers. Recently, the NFIP and the Federal Emergency Management Agency (FEMA) have publicly embraced this sort of competition as a way of improving flood insurance for all. Indeed, the NFIP’s prospective expanded investment in reinsurance and the purchase of its first-ever catastrophe bond to contend with the realities of flood risk are positive steps toward finding private-sector solutions to existing problems.

We applaud Kunreuther’s call to make the NFIP “more transparent, more cost-effective, more equitable, and more appealing to property owners.” Moreover, we share his conclusion that a successful, more evolved NFIP would benefit greatly by building on the “support and interest of real estate and insurance agents, banks and financial institutions, builders, developers, contractors, and local officials concerned with the safety of their communities.”

Sean M. Kevelighan
Chief Executive Officer
Insurance Information Institute

Howard Kunreuther identifies a central challenge when it comes to reducing flood damage: motivating people to take action ahead of time. The systematic biases that he articulates (myopia, amnesia, optimism, inertia, simplification, and herding) apply not only to purchasing insurance but also to making life-safety and property-protection decisions before, during, and after disasters. These biases result in a serious protection gap that stymies efforts to improve the National Flood Insurance Program and hampers mitigation efforts to reduce damage from wind, hail, wildfire, winter storms, and other severe weather.

Kunreuther writes, “A large body of cognitive psychology and behavioral decision research over the past 50 years has revealed that decision-makers (whether homeowners, developers, or community officials) are often guided not by cost-benefit calculations but by emotional reactions and simple rules of thumb that have been acquired by personal experience.”

Intuition will get you only so far, and it’s not far enough to push back against the wiles of Mother Nature. To meet Kunreuther’s challenge we need well-founded, realistic, scientific testing and research on building systems subjected to real-world wind, hail, wind-driven rain, and wildfire conditions.

The Insurance Institute for Business & Home Safety (IBHS) that I lead not only takes on this challenge of sound building science, but also educates consumers on how to put that insight into action. Groundbreaking building science findings need to be translated into improved building codes and standards as well as superior design, new construction, and retrofitting standards that exceed building code requirements. Yet in translating our work into bricks-and-mortar improvements, we are confronted with the same risk biases that undermine sound NFIP purchase decisions.

The cognitive psychology and behavioral decision research cited by Kunreuther has opened our eyes to the impediments to rational risk management behavior. A discipline once confined to classrooms and conferences now fills Kindles and personal studies in books such as Misbehaving; Nudge; and (my favorite) Thinking, Fast and Slow. That’s progress, but the next step is to translate these ideas into messages and strategies that change the way people think about risk and act to reduce it. In this regard, we support Kunreuther’s idea of conducting behavioral risk audits to promote a better understanding of the systemic biases that block appropriate risk decision-making and set the stage for development of strategies that work under real-world conditions.

While at the Federal Emergency Management Agency, where I worked prior to joining IBHS, we launched two ambitious moonshots: the first, to double the number of policyholders with flood insurance, and the second, to quadruple mitigation investment nationwide. Achieving these goals will require dedicated efforts by government, the private sector, and the nonprofit community. To best cope with rain, flood, hail, wind, wildfire, earthquakes—choose your peril—we need a whole set of players to put in motion actionable strategies that, as Kunreuther suggests, “work with rather than against people’s risk perceptions and natural decision biases.”

Roy Wright
President & Chief Executive Officer
Insurance Institute for Business & Home Safety

Formerly led the Federal Emergency Management Agency’s mitigation and resilience work and served as the chief executive of the National Flood Insurance Program.

Debates about autonomous vehicles (AVs) can sometimes become quite polarized, with neither side willing to acknowledge that there are both risks and opportunities. In “We Need New Rules for Self-Driving Cars” (Issues, Spring 2018), Jack Stilgoe provides a much-needed, balanced overview of AVs, calling for increased public debate and governance of the development, not just deployment, of AV technology. We agree with many of his recommendations, particularly about the importance of data sharing with the public and other parts of the industry. We also share his concern that AVs may exacerbate social inequalities, or shape the public environment in ways that disproportionately harm the worse off. At the same time, we raise two issues that are deeply important to our understanding and policy-making, but that receive only passing mention in his article.

First, “reliability” and “safety” are oft-used terms in these discussions, but almost always without reference to the contexts in which AVs are hopefully reliable or safe. Stilgoe notes challenges presented by “edge cases,” but the importance of deployment context is not limited to unusual or anomalous cases. Rather, those contexts shape our whole understanding of AV safety and reliability. In particular, proposals for approval standards based on criteria such as “accident-free miles driven” or “number of human interventions per mile” are misguided. Such criteria are appropriate only if test conditions approximately mirror future deployment contexts, but many technologies used in AV development, such as deep networks, make it difficult to determine if test and deployment conditions are relevantly similar. We have thus proposed that the approval process for AVs should include disclosure of models used by developers to link AV testing and deployment scenarios, including the validation methodology for that model, along with credible evidence that the developer’s test scenarios include a sufficient range of likely deployment contexts and that the AV performs safely and reliably in those contexts.

Second, development of AVs requires many technological decisions, where different technological options are acceptable. As one simple example, an AV can be designed to always drive as safely as possible, or to always follow the law, but not to always maximize both values. Driving as safely as possible might involve breaking the law, as when other vehicles are themselves breaking the law (say, by speeding excessively). Moreover, the technology does not dictate which value to prioritize; the developer must decide. Importantly, there is no unique “right” answer: one could defensibly prioritize either value, or even try to balance them in some way. Nonetheless, some choice is required. Technology is not value-neutral, but rather encodes developer decisions about, for example, what counts as “success” at driving. Public discussion should thus go beyond the issues mentioned by Stilgoe, and further include debate about the values that we want AVs to embody and what AV developers should be required to tell us about the values in their particular technologies.

David Danks
L.L. Thurstone Professor of Philosophy and Psychology

Alex John London
Clara L. West Professor of Ethics and Philosophy
Carnegie Mellon University

In a rapidly changing world, communities must deal with a wide range of complex issues. These issues include adapting to a rapidly changing economy and providing a safe and healthy environment for their residents. In rural areas, addressing such issues is made more difficult because communities often lack the capacity to hire professionals and elected officials are typically stretched thin. In addition to their work with the community, leaders often have full-time employment, a family, and other responsibilities.

At the same time, many new data and other resources are available to assist communities. Data can help them better understand their current circumstances and prepare for the future. The problem is finding the time and expertise to effectively collect and analyze these data. I am convinced that the Community Learning through Data Driven Discovery (CLD3) model that Sallie Keller, Sarah Nusser, Stephanie Shipp, and Catherine E. Wotek describe in “Helping Communities Use Data to Make Better Decisions” (Issues, Spring 2018) represents an effective way to provide much-needed help to rural communities. The brilliance of the model is that it uses existing entities (Cooperative Extension and the Regional Rural Development Centers) to provide expertise that rural communities often lack. Cooperative Extension has representation in virtually every county in the country. The Regional Rural Development Centers have the capacity to make the necessary connections with Cooperative Extension in all 50 states. As director of one of the Regional Rural Development Centers, I am excited about this program and the potential benefits it provides to rural areas.

Don Albrecht
Executive Director, Western Rural Development Center
Utah State University

In “Reconceptualizing Infrastructure in the Anthropocene” (Issues, Spring 2018), Brad Allenby and Mikhail Chester offer a challenge for large-scale infrastructure designers and managers, as well as for policy-makers. It is certainly true that major infrastructure projects can have significant effects on both human and natural systems. As Allenby and Chester emphasize, we need new tools, education approaches, and management processes to deal with them.

We also need political leadership and extensive institutional partnerships to be effective in tackling major challenges and projects in the Anthropocene era. Advisory and research entities, such as the Natural Research Council (NRC) of the National Academies of Science, Engineering, and Medicine, can also be critical for success. Let me give a few examples of such leadership and partnerships.

First, building the US Interstate Highway System over the past 50 years has provided extraordinary mobility benefits for the nation. The 1956 Federal Aid Highway Act provided a vision and a funding mechanism with the Highway Trust Fund. The system implementation relied on a partnership between federal agencies and state departments of transportation, as well as with private contractors who designed and built the system. However, it was only after a period of decades that the system began to deal with the environmental and social impacts of roadbuilding, especially in sensitive natural environments and urban neighborhoods. The American Association of State Transportation Officials and the NRC’s Transportation Research Board (TRB) became other institutional partners. TRB played a major role in convening forums to incorporate more holistic approaches to roadway development and operation.

Second, the South Florida Everglades Restoration project is an ambitious effort to restore natural systems and alter the overall flows and use of water in an area of 4,000 square miles (10,000 square kilometers). Political leadership came when Congress passed the Water Resources Development Act of 2000. The effort has required active partnership among several federal agencies (Army Corps of Engineers, Fish & Wildlife Service, and National Park Service), Florida state agencies, local agencies, and various private groups. The project is still coping with issues such as legacy pollution (especially nitrates) and climate change (including sea level rise). Again, the NRC is a partner with an interdisciplinary advisory group producing a biennial report on progress and issues.

Finally, the National Academy of Engineering identified the electricity grid (as well as the Interstate Highway System) as ranking among the greatest engineering accomplishments of the twentieth century. Now, the rebuilding of the national electricity grid is under way with goals of efficiency, resiliency, and sustainability, including the incorporation of much more extensive renewable energy and new natural gas generators. Unlike the original grid development that relied on vertically integrated utilities (and utility regulators), this new rebuilding relies on market forces to involve private and public generators (including private individuals with solar panels), transmission line operators, and distribution firms. Though markets play a major role, institutional partnerships among grid operators, state regulators, and federal agencies are essential. Again, study forums such as the Electric Power Research Institute and the NRC have been instrumental in moving the effort forward.

Some Anthropocene challenges have not had effective political leadership or formed these institutional partnerships to accomplish infrastructure change. Allenby and Chester provide a good example with the “dead zone” in the Gulf of Mexico below the Mississippi River. In addition to developing new education, tools, and design approaches, we need both political leadership and new partnerships to make progress coping with large-scale problems in the Anthropocene era.

Chris Hendrickson
Hamerschlag University Professor Emeritus, Civil and Environmental Engineering
Carnegie Mellon University

In “What is ‘Fair’? Algorithms in Criminal Justice” (Issues, Spring 2018), Stephanie Wykstra does a masterful job summarizing challenges faced by algorithmic risk assessments used to inform criminal justice decisions. These are routine decisions made by police officers, magistrates, judges, parole boards, probation officers, and others. There is surely room for improvement. The computer algorithms increasingly being deployed can foster better decisions by improving the accuracy, fairness, and transparency of risk forecasts.

But there are inevitable trade-offs that can be especially vexing for the very groups with legitimate fairness concerns. For example, the leading cause of death among young, male African Americans is homicide. The most likely perpetrators of those homicides are other young, male African Americans. How do we construct algorithms that balance fairness with safety, especially for those mostly likely to be affected by both?

An initial step is to unpack different sources of unfairness. First, there are the data used to train and evaluate an algorithm. In practice, such data include an individual’s past contacts with the criminal justice system. If those contacts have been significantly shaped by illegitimate factors such as race, the potential for unfairness becomes a feature of the training data. The issues can be subtle. For example, a dominant driver of police activity is citizen calls for service (911 calls). There are typically many more calls for service from high-crime neighborhoods. With more police being called to certain neighborhoods, there will likely be more police-citizen encounters that, in turn, can lead to more arrests. People in those neighborhoods can have longer criminal records as a result. Are those longer criminal records a source of “bias” when higher-density policing is substantially a response to citizen requests for service?

Second is the algorithm itself. I have never known of an algorithm that had unfairness built into its code. The challenge is potential unfairness introduced by the data on which an algorithm trains. Many researchers are working on ways to reduce such unfairness, but given the data, daunting trade-offs follow.

Third are the decisions informed by the algorithmic output. Such output can be misunderstood, misused, or discarded. Unfairness can result not from the algorithm, but from how its output is employed.

Finally, there are actions taken once decisions are made. These can alter fundamentally how a risk assessment is received. For example, a risk assessment used to divert offenders to drug treatment programs rather than prison, especially if prison populations are significantly reduced, is likely to generate fewer concerns than the same algorithm used to determine the length of a prison sentence. Optics matter.

Blaming algorithms for risk assessment unfairness can be misguided. I would start upstream with the training data. There needs to be much better understanding of what is being measured, how the measurement is actually undertaken, and how the data are prepared for analysis. Bias and error can easily be introduced at each step. With better understanding, it should be possible to improve data quality in ways that make algorithm risk assessments more accurate, fair, and transparent.

Richard Berk
Professor of Criminology and Statistics
University of Pennsylvania

As decision-makers increasingly turn to algorithms to help mete out everything from Facebook ads to public resources, those of us judged and sorted by these tools have little to no recourse to argue against an algorithm’s decree against us, or about us. We might not care that much if an algorithm sells us particular products online (though then again, we might, especially if we are Latanya Sweeney, who found that black-sounding names were up to 25% more likely to provoke arrest-related ad results on a Google search). But when algorithms deny us Medicaid-funded home care or threaten to take away our kids through child protective services, the decisions informed by these tools are dire indeed—and could mean life and death.

Here in Philadelphia, a large group of criminal justice reformers is working to drive down the population of our jails—all clustered on State Road in Philly’s far-Northeast. Thousands of people are held on bails they can’t afford or on violations of previous probation and parole convictions, and reformers have already been able to decrease the population by more than 30% in two years—and the city’s new district attorney has told his 600 assistant DAs to push to release on their own recognizance people accused of 26 different low-level charges. But in an effort to go further, decision-makers are working to build a risk-assessment algorithm that will sort accused people into categories: those deemed to be at risk of not showing up to court, those at risk of arrest if released pretrial, and those at risk of arrest with a violent charge.

In her compelling and helpful overview of the most urgent debates in risk assessment, Stephanie Wykstra lifts up the importance of dividing the scientific questions associated with risk assessment from the moral ones. Who truly decides what “risk” means in our communities? Wykstra includes rich commentary from leaders in diverse fields, including data scientists, philosophers, scholars, economists, and national advocates. But often, the risk of not appearing in court is conflated with risk of arrest. We need to understand that communities fighting to unwind centuries of racism in practice define what level of risk their communities might tolerate very differently from how others active in criminal justice reform do.

I would posit that no government should put a risk-assessment tool between a community and its freedom without giving that community extraordinary transparency and power over that tool and the decisions it makes and informs. Robust and longstanding coalitions of community members have prevented Philly legislators from buying land for a new jail, have pushed the city to commit to closing its oldest and most notorious jail, and have since turned their sights on ending the use of money bail. That same community deserves actual power over risk assessment: to ensure that any pretrial tool is independently reviewed and audited, that its data are transparent, that they help to decide what is risky, that tools are audited for how they are used by actual criminal justice decision-makers, and that they are calibrated to antiracist aims in this age of mass incarceration in the United States.

Our communities have understandable fear of risk assessment as an “objective” or “evidence-based intervention” into criminal justice decision-making. Risk-assessment tools have mixed reviews in practice—and have not always been focused on reducing jail populations. Because black and brown Philadelphians are so brutally overpoliced, any algorithm that weights convictions, charges, and other crime data heavily will reproduce the ugly bias in the city and the society. But we also see these tools spreading like wildfire. Our position is that we must end money bail and massively reduce pretrial incarceration, without the use of risk-assessment algorithms. But if Philly builds one as a part of its move to decrease the prison population in the nation’s poorest big city, we have the right to know what these tools say about us—and to ensure that we have power over them for the long haul.

Hannah Sassaman
Policy Director
Media Mobilizing Project

Scholars, activists, and advocates have put an enormous amount of intellectual energy into discussing criminal justice risk-assessment algorithms. Stephanie Wykstra sifts through this literature to provide a lucid summary of some of its most important highlights. She shows that achieving algorithmic fairness is not easy. For one thing, racial and economic disparities are baked into both the inputs and the outputs of the algorithms. Also, fairness is a slippery term: satisfying one definition of algorithmic fairness inherently means violating another definition.

The rich garden of observer commentary on criminal justice algorithms stands in contradiction to the disinterest or even disregard of those who are responsible for using them. As Michael Jordan put it in his recent essay on artificial intelligence, “the revolution hasn’t happened yet.” Angele Christin’s research shows significant discrepancies between what managers proclaim about algorithms and how they are actually used. She found that those using risk-assessment tools engage in “foot-dragging, gaming, and open critique” to minimize the impact of algorithms on their daily work. Brandon Garrett and John Monahan show broad-reaching skepticism toward risk-assessment among Virginia judges. A significant minority even report being unfamiliar with the state’s tool, despite it having been in use for more than 15 years. My own research shows that a Kentucky law making pretrial risk assessment mandatory led to only a small and short-lived increase in release; judges overruled the recommendations of the risk assessments more often than not.

Algorithmic risk assessments are still just a tiny cog in a large, human system. Even if we were to design the most exceptionally fair algorithm, it would remain nothing more than a tool in the hands of human decision-makers. It is that person’s beliefs and incentives, as well as the set of options available within the system, that determine its impact.

Much of the injustice in criminal justice stems from things that are big, basic, and blunt. Poor communication between police and prosecutors (and a rubber-stamp probable cause hearing) means that people can be held in jail for weeks or months before someone realizes that there is no case. Data records are atrociously maintained, so people are hounded for fines that they already paid or kept incarcerated for longer than their sentence. Public defenders are overworked, so defendants have access to only minutes of a lawyer’s time, if that. These are issues of funding and management. They are not sexy twenty-first century topics, but they are important, and they will be changed only when there is political will to do so.

Wykstra’s summary of the fairness in criminal justice algorithms is intelligent and compassionate. I’m glad so many smart people are thinking about criminal justice issues. But as Angele Christin writes in Logic magazine, “Politics, not technology, is the force responsible for creating a highly punitive criminal justice system. And transforming that system is ultimately a political task, not a technical one.”

Megan Stevenson
Assistant Professor of Law
George Mason University

Opening Up the Climate Policy Envelope

Policy action is required to mitigate and adapt to human-caused climate change, but current efforts to develop a global climate policy cannot fly. What the world’s leaders have been able to agree on will not prevent the steady increase in greenhouse gases in the atmosphere and the risks of climate disruption that will result.

For an aircraft to fly it must operate within a flight envelope, the combination of conditions such as airspeed, altitude, and flight angle necessary for successful operation. For a specific approach to climate action to succeed, it must operate within a policy envelope, the combination of policy design and political, economic, technological, and other conditions necessary for the approach to be effective.

If aircraft designers sought to improve the performance of a poorly designed aircraft not by improving its design, but by rejiggering their claims about aerodynamics, or airfoil design, or jet fuel combustion thermodynamics, to match the aircraft performance they desire, it is obvious that the aircraft would still perform badly. In the case of climate change, policy-makers and climate experts are doing something similar. In the face of ongoing failure to reduce global greenhouse gas emissions, they are rejiggering the way they define the climate change challenge as if that will somehow allow policies that have been failing for over 25 years to become successful.

Understanding the unexplored dimensions of a policy envelope can be particularly important in situations of policy failure or gridlock. Sometimes new options are needed in order to break a stalemate, enable political compromise, or create new technological possibilities. The exploration of options can also give confidence that the policies being implemented do not have better alternatives. Thus, an important role for policy analysts, especially in the context of wicked or intractable problems, is to understand the ever-changing dimensions of the policy envelope in a particular context to assess what might be possible in order for progress to be made, perhaps even expanding the scope of available actions.

The failure of global climate policies to date suggests that new policy options should be explored—that we may need a significantly expanded policy envelope to begin to make satisfactory progress. But rather than exploring such options, we have instead been protecting the current policy envelope from critical scrutiny. One mechanism of such protection is via scenarios and assumptions that underlie the authoritative policy assessments of the Intergovernmental Panel on Climate Change (IPCC).

Inside the envelope

The dynamics at play are complex and somewhat circular. In the 1980s and ’90s scientists and policy-makers concluded that accumulating carbon dioxide in the atmosphere would be best addressed via an international treaty that focused on incremental reductions in emissions, based on negotiations among countries. This approach culminated in the 1992 United Nations Framework Convention on Climate Change (UNFCCC), followed by its various incremental offspring, including the Kyoto Protocol of 1997 and the celebrated Paris Agreement of 2015. The social scientist Steve Rayner of Oxford University argues that this approach incorporated “borrowings from other international governance regimes addressing stratospheric ozone and nuclear weapons as well as the USA’s experience with SO₂ [sulfur dioxide] emissions trading.” The focus was thus on an international process with successive and incremental commitments to emissions reductions made by participating parties, supported by mechanisms of measurement and verification.

An important role of climate research in this regime was to provide ever more certain justifications for action and key guidance on the details of implementation. The IPCC thus focused on providing technical support for the UNFCCC. A key element of this support has been the development of scenarios of the future that show how the world might look decades hence with and without climate policies under the UNFCCC, so that policy-makers might understand costs and benefits of proposed actions. These scenarios thus support the formulation and implementation of climate policies within a policy envelope that was established by and has been pursued under the UNFCCC.

The restricted policy envelope that results from the scenarios of the IPCC—typically formalized in the form of so-called integrated assessment models—is the result of two reinforcing sets of assumptions. One is that the costs of inaction will be high due to projected large changes in climate resulting from a massive increase in future emissions and resulting negative impacts on societies. The second is that necessary incremental actions to reduce and ultimately eliminate emissions will be technologically feasible at low cost, or even at no net cost at all—that such actions are economic and political no-brainers.

Both sets of assumptions may well prove to be correct. But what if the costs of inaction are not so high or the costs of incremental action are not so low? What if the current approach to climate policy reinforces a partisan divide and fuels its own opposition? What if there are other ways to address the challenge? What if our view has focused on a policy regime that cannot succeed? What if we need to think differently in order to succeed? How many more decades of failing to make real progress will be necessary before asking such questions is not only politically acceptable but unavoidable? What if it is then too late?

At the center of the current approach is a target and a timetable. The target is to stabilize concentrations of carbon dioxide in the atmosphere at a low level. In the past this level was commonly expressed as 450 parts per million carbon dioxide equivalents, and more recently has been expressed as a temperature target such as 2 degrees Celsius (2°C). Under the Kyoto Protocol, the timetables for emissions reduction were specified quantitatively for certain participating counties, and when this did not work the Paris Agreement allowed countries to specify targets for themselves. Neither approach has proved effective at securing emissions reductions, much less making progress toward a stabilization or temperature target.

In 2012 the physicist Robert H. Socolow argued in the Vanderbilt Law Review that the 2°C target was not so much a policy goal but rather a political motivation, reflecting “a mindset that is common to the entire exercise: to create maximum pressure for action. The action most on the minds of the proponents of ‘two degrees’ is deep transformation of lifestyles and the industrial structure of the OECD [Organization for Economic Co-operation and Development]. Implications for developing country industrialization receive little attention.” Under the Kyoto Protocol, developing countries were not expected to reduce emissions, nor are they under the Paris Agreement, at least not until far into the future. Vast parts of the world have yet to achieve the full bounty of energy services available in OECD countries, and climate politics has tiptoed around this inequity for decades. Not surprisingly, emissions have increased dramatically in parts of the world experiencing rapid economic growth.

If climate policy can’t be made to work in the real world thus far, then at least it can be made to work in the scenarios and models of the future that underpin the debate. It is Policy Analysis 101 to consider the consequences of alternative policy interventions, and economic and other types of models can often help us to productively understand these consequences and associated uncertainties. But in addition to supporting insight, models and scenarios can obstruct understanding and discourage critical thinking.

Discussion of climate policy options has long depended on the generation of scenarios of the future that include a wide range of assumptions, such as growth in population and wealth and prospects for technological innovation in energy production and consumption. These scenarios serve multiple purposes. For instance, they are used to generate projections of future greenhouse gas emissions, which can then be used as a key input for physical climate models, which in turn project future changes in climate. These projections can tell us something about the consequences of alternative interventions, and of inaction as well. Scenarios are also used as the basis for projecting the scale and scope of possible policy actions, including the projected costs and benefits of different approaches to climate mitigation.

Typically, two categories of scenarios are developed. Baseline scenarios are used to describe a future in which society continues to change, but no intentional action is taken on climate mitigation. Policy scenarios are used to describe a departure from the baselines, to describe a future in which action is taken on climate mitigation. Both baseline and policy scenarios have many assumptions about the future built into them, and both have profound implications for our view of the climate policy envelope, and by extension for which policies are considered as worth pursuing and which are not.

Scenarios are essential because to move into the future intentionally we need some expectation of how actions and outcomes may be related. But scenarios may become captured by assumptions and beliefs about how the world does or should work, and thus can limit our vision of possible futures, and make us vulnerable to surprises.

Assume a light saber

An obvious example of myopia induced by climate policy scenarios can be found today in the role played in scenarios of bioenergy with carbon capture and storage, called BECCS, a technology that combines biomass energy production with the storage of carbon dioxide. BECCS allows for “negative emissions”—the removal of carbon dioxide from the atmosphere through the large-scale growing of plants, which are then combusted to create energy, with the carbon dioxide emissions from the combustion then captured and stored, presumably in deep geological formations. BECCS technologies thus serve two important functions in scenarios of the future: they serve as a source of carbon-free energy supply needed to replace fossil fuels, and as a sink for carbon dioxide in the atmosphere. Large-scale BECCS technologies do not yet exist.

Negative emissions technologies are a relatively new addition to climate policy discussions, appearing in the academic literature in the past 20 years and then making their way into integrated assessment models about a decade ago. A 2005 IPCC report on carbon capture and storage mentioned “negative emissions” in passing and cautiously suggested it as a “possibility … [that] may provide an opportunity to reduce CO₂ [carbon dioxide] concentration in the atmosphere if this option is available at a sufficiently large scale.” The report noted that BECCS “is a new concept that has received little analysis in technical literature and policy discussions to date.” Not surprisingly, at that time BECCS was not a technology incorporated in IPCC scenarios and models of the future.

In 2007 the IPCC Fourth Assessment Report noted that “current integrated assessment BECCS scenarios are based on a limited and uncertain understanding of the technology. In general, further research is necessary to characterize biomass’ long-term mitigation potential.” Yet by 2013, such caution had been left far behind, and negative emissions were central to nearly all scenarios of the IPCC Fifth Assessment Report that are compatible with a 2°C target. In less than a decade negative emissions went from an afterthought to being absolutely essential to international climate policy. No government had actually debated the merits of BECCS, there were no citizen consultations, and very little money was being devoted to research, development, or deployment of negative emissions technologies. Yet there it was at the center of international climate policy.

In a 2013 paper, Massimo Tavoni and Socolow observed that negative emissions first appeared in the scenarios of the IPCC Fourth Assessment Report associated with new demands for increasingly strict emissions targets, such as 2°C. “Subsequently, many models have incorporated [carbon dioxide removal], thereby increasing the options for achieving stringent climate targets,” they noted. “Thus, paradoxically, despite little progress in international climate policy and increasing emissions, long-term climate stabilization through the lens of IAM [integrated assessment models] appears easier and less expensive.” An option was created, not in the real world, but in models that sustain the current policy envelope.

The result was to make business-as-usual climate policy under the UNFCCC appear to be on track. The paradoxical result, as Tavoni and Richard S. J. Tol noted in 2010, was that demands for more stringent climate targets were accompanied by greater apparent policy feasibility at lower cost. “In order to be able to satisfy this new policy demand,” they said, “the models that have analyzed the more ambitious policies have been pushed towards implementing more optimistic assumptions about the range and availability of their mitigation portfolio, which has the effect of lowering the costs of climate policies.” Because the proposed technologies were speculative and at best well off into the future, estimates of the costs and feasibility of their implementation could be tailored to the needs of sustaining the policy regime. Peter A. Turner and colleagues have observed that whereas “BECCS appears to be cost-effective in stylized models, its feasibility and cost at scale are not well known.” Of course not. If nothing else, full implementation of BECCS “at scale” would require the use of a global land area one and a half times the size of India (land that will therefore not be available for agriculture or other uses). In the absence of any justifiable method for predicting actual costs, why not just assume that BECCS will be affordable?

BECCS technologies are fundamental to the new family of IPCC scenarios designed to take climate policy discussion into the 2020s. Sabine Fuss and colleagues observe that some extreme scenarios foresee BECCS accounting for more than 1,000 gigatons of carbon dioxide over the twenty-first century, with a median removal across the new generation of IPCC scenarios of 12 gigatons per year. For comparison, in 2017 carbon dioxide emissions from fossil fuels emissions worldwide totaled about 37 gigatons. Glen Peters, of the Center for International Climate Research in Oslo, Norway, notes that in IPCC’s forthcoming Sixth Assessment Report, BECCS plays a central role and can be found in 78 of 80 of its new family of policy scenarios.

The Paris Agreement on climate change does not mention BECCS technologies, yet achievement of its goal to hold temperature increases to below 2°C is fully dependent upon them in the IPCC’s scenarios of policy success. Wilfried Rickels and colleagues concluded in 2017, “Achieving the 2°C and even more the 1.5°C goal is unrealistic without intentional atmospheric carbon dioxide removal.” Take away the speculative technology embedded across scenarios and models and the entire policy architecture of the Paris Agreement and its parent, the UNFCCC, falls to pieces, just as would an aircraft fatally outside its flight envelope. Oliver Geden, of the German Institute for International and Security Affairs, argues that scenarios thus hide policy failure: “By establishing the idea of negative emissions, climate researchers have helped, unintentionally, to mask the lack of effective political mitigation action.”

Carbon dioxide removal at massive scale is science fiction—like a light saber, incredible but not real. Yet BECCS plays a very real role in today’s climate policy arena, by helping to maintain the climate policy envelope and save us from having to do the enormously difficult and uncomfortable work of thinking how we might go about addressing accumulating carbon dioxide in the atmosphere differently than we have since the 1980s and ’90s. Yet BECCS is not the only important assumption in scenarios that serves to limit the scope of the climate policy envelope.

Assume you’re running downhill

Another key assumption found in scenarios and models used by the IPCC has to do with the rates at which the global economy will become less energy intensive (for example, through gains in energy efficiency, and the dematerialization and increasing productivity of economic activity) and how fast global energy production will become less carbon intensive. Such decarbonization (technically, the ratio of carbon dioxide emissions to global gross domestic product) occurs in the absence of climate mitigation policies, and is thus termed spontaneous decarbonization. It is an important feature of baseline IPCC integrated assessment scenarios for the future.

Policies to mitigate climate change seek to accelerate the rate of decarbonization of the global economy, a job made easier to the extent that the world is also decarbonizing due to normal processes of technological change. Small changes in assumptions of future rates of such spontaneous decarbonization can thus have an enormous impact on the role for and costs of future climate mitigation policies.

Consequently, assumptions of spontaneous decarbonization are essential to understanding the magnitude of the challenge of stabilizing concentrations of carbon dioxide in the atmosphere. For instance, a 2017 report by the PwC corporation, based in the United Kingdom, estimated that the global economy will have to decarbonize at a rate of 6.4% per year through the end of the century if the 2°C target of the Paris Agreement is to be achieved—where 6.4% is the sum total of spontaneous and intentional decarbonization. From 2000 to 2016 the world decarbonized at a rate of 1.4% per year. That leaves a decarbonization gap of about 5% per year between recent rates of decarbonization and that demanded by a 2°C target. Simple mathematics means that every year the world fails to achieve the needed rate of decarbonization, that gap increases.

The size of the 5% decarbonization gap can be instantaneously much reduced in the future if for the remainder of the century we assume that instead of the 1.4% rate of decarbonization observed so far this century, the world will spontaneously decarbonize at, say, 2.4% or 3.4% or even more per year, in coming decades. For instance, if we assume future spontaneous decarbonization of 3.4%, then the decarbonization gap to be addressed by new climate policies would be just 3% per year. Of course, such optimistic assumptions might not be correct, as the future is uncertain. The point here is not that such assumptions of the future are necessarily wrong or not justifiable, but rather that they represent a subset of what might be possible. We can modulate how much policy ambition might be needed to stabilize carbon dioxide in the atmosphere simply by changing our assumptions for future spontaneous decarbonization.

Tom Wigley, Christopher Green, and I observed a decade ago that the vast majority of projected emissions reductions found in the IPCC scenarios produced in 2000 came from assumptions of spontaneous decarbonization, and not from climate policies. Many such assumptions were far more aggressive than what had been observed in the recent past. This result was consistent across all the IPCC scenarios. The reliance on spontaneous decarbonization to carry a large proportion of future emissions reductions was repeated in the next generation of IPCC scenarios, which were the basis of IPCC’s Fifth Annual Assessment, and is once again found in the newest scenarios to inform the Sixth Assessment.

Before there was BECCS, assumptions of spontaneous decarbonization did a lot of the work of reducing future emissions in climate scenarios and models. And as with BECCS, assumed rates of spontaneous decarbonization necessary to achieve desired results had yet to occur in the real world but were assumed to be reasonable in the future. Today with scenarios and models having both BECCS and spontaneous decarbonization in them, progress toward reducing emissions seems to be inevitable and easy—in imaginary future worlds.

Moreover, spontaneous decarbonization and BECCS serve a sort of Goldilocks role in climate policy. More aggressive assumptions about each of them can make the emissions problem largely or entirely go away on its own. Less aggressive assumptions would indicate the impotence of current approaches and suggest potentially higher costs in the future to mitigate emissions. And so, magically, the IPCC scenarios assume just the right amount of each in order to preserve the plausibility of the climate policy envelope that has characterized the issue for almost 30 years.

In 2007, Ian Sue Wing and Richard S. Eckhaus called such assumptions a “fudge factor” that “allows the results of climate-economy simulations to be tuned according to the analyst’s sense of plausibility.” Of course, some assumptions found in scenarios will likely prove to have correctly anticipated the twenty-first century. But shaping policy to address a limited set of futures represents a bet on those futures, and against other possibilities. The risk in such a bet is that the future plays out differently than assumed and the policies designed for that narrow set of futures prove not to address the problems that motivate the policies to begin with.

Predicting a different past

A third example of a key assumption used to reinforce the boundaries of the business-as-usual climate policy envelope has been the heavy reliance on a particular emission scenario, called RCP (Representative Concentration Pathway) 8.5, in climate impact studies. The RCP 8.5 scenario is based on an assumption of the dramatic expansion of coal energy around the world over the twenty-first century that results in extremely high carbon dioxide emissions, in fact the most emissions of any scenario used by the IPCC in its Fifth Assessment Report. Despite its outlier status, RCP 8.5 is the most commonly used scenario in climate impact studies, appearing in thousands of academic papers. Climate impact studies generally use physical climate models to project how the climate system might change in the future, with particular attention to such phenomena as heat waves, floods, drought, and hurricanes. The use of an extreme emissions scenario yields larger and more significant changes to climate in the future. The characterization of an extreme scenario as “business as usual” implies that it is a baseline scenario, a vision of what is likely to happen in the absence of climate policies.

However, RCP 8.5 is not a business-as-usual scenario and has been criticized for its unrealism. Justin Ritchie and Hadi Dowlatabadi concluded in 2017 that “evidence indicates that RCP 8.5 does not provide a physically consistent worst case BAU [business-as-usual] trajectory that warrants continued emphasis in scientific research, it does not provide a useful benchmark in policy studies.” In early 2017 the team of researchers responsible for producing the scenarios that will underpin IPCC’s Sixth Assessment observed that emissions consistent with RCP 8.5 “can only emerge under a relatively narrow range of circumstances.”

Yet RCP 8.5 remains a scenario favored in most climate impacts studies published in the academic literature. One reason for this is obvious: because the scenario generates very high carbon dioxide emissions, the associated climate impacts projected in climate models can also be very large, and thus lend continued urgency to calls for emissions reductions, and supporting economic models that show very high costs of future climate change impacts.

RCP 8.5 can also be deployed to demonstrate large impacts of climate change today. For instance, following the disastrous flooding associated with Hurricane Harvey in Houston, Kerry Emanuel, a researcher at the Massachusetts Institute of Technology, published a paper in the Proceedings of the National Academy of Sciences concluding that floods such as Harvey’s had become six times more likely in the past 25 years due to greenhouse gas emissions. The paper’s conclusion was widely covered in the media.

To generate these large changes over the past 25 years, Emanuel used a unique methodology that relied on no actual data from the past. Using climate model simulations based only on the RCP 8.5 emissions scenario, the paper projected how climate impacts associated with hurricane-related flooding might change to 2100. The study then calculated how much change had occurred over the recent past by assuming that the changes that have occurred to date fall on the same trend line that RCP 8.5 projects for the future.

Why describe the impacts of greenhouse gases on hurricane-related flooding that have already happened by using a particular scenario for 2100, when those impacts can be empirically described by enormous amounts of actual data, from which empirically validated trends can be determined? Consider that if Emanuel’s study had used a scenario for 2100 that had emissions stabilized at a low level, then the change in the risk of flooding like that of Hurricane Harvey over the past several decades in the study would have been minimal or none at all. The choice of scenario for 2100 thus changes how our recent history is viewed. In fact, empirical research on long-term trends of hurricane-related heavy rainfall and flooding in the United States has found no trends.

Dramatic findings of climate disasters resulting from greenhouse gas emissions serve an important function in helping to maintain the boundaries of the business-as-usual climate policy envelope. Large projected impacts typically imply large future costs of changes in climate under baseline scenarios. In a benefit-cost exercise, the avoidance of large projected costs is a benefit of climate mitigation policies. Carbon dioxide removal, spontaneous decarbonization, and large future impacts under RCP 8.5 thus all work in the same direction in cost-benefit analyses, but on different sides of the equation. Indeed, the larger the projected costs of inaction under RCP 8.5, the larger the allowable set of scenarios under which climate mitigation policies can provide a positive benefit-cost ratio (above 1.0).

Under scenarios that generate impressive benefit-cost ratios across the twenty-first century, an unwillingness to support climate policies with high short-term costs but higher long-term benefits looks economically irrational at best, a disastrous manifestation of “climate denialism” at worst. Using RCP 8.5 to project future climate impacts can help us understand a potential worst-case scenario, but using it as a generic business-as-usual scenario thus contributes to the toxic politics of climate policy. It feeds into suspicions that scientists are putting their thumb on the scale of climate models in order to generate projections that emphasize the more dire possible futures of climate change, however unlikely these may be.

The three key assumptions discussed above—negative emissions, spontaneous decarbonization, and reliance on RCP 8.5 as climate policy business as usual—are far from the only ones that can be used to modulate or restrict the boundaries of the climate policy envelope. For example, as I’ve explained elsewhere, one can help meet emissions targets in emissions scenarios simply by assuming that hundreds of millions of people in developing countries will fail to achieve significant levels of energy access.

But we cannot assume away the failure of climate policy today. Climate policy business as usual has been accompanied by an increase in global fossil fuel consumption of almost 60% since the 1992 Rio Earth Summit that gave rise to the United Nations Framework Convention on Climate Change, as well as to a corresponding increase in carbon dioxide emissions. What the world is doing is not working.

Climate denial of another kind

Some observers have pointed out the obvious. For instance, in 2012 Robert Socolow warned, “No one appears to be preparing for a time—possibly quite soon—when a consensus develops that a peaking of emissions in the 2020s will not occur and that therefore (at least in this meaning) ‘two degrees’ will not be attained.” Yet rather than open up discussion of climate policy to new possibilities, the main response to such observations has been climate denialism of another sort, manifested in the Paris Agreement’s call for a more stringent target (1.5°C), made seemingly feasible by the incorporation of assumptions about the future that are at best wildly optimistic.

We need to break free of such assumptions in order to recognize that the current policy envelope does not contain the pathways to meaningful progress, but rather is an obstacle to discovering such pathways.

If the IPCC is unable or unwilling to consider a more expansive climate policy envelope, then others in leadership positions might explicitly take on this challenge. It won’t be easy. Business-as-usual climate policy has a large and powerful political, economic, and social constituency. Repeated policy failures, most obviously the Kyoto Protocol, have been insufficient to motivate a change in thinking or direction. Although the Paris Agreement helpfully abandoned pretentions of a top-down fix, it did little to change thinking about how its targets were to be achieved. And whereas it’s easy to blame the intransigence of the United States for lack of progress, such a tack is just another way to try to protect business-as-usual policy, for the fact is that the rest of the world isn’t making progress either.

The work on the IPCC’s Sixth Assessment Report looks to be similar in form and function to that of past reports, designed to support the UNFCCC but certainly not to open up new possibilities that might require different institutional arrangements.

An expansion of the boundaries of a climate policy envelope is different from a search for specific solutions to a narrowly defined problem. Rather, it represents a search for circumstances under which alternative, effective policy interventions might be possible. In the best cases such an exploration can result in practical options previously not considered, and in new coalitions of actors coming together in new political arrangements to seek progress.

What might an exploration of a more expansive climate policy envelope look like? Below are some questions that push toward an expanded set of options, but once we set our collective attention to the task, no doubt a dramatic expansion of ideas and possibilities would multiply quickly.

• What do climate policy options look like if BECCS is not assumed in scenarios and models?
• What happens if we abandon the 2°C temperature target? Oliver Geden observes: “Worldwide, there has been almost no questioning of the parties’ intention to hold the temperature increase to below 2 or 1.5°C.” What alternative long- or short-term targets might be used to track climate policy progress?
• One possibility might include a commitment to the expansion of carbon-free energy in national energy mixes, as achieving zero emissions will require that almost all energy consumption come from sources that are carbon neutral. The world currently is at less than 15% carbon-free energy consumption. What might a technology-focused climate policy architecture focused on targets and timetables for the adoption of carbon-free energy sources look like (rather than emissions or temperature targets)?
• Succeeding in the stabilization of carbon dioxide at low levels in the atmosphere will require a massive reduction in the use of fossil fuels. There has been essentially no serious international policy focus on how this might actually be done. Consider that the world consumes more than 11,000 million metric tons of oil equivalent (MTOE) of fossil fuels each year, according to the multinational energy corporation BP. If this number is to approach zero, then the world would need to retire and replace about 1 MTOE each day until 2050. That is the equivalent of more than a nuclear power plant’s worth of carbon-free energy, every day. How might the world decommission such a magnitude of fossil fuel energy? The UNFCCC policy envelope has been an exercise in avoiding this question. What would it mean to get serious about answering it?
• The massive scaling of technologies that do not yet exist or do not exist at scale would require a commitment to dramatically enhanced national and international innovation policies. What policy options would support innovation at the scale needed to transform the global energy system? Are there innovation investments or practices that would be amenable to targets and timetables? Above all, what magnitude of investments is likely to be necessary for a massive scale-up in carbon-free technologies?
• Climate policy discussions have tended to emphasize worst case scenarios of the future. What might climate policy look like if scenarios expected to represent more likely futures are placed at the center of climate policy discussions? How might costs and benefits look under such scenarios? What new policy options might become politically plausible with changes in predicted costs and benefits focused on central tendencies and not extremes?
• What might climate policy look like if costs and benefits of proposed policies are not calculated over decades and longer (e.g., under assumptions of future spontaneous decarbonization), but instead are examined from a perspective of one or several years, so as to be more consistent with political calendars?

These a just a very few possibilities for the sorts of questions that might be asked that would lead to an expanded climate policy envelope. Additional challenges and opportunities for an expanded policy envelope could come from consideration of factors such as non-carbon climate forcings, land use change, adaptation, global energy access, and expanded consumption, or the many other dimensions of energy and climate policies, any of which might add fruitful new options to the policy envelope.

One possible outcome of such an exercise could be that we will quickly learn that we don’t currently know how to fully address the challenge of stabilizing concentrations of carbon dioxide in the atmosphere at a low level. There is nothing wrong with acknowledging ignorance, and doing so can be an important motivation for a search for new, more effective approaches that emphasize a variety of first possible steps toward meeting a difficult challenge, rather than continuing to focus, as we have been doing, on a narrow pathway toward its completion. As a group of my colleagues wrote in “The Hartwell Paper,” in the aftermath of the disastrous 2009 Copenhagen climate conference, “for progress to occur on climate policy, we must reframe the issue in a fundamental way.”

We don’t know much about the scope of the climate policy envelope because we have done little to explore its dimensions since it was first locked in more than 25 years ago. Climate policy business as usual means that we go exactly where we have been headed, repeating the same behavior, and modifying our assumptions to accommodate our continuing failure to make progress.

The uncomfortable alternative is to open up debate beyond the constraints imposed by scenarios and models that have reinforced boundaries on the discussion of policy options, and begin to explore a future that is at once more daunting and more rich with opportunities for making progress.

Forum – Spring 2018

Helping fathers be parents

The law is evolving in response to behavioral science. If Criminal Law 1.0 was capital punishment for all felonies and mutilation of offenders, and Criminal Law 2.0 meant due process, public defenders, and “humane” incarceration, we are now implementing the more effective Criminal Law 3.0—restorative justice practices, treatment courts for a growing range of offenders, and probation officers trained in “motivational interviewing”—which essentially asks what do you want to accomplish in this probation and how can we help you?

Kathryn Edin’s field research has educated us all about unmarried parenting and poverty. Now she is using her knowledge of life on the ground to give us the most comprehensive program I have yet seen for Child Support 3.0, and her ideas are clearly presented in “Child Support in the Age of Complex Families,” (Issues, Winter 2018).

A wise friend, a psychologist, once told me, “The normal response of a healthy adult when faced with coercion is to resist.” The truth of this statement has been borne out to me time and time again over 20 years on the bench. By the time I have to order someone to do something, the battle is just about lost. Far better to ask, What do you want to accomplish at this point in your life and how can we help you? This approach, soundly grounded in behavioral science, undergirds the 3.0 wave of legal processes.

It is not just the court system that is following behavioral science to develop the wave of “3.0” versions. Consumers now can easily research any medical condition, and doctors are finding that issuing “doctor’s orders” is less effective than asking people about their health goals and pointing out the tools to accomplish them. My wife is a psychologist, and the continuing education brochures I keep seeing in our mailbox advocate mindfulness training for every variety of mental distress from anxiety to addiction to pain.

Business too, always on the lookout for effectiveness, is well into the 3.0 wave. Bottom-up engagement is proving more successful than top-down commands. I just finished reading the book about the fabulous success of Bridgewater Capital, Principles, in which the author, founder Ray Dalio, keeps coming back to his fundamental principle of “radical truth and transparency.”

The Co-Parent Court Edin uses to illustrate the value of empowerment and respect was started because some of us in Minneapolis were dissatisfied with the Child Support 2.0 process that summoned droves of young men into court to tell them, “Congratulations, you are the father. Here is your child support order.” We suspected it would be much more productive to ask them what kind of father they wanted to be and how could we help.

The carefully evaluated results demonstrate that behavioral science applies to parenting just like everything else. My biggest fear in starting the program was that people would blow us off—they just wouldn’t come. Instead, in a highly unstable population, two-thirds of the parents completed the workshops. And I worried that we would push troubled people together and just foment conflict. Instead, most of the parents worked out comprehensive parenting plans together. Just as Edin predicted, the vast majority of the low-income men I encountered, despite all the employment, health, housing, and legal issues in their lives, very much wanted to take pride in being good fathers. We just empowered and respected them.

The motivation behind Child Support 2.0—formal legal processes to try to coerce men into being responsible fathers—was always admirable. But now we just know better.

By the way, Edin mentions in passing the “withering” of the current welfare (TANF) system, which is based on version 2.0 sanctions to coerce work. What would Welfare 3.0 look like? How about asking a parent what do you want to accomplish for your family in the next year and how can we help you financially to do it?

Bruce Peterson
Judge, Fourth Judicial District
Hennepin County, Minnesota

Fostering economic growth

Robert Solow famously developed the field of growth economics by demonstrating that what he termed technological and related innovation was the dominant causative factor in economic growth. Only 211 years after publication of the iconic Wealth of Nations, Solow in 1987 was awarded the Nobel Prize in Economic Sciences for finally identifying the long-invisible monster in the economics room: a demonstrated theory of economic growth. But there was a catch-22 imbedded in his efforts; he found that economic growth was “exogenous” to the approaches of his still-dominant economic school of neoclassical economics. The elements and variables behind innovation were simply too complex to fit within more simplistic, metrics-driven neoclassical theories. Of course, an economics school without a functioning theory of growth appeared entirely unacceptable to many, and a group of “New Growth Theory” economists, led initially by Paul Romer, worked to make growth theory “endogenous,” to put it into an analytical, neoclassical box. But this has proven to be a very hard problem, and economists have gone off on more manageable projects such as behavioral economics.

Gregory Tassey remains one of the small number of card-carrying economists still pursuing economic growth policy. His succinct article “Make America Great Again” (Issues, Winter 2018)—which is more of a cri de coeur—lays out an analysis of the failure of the past decade of national economic policy, particularly the past year of it, to focus on the underlying necessities for renewal of American economic growth. As he notes, the nation’s economy has been beset with low growth—and behind that is low productivity growth and behind that is low investment in capital plant, equipment, and technology. This low growth is breaking us apart: there is a dramatic increase in income and asset inequality and a declining middle class in a nation founded on the ideas that everybody gets better, the next generation is better off than the last. This economic success has been at the heart of America’s democratic experiment. But now we seem to be systematically striving to leave our working class behind by failing to advance a broad-based, innovation-based growth agenda that might create the societal resources that could put them ahead again.

Tassey finds fault with the political parties, neither of which seems to “get” the basics of the growth economics that he lays out. How did the political parties completely miss growth economics? As John Maynard Keynes famously wrote, “Practical men who believe themselves to be quite exempt from any intellectual influence, are usually the slaves of some defunct economist.” Our political parties appear to have locked-in long ago on classical economics. The politics of each is organized around one of the two dominant factors that classical economics thought (wrongly) was behind growth: capital supply and labor supply. Republicans have focused on capital supply, with its leaders returning again and again to the popular political well of lowering marginal tax rates. Democrats focus on labor supply: improving education, health, and income in labor markets. Neither of these are bad pursuits; they are still significant. But they miss the monster in the room that Solow and Romer—and Tassey—want us to understand: technological innovation and its role in technology-driven growth.

In 2005 the economist Benjamin Friedman’s noted book, The Moral Consequences of Economic Growth, showed from international studies that periods of higher economic growth tend to be accompanied historically by more tolerance, optimism, and egalitarian perspectives, while periods of declining economic growth are characterized by pessimism, nostalgia, xenophobia, and violence. Today, though the American upper middle class is doing fine, the remainder, as Tassey’s data indicate, has been in decline. Unless growth agendas such as Tassey’s are followed, we’re in for a difficult time; we now are seeing that the social externalities of economic well-being are affecting the American working class.

William B. Bonvillian
Lecturer at MIT

Author (with Peter L. Singer) of Advanced Manufacturing, The New American Innovation Policies(2018)

Coal facts

In “President Obama’s War on Coal? Some Historical Perspective” (Issues, Winter 2018), Charles Herrick offers some helpful insight, but he misses a key set of developments. In 1987, Congress passed the Powerplant and Industrial Fuel Use Act, repealing an existing federal ban on the use of natural gas for new power generation—a ban that had been instituted in part to promote coal. This repeal, combined with natural gas price deregulation, the advent of wholesale electric competition, improvements in gas generation technology, and finally the onset of unconventional gas exploitation through fracking, opened the floodgates for inexpensive natural gas-fired power generation. Today, natural gas fuels over a third of US power generation, roughly equal to coal’s share, which is down from roughly 50% a decade ago.

In short, this is largely a story of market forces and technical innovation, not regulation, much less a “war on coal.” An April 2017 report by the Columbia University Center on Global Energy Policy titled “Can Coal Make a Comeback?” found that competition from natural gas accounts for about 50% of coal’s erosion in the past decade; another 26% comes from lower than expected energy demand that reduced headroom for the more-expensive-to-operate coal plants; and about 18% stems from wind and solar, which steadily improved in cost and performance over this period, aided by federal tax policies supporting research and development. (The same market forces have challenged US nuclear power generation, but no one complains of a “war on nuclear.”) In contrast, the Columbia study found that less than 10% of coal’s loss was due to environmental regulation of harmful air pollutants—regulations, by the way, that have already saved tens of thousands of lives, reduced neurological damage to children, reduced ecosystem damage, and improved visibility. As for prospective regulation, most analyses of President Obama’s Clean Power Plan have concluded that much of the anticipated emission reduction would already be achieved due to these ongoing trends; the rule simply locks in and makes enforceable these early reductions.

Beyond the Clean Power Plan, in a world more constrained by carbon, coal can continue to play a role by adopting carbon capture utilization and storage (CCUS) technology—at both new plants and as retrofits on existing coal-fired plants. CCUS technology has been commercially demonstrated in Texas and Canada. Congress, with the support of a broad coalition of environmentalists and industry, recently approved tax credits for CCUS projects roughly equivalent in value to the wind production tax credit.

Though some people may find it convenient to blame regulation for the problems of the US coal industry, free markets and American ingenuity are the real culprits, if that is the word. If coal is to have a significant future, more technical innovation is essential.

Armond Cohen
Executive Director
Clean Air Task Force
Boston, Massachusetts

Coal is abundant, with America possessing a 200-400-year supply, and is more reliable and less expensive than every other source of electric power except natural gas. Competition with natural gas plants is the primary reason why many coal-fired power plants have recently closed or are slated for early retirement. But if one is concerned about price volatility, coal beats natural gas in many instances as a long-term fuel source because its price is less prone to rise or fall in response to weather or surging demand from alternative uses. Coal can compete under fair conditions.

Charles Herrick’s article indicates that he believes it misleading to call former President Obama’s actions vis-à-vis coal a “war on coal.” I disagree. Nearly a year into his presidency, Obama’s Environmental Protection Agency (EPA) issued an endangerment finding ruling carbon dioxide, the gas plants need for life and every human and animal exhales, a danger to human health or the environment. Never before had EPA found a naturally occurring chemical dangerous at levels that have no toxic effect. During his tenure, Obama also successfully pressured Congress to increase the subsidies to wind and solar power plants and directed agencies such as EPA to expand their regulatory authority to tighten regulations on coal-fired power plants. Combined with competition from natural gas, these regulations and subsidies caused the premature closure of more than 250 coal-fired power plants nationwide.

President Donald Trump is in the process of reversing many of these policies by withdrawing the United States from the Paris climate agreement and rescinding the Obama era clean power plan. However, additional federal and possibly state actions are required to fully level the playing field and provide electric utilities with coal-fired generating units some degree of confidence that they will be allowed to compete fairly in the marketplace and thus should keep vital power plants open.

Ultimately, Congress should eliminate all energy subsidies. Absent subsidies, wind and solar power would largely disappear from the marketplace, being less reliable and more expensive than other sources of power. Coal, hydropower, natural gas, and nuclear power would thrive, competing solely on the basis of reliability and price, rather than government favors. This action would also strongly encourage state legislators to repeal renewable energy mandates. Absent generous federal subsidies and tax credits, electric power users would have to pay the full costs for renewable power. The result would be a dramatic increase in the price of power in states with renewable mandates. The howls of outrage over the price shock would likely cause state legislators to repeal renewable mandates.

Most important, Congress should bar the EPA from regulating carbon dioxide emissions unless and until it adopts a law specifically regulating carbon dioxide by name.

Sterling Burnett
Senior Fellow
The Heartland Institute
Arlington Heights, Illinois

Reinvigorating nuclear energy

In “How to Reinvigorate US Commercial Nuclear Energy” (Issues, Winter 2018), Steven Aumeier and Todd Allen make a compelling and timely case about the importance of the nuclear energy sector in the United States and why it should be encouraged.

A broad range of imperatives—including national commitments to reduce air pollution and carbon emissions and the need to meet growing energy demands, diversify energy supplies, ensure long-term price stability, and conserve land and natural resources—are driving an increasing number of countries to embark on nuclear energy development programs. Unlike during the 1970s and ’80s when the first global wave of nuclear construction took place, there are now multiple alternatives to US suppliers and several nations to which aspiring nuclear programs can turn to acquire civilian nuclear reactor technology.

Two nations in particular, Russia and China, have demonstrated their intent to employ nuclear energy exports as a means of expanding and strengthening their global influence. As Aumeier and Allen show, these nations are pursuing aggressive nuclear export strategies because they understand the long-term influence that comes from building a nuclear power plant in another nation. A nuclear plant is an extremely long-lived piece of energy infrastructure, designed to operate for 60 to 80 years, possibly more. So when a country sells a nuclear reactor to another nation, that transaction marks the beginning of what can be a century-long relationship. And the relationship is not just a commercial one; as I saw firsthand when I led the US Department of Energy’s nuclear energy program, bilateral relationships forged through commercial nuclear energy span a wide range of areas including education, training, safety regulation, environmental protection, physical and cyber security, and nuclear nonproliferation. These relationships lead to the types of long-term alliances that have been the hallmark of US global leadership.

As Aumeier and Allen demonstrate, there is no question that the best outcome for the US economy, national security, and global standing is for the United States to reestablish itself as the nuclear energy supplier of choice in burgeoning markets in Asia, the Middle East, and elsewhere. Given that the private US companies pursuing global nuclear energy opportunities are competing against state-owned enterprises, the US government must concern itself with helping US firms compete—by expanding investments in clean energy research, development, and demonstration to ensure that the nation retains its place as the leader in nuclear energy innovation; by restoring a functional export credit agency; and by creating opportunities for US firms through government-to-government agreements. And finally, the nation must preserve its nearly 100 domestic commercial nuclear reactors. Keeping these reactors running isn’t just a crucial step in preserving America’s largest source of carbon-free electricity generation. The decades of safe, efficient, and reliable operation of these reactors have set the global standard for nuclear power operations. The United States will not be seen as a credible global nuclear leader if it allows this domestic nuclear fleet to atrophy.

John F. Kotek
Vice President for Policy Development and Public Affairs
Nuclear Energy Institute
Washington, DC

Steven Aumeier and Todd Allen make a timely and convincing case for the need to reexamine the present export rules and regulations toward China that impede US industry in competing in the global nuclear energy market.

Clearly, the dwindling domestic market will not be able to support a robust nuclear industry. It is also arguable that from a national security perspective, the United States needs to maintain a viable civilian nuclear industry to retain its influence in the nonproliferation arena.

BP Energy has projected that over the next 20 years three-quarters of the new nuclear reactors built will be in China. China has 36 operating plants today, but the number will double in 20 years, and by 2050 China will have over 100 plants, the largest operating fleet in the world.

Currently, the United States still has the most sought after advanced technology, but US nuclear exports to China have been severely curtailed by federal export regulation called 10 CFR Part 810. The Part 810 approval process can take more than 300 days, according to a recent study by the Nuclear Innovation Alliance.

The 810 regulation was instituted in the 1950s mainly for nonproliferation purposes when the United States was the only exporter of nuclear power. Today, US competitors include France, Russia, Canada, Korea, and China. Importantly, it is generally recognized that the widespread use of light-water reactor (LWR) technology globally has not created a corresponding proliferation threat. On the contrary, the countries that adopted US civilian nuclear technology have less of a tendency to pose a proliferation threat amid continuous engagement with the United States.

The export regulation toward China has been inconsistent. While the US government has granted permission for the transfer to China of the most advanced technologies—such as Gen III+ reactors, AP1000 reactors, traveling-wave reactors, and molten salt reactors—China is still subject to more stringent review with a case-by-case specific authorization on all commercial activities, large or small. Other countries competing with the United States do not impose the same scrutiny and restriction on China.

It is time for the United States to adopt a coherent export policy toward China. The two nations must build a stronger and trustworthy bilateral relationship in the civilian nuclear program to help ensure continued nonproliferation engagement.

The United States is running the risk of being a lame duck in the sphere of nonproliferation if it stays on its current course regarding nuclear export policy toward China. It is time for the United States to modernize its export control to reflect the world market reality. In this light, the United States should designate China the same as other recipient countries with which it has Section 123 agreements. (Section 123 of the US Atomic Energy Act of 1954 requires an agreement for cooperation as a prerequisite for nuclear deals between the United States and any other nation.) The United States could grant general authorization for LWR technology export to China that has no implication on proliferation concerns.

The days are gone when the one who provides the ball for the game can dictate the rules, because now there are many other balls available. The United States may still have the best ball around, but before long it will be standing on the sidelines unless it focus on how it can stay engaged today.

P. Lau
President
Fraser Energy Consulting LLC
Oakville, Ontario, Canada

Climate and character

In “Character and Religion in Climate Engineering” (Issues, Fall 2017), Forrest Clingerman, Kevin J. O’Brien, and Thomas P. Ackerman make a strong case that religion, particularly virtue ethics, can substantially contribute to the debate about whether and how to conduct geoengineering research. They argue that focusing on character, particularly virtues such as responsibility, humility, and justice, can help guide such decision-making. I agree that more attention to character evaluation and development is needed in American political life in general and in the assessment of geoengineering in particular. Yet I am concerned that the individualism of American culture will hinder the implementation of their suggestions unless even more attention is paid to the virtues of institutions and communities alongside those of individuals.

Discussions of character in American popular discourse focus almost entirely on virtuous individuals. Whether as part of famous moral exemplars (Martin Luther King Jr., Lois Gibbs) or news stories about whistle blowers, individuals rather than institutions or communities receive the attention. Certainly, individuals are important agents of change, but overemphasizing them can hinder ethical decision-making and action. It is too easy to see the virtuous person as a saint or superhero rather than as someone in whose footsteps we regular people can follow. It is tempting to wait for the as-yet-unidentified virtuous leader to rescue us rather than figure out how we can act more virtuously ourselves. Though Clingerman and his colleagues do not advocate this extreme individualism, I worry that readers may go down that path.

To counter this possibility, we can look to two other features of religions. First, they are practicing communities in which, at their best, they offer moral support and training, both informal and institutional, in the development of their members’ virtues. Individuals are not virtuous on their own. Thus, while an individual may provide insight about geoengineering, we should not wait for such a moral exemplar, but should foster the virtues together so that we might support each other in this difficult decision-making.

Religious communities also develop a collective character in which they work to uphold virtues as a group. Similarly, decision-making bodies about geoengineering should develop both policies and a culture that uphold virtuous decision-making and action. Focusing on those virtues discussed by Clingerman and colleagues, we can ask whether organizations focused on research, policy-making, or implementation of geoengineering are willing to take responsibility for the potential and actual effects of their actions. Do they create a culture in which everyone can rise to the occasion, creatively identifying and solving problems, or do they encourage buck-passing, micromanaging, or dictatorial style decision-making that erodes responsibility? Are they humble about the limits of their knowledge and power as an organization? Do they strive for justice within and outside of their organization? Developing methods of cultivating organizational virtues is outside the scope of this short response. Looking to the theory and practice of religions as well as that of businesses, governments, and nonprofits can enhance the virtues necessary for making decisions about geoengineering.

Sarah E. Fredericks
Assistant Professor of Environmental Ethics
University of Chicago Divinity School

From the Hill – Spring 2018

A funny thing happened on the way to the White House 2019 fiscal year budget, which the Trump administration released on February 12, 2018: Congress adopted a sweeping deal to substantially increase the budget’s spending caps. The administration had originally planned a repeat performance of last year’s budget, according to documents posted to the White House Office of Management and Budget website. This would have included a cut of more than 20% for basic science, with massive reductions to the National Institutes of Health (NIH) and the National Science Foundation (NSF), among other agencies. The administration also again planned on targeting environmental science programs, applied technology research within the Department of Energy (DOE), and other such programs.

But with the passage of the Bipartisan Budget Act of 2018 on February 8, Congress demonstrated (not for the first time) an unwillingness to move the budget in the direction preferred by the administration. Congress added $68 billion to the nondefense discretionary cap in FY 2019, whereas the administration had intended to cut nondefense spending by $65 billion, according to its official documents.

This left a gap of $132 billion between what the law called for and what the White House was planning to request. In response to events, the White House made a last-minute decision to add $75 billion back into its budget. From a science and technology perspective, those changes primarily directed spending back into the budgets for a few big basic science agencies: NIH, NSF, and the DOE’s Office of Science, allowing them to stay closer to flat.

As a result, the budgets for those agencies look far different in this year’s request than last year’s request. (See the chart for a side-by-side comparison of the administration’s FY 2019 and FY 2018 proposals.) Note that the NIH increase is accounted for by additional funding in the area of opioids and the (unlikely) consolidation of NIH with the Agency for Healthcare Research and Quality, the National Institute for Occupational Safety and Health, and the National Institute on Disability, Independent Living, and Rehabilitation Research; many individual NIH institutes would see funding reductions under the administration’s plan.

Below are estimates of major features of the administration proposal. It must be noted that the administration has not produced a fully detailed budget, so the figures cited are drawn from a variety of sources.

Department of Defense. Defense science & technology funding would receive a small 2.3% increase over FY 2017, and the Defense Advanced Research Projects Agency budget would get a larger 19.1% boost spread across several areas including electronics, information and computer technology, aerospace programs, and biodefense.

National Science Foundation. The total budget would be just slightly below FY 2017 enacted level, with the introduction of a new “10 Big Ideas” initiative but with a reduction in the number of new research grant awards.

National Aeronautics and Space Administration. The administration focuses the agency’s investments on space exploration and planetary science, but terminates several earth science missions as well as development of the Wide Field Infrared Survey Telescope.

Department of Energy. The budget request keeps the Office of Science at FY 2017 levels with advanced computing the priority and again recommends elimination of the Advanced Research Projects Agency-Energy and reductions in other technology areas.

National Institutes of Health. Most individual NIH institutes would see roughly 2%-3% decreases below FY 2017 levels, though the NIH total would rise with opioids-related funding and the consolidation of multiple other federal agencies within NIH.

Details on many agencies are still emerging and may not be fully available for some time. On the other hand, as the chart illustrates, several research and development (R&D) areas outside the basic research realm would again be targeted for steep cuts, and many of these are copies of proposals from last year’s budget. Examples include:

• Elimination of climate research within the Environmental Protection Agency (EPA)
• Support for only three of the eight Climate Science Centers within the US Geological Survey
• Termination of the Hollings Manufacturing Extension Partnership within the National Institute of Standards and Technology (NIST)
• Termination of the National Oceanic and Atmospheric Administration’s Sea Grant College Program and the National Estuarine Research Reserve system

Congress appears to be on the verge of rebuffing many of these proposals for FY 2018, though final decisions on appropriations are yet to be determined.

Although the FY 2019 budget is ultimately a bit friendlier for many research programs, the White House could actually have gone further. Following the February 8 budget deal, the cap for FY 2019 stands at $597 billion—but the White House is recommending that Congress spend only $540 billion. That means the administration could have chipped in some extra funding to either roll back some of the reductions shown in the chart or increase investment in apparent priorities such as high-performance computing, cybersecurity, agricultural biotechnology, or other fields. Instead, the White House apparently decided that it would rather see that funding go uncommitted and unspent.

Of course, Congress may have no such qualms spending the money it has given itself. But before it can address FY 2019 spending, it must first finalize FY 2018 spending. Those negotiations may provide some clues into what Congress will do regarding FY 2019.

American Innovation and Competitiveness Act hearing

On January 30, the Senate Committee on Commerce, Science and Transportation convened a hearing on implementation of the American Innovation and Competitiveness Act (AICA), passed in the final days of the Obama administration. The AICA was the bipartisan compromise bill that emerged from dueling attempts at reauthorizing the landmark America COMPETES Act, which set ambitious budget-doubling goals for NSF and NIST, among other things; by contrast, the AICA authorizes 4% funding increases at both agencies. This funding level was scrutinized by the committee, which noted that China is now the world’s second-largest R&D funder, according to newly released data in NSF’s Science & Engineering Indicators report.

Noting tightening budgets and intensifying global competition, the committee discussed measures that NSF and NIST are taking to meet performance targets outlined in the AICA legislation. NSF Director France Córdova updated the committee on several foundation priorities, including increasing transparency and accountability of the merit-review process, strengthening management of multiuser facilities, and broadening STEM education and the I-Corps program. She also highlighted a foundation policy being renewed aimed at streamlining and reducing regulatory burdens for researchers. NIST Director Walter Copan discussed continuing efforts in cybersecurity research set forth in the institute’s Cybersecurity Framework and plans to increase laboratory work in the areas of bioeconomy, artificial intelligence, and the internet of things. He also expressed satisfaction with the AICA-mandated cost-share and re-competition guidelines for Manufacturing Extension Partnership Centers. Overall, Copan stressed that NIST could be doing more to facilitate technology transfer, his specialty. In lieu of giving formal testimony to the committee, the White House Office of Science and Technology Policy provided a letter outlining the administration’s response to the AICA.

HHS Secretary Azar backs gun violence research

Health and Human Services Secretary Alex Azar told a subcommittee of the House Energy and Commerce Committee that he is encouraging federally funded research on gun violence, despite an amendment that Congress attached to a spending bill in 1996 that forbade the Centers for Disease Control and Prevention (CDC) from using money to “advocate or promote gun control.” Following passage of that amendment—which the National Rifle Association supported—Congress lowered the CDC’s budget by the exact amount it had been spending on such research, a move that has had a chilling effect on gun violence research for the past two decades. In 2016, more than 100 medical societies signed a letter asking Congress to lift the amendment, and following the Sandy Hook school shooting, President Obama signed an executive order directing NIH to fund research into gun violence, though that program has since ended. Azar told the committee, “We’re in the science business and the evidence-generating business, and so I will have our agency certainly working in this field, as they do across the broad spectrum of disease control and prevention.”

USGS officials resign over confidential data request

Two senior US Geological Survey officials have resigned following a request by Interior Secretary Ryan Zinke that they provide him with data on the National Petroleum Reserve-Alaska. The officials who stepped down said the request violated the survey’s scientific integrity policy that sensitive and potentially commercially valuable information was not to be shared prior to its public release. A spokesperson for the Interior Department said its Solicitor’s Office had determined that the secretary has the right to review data, draft reports, or other information that he deems necessary. The officials who resigned are Murray Hitzman, associate director for energy and minerals, and Larry Meinert, acting deputy associate director for energy and minerals. Meinert said that he was already planning to retire but that this incident spurred him to retire now. He added that he had no reason to believe that either Zinke or any of his deputies planned to use the information for personal gain.

Senate fails to pass immigration bill

The Senate failed to reach consensus on immigration reform, including legislation to set a path to citizenship for individuals covered under the Deferred Action Childhood Arrivals (DACA) program. A bipartisan measure proposed by Sens. Mike Rounds (R-SD) and Angus King (I-ME) that included language to provide legal status to individuals covered under DACA and funding for a border wall failed to reach the requisite 60 votes to pass that chamber without a filibuster. A counterproposal sponsored by Sen. Chuck Grassley (R-IA) that reflected the president’s position would have included DACA protections and funding for a wall, but also included more controversial policies such as the elimination of the diversity lottery visa and chain migration policies. The Grassley proposal also failed to pass the Senate, leaving the future of DACA recipients in limbo. Both the House and Senate claim they will continue to consider immigration legislation, but time is running out. Meanwhile, a second district court moved to block the termination of the DACA program, which was set to expire in early March.

Higher education groups express concern over Prosper Act

More than 35 higher education organizations, including the Association of American Universities, the Association of Public and Land-grant Universities, and the American Council on Education, sent a letter to the leaders of the US House of Representatives expressing concerns about the PROSPER Act. The act is a reauthorization of the Higher Education Act and would make changes to loan repayment, scholarship, and other financial aid programs. The higher education groups warned that the legislation could “undermine the ability of students to afford and attend college.”

EPA’s Pruitt recaps first year

On January 30, EPA Administrator Scott Pruitt testified for more than two hours before the Senate Committee on Environment and Public Works—his first appearance before the committee since his confirmation hearing in January 2017. Committee Republicans praised the dramatic changes in policy at the agency under Pruitt’s leadership, while Democrats zeroed in on ethics issues under Pruitt’s tenure at an agency that in the past year has been accused of discrediting the role of science and intimidating career employees who appear not to agree with the Trump administration’s agenda. During questioning, Pruitt explained his policy for prohibiting EPA-funded scientists from serving on EPA advisory boards and councils as necessary to keep up the “independence of the review with respect to the integrity of that process,” and he evaded questioning from Sen. Sheldon Whitehouse (D-RI) on why three EPA scientists were instructed not to present their climate research at a conference on the Narragansett Bay in the fall. At the end of the hearing, Pruitt stated that his proposed red team/blue team exercise on climate change is still “under consideration” and that public reports that the White House has asked the agency not to move forward with the exercise are “untrue.” The New York Times reported on March 8 that there would be no such public debate.

Federal agencies develop toxicity testing road map

Sixteen federal agencies have partnered to develop a strategic road map for testing the safety of drugs and chemicals, saying the plan aims to “provide more human-relevant toxicology data while reducing the use of animals.” The road map was developed by the Interagency Coordinating Committee on the Validation of Alternative Methods, which focuses on the development of toxicological testing methods that replace, reduce, or refine the use of animals, and it was published by the National Toxicology Program, part of NIH’s National Institute of Environmental Health Sciences.

Trump endorses right-to-try bill

In the State of the Union address, President Trump expressed his support for a so-called right-to-try bill, which passed the Senate in August 2017 but stalled in the House. The bill would allow terminally ill patients access to experimental drugs if they have exhausted all other options. In 2007, the US Court of Appeals for the DC Circuit ruled that there is no constitutional right to experimental drugs that might only worsen a patient’s condition. Since then, several states have passed their own right-to-try laws. In 2009, Congress expanded the Federal Food, Drug, and Cosmetic Act to allow patients access to experimental drugs when there is no clinical trial available. The Food and Drug Administration’s (FDA) expanded access policy guarantees a patient a response to a request within 24 hours—with a 99% approval rate, according to FDA administrator Scott Gottlieb. The policy requires that the drug must have shown “sufficient evidence of safety and effectiveness,” which usually means relying on data from Phase III trials or very compelling evidence from Phase II trials. Given the FDA’s expanded access policy, it may be unnecessary to adopt federal right-to-try legislation.

A Task for Democracy

Over the past decade, a tool has emerged that enables very precise editing of the human genome, bringing with it tremendous value for research. Yet this gene-editing tool, called CRISPR, also generates worries that society may not be able to control its use and that it may deliver significant harm by allowing scientists to reshape the very basis of life. The tool, which enables scientists to inexpensively manipulate targeted bits of genetic code with relative accuracy, could have a broad range of applications. These uses include treating human genetic disorders, breeding new plants and animals, and controlling disease-carrying insects or invasive species. A Crack in Creation tells the story of this revolutionary technology from the perspective of biochemist Jennifer Doudna and one of her former graduate students, Samuel Sternberg, who were central figures in the technology’s development.

This book has much to like: Charismatic authors who are indisputably leaders in CRISPR. A story of the excitement of discovery in science, and related stories of the discovery of social issues relating to science. Lively writing that is wonderfully clear and pulls the reader along, eager to learn what’s next. A Crack in Creation shows why CRISPR is so compelling as a tool for gene editing, with prospects for groundbreaking medical and agricultural advances. And it shows why both the public and scientists should think carefully about CRISPR’s uses and whether there should be limits on the research.

The book is more important and far more nuanced than another tale told by a scientist about his important discoveries and their impacts, James Watson’s Double Helix, the 1968 autobiographical account of his and Francis Crick’s discovery of DNA’s structure. A Crack in Creation will (or should) be widely read and widely discussed, but also widely critiqued, if we accept the authors’ call for public engagement and communication.

First a note about authorship: The book is purportedly written by Doudna and Sternberg. Yet the voice is entirely first-person Doudna, with Sternberg remaining largely invisible—more like a ghostwriter—even though he was there during part of the laboratory work being discussed. The approach nonetheless works because the writing is vibrant and comes across as authentic and personal in a way that will undoubtedly attract a lot more readers than the usual secondhand narratives by science writers. Sternberg deserves congratulations for being willing to fade into the background. And it makes sense to refer to this book as Doudna’s because it is told from her point of view.

The first part of the book is about the science of CRISPR, or what is more fully called the CRISPR-Cas system. (CRISPR stands for Clustered Regularly Interspaced Short Palindromic Repeats, which refers to repeated sequences of genetic code; Cas refers to a CRISPR-associated enzyme that can be used to cut DNA in specific places.) Doudna offers vignettes about her own development as a researcher, moving from doing work on RNA to discovering CRISPR. She explains step-by-step the reasoning and process of discovery in her lab, including key moments along the way. This discussion shows the process of doing science, including the need to draw on expertise from many different contributors, and the way that the principal investigator recruits and benefits from many others. She names these other researchers, and gives insights into steps that proved more elusive than she had initially thought.

This part is so well written and clear that it is a pleasure to read the story of how Doudna followed various twists and turns to triumph and become a leader in the CRISPR work. She probably intended to be humble and to show that science is teamwork. Yet her personal excitement at thinking of herself as the discoverer comes through as well.

Doudna suggests that CRISPR is a new kind of tool with exceptional gene-editing power. The framing is clever, starting with a story about a patient with a genetic disease whose genome healed itself and overcame the forces of mutation. That one amazing case raised hopes that gene editing in the lab could accomplish the same thing in more patients. Indeed, the story is offered to justify the search for a technological fix to assist nature. The authors briefly acknowledge that CRISPR is not entirely new; earlier gene-editing tools included recombinant DNA, zinc fingers, and TALENs, each with its own strengths and limits. Excitement, a sense of discovery, and the feeling of being at the center of something new and important come through clearly in this part of the book. The section ends with Doudna clicking her computer to submit to Science her 2012 paper demonstrating the use of CRISPR for genome editing. She notes that the publication “was, I now realize, the calm before the storm.”

The second part of the book shifts to social impacts and issues. Here Doudna moves from the swirling scientific world of CRISPR’s discovery, where she was interacting with and learning from many other scientists and colleagues. Here she becomes the leader, with the implication that she sees herself as the one who caused the storm. She seems to believe that she is the one who should lead the way through that maelstrom, to provoke and promote public discussion that can inform decisions about when and whether to use CRISPR. This seems less a matter of hubris on her part than a result of her newfound and admirable recognition that scientists need to take responsibility for the social applications of their work.

This part becomes a bit disappointing, though. Whereas earlier in the book Doudna explains how important it was for her to read the literature and learn new scientific information and tools, here she seems to feel that intuition is a valid approach for social issues. She feels excited by the hopes for curing disease, but queasy about the possibilities for human germline editing—that is, editing the human genome so that the changes are heritable. She gives no evidence that she felt the need to immerse herself in the professional literature related to bioethics, policy, or law. Why not? She seems motivated by good intentions to become “engaged” with public issues, but she could learn from many others who have thought deeply about these concerns.

Doudna discusses a variety of ways CRISPR has been or can be used. The language becomes exaggerated: “It amazes me to realize that we are on the cusp of a new era in the history of life on earth—an age in which humans exercise an unprecedented level of control over the genetic composition of the species that co-inhabit our planet. It won’t be long before CRISPR allows us to bend nature to our will in the way that humans have dreamed of since pre-history.” Really? Then follows a litany of ways the tool can be, and in some cases has been, used to improve agriculture, in both plants and animals. She describes uses of CRISPR that seem close to being able to edit out human disease-causing genes.

All this amazing work is good, well-intentioned, and clearly going to improve human lives, Doudna suggests. Yes, some people worry about genetically modified organisms—but those critics are misguided or uninformed. “It’s easy to get caught up in the excitement,” she writes, and she clearly has. These are “some of the most incredible developments in medical science I’ve ever seen.” In sum, “I am extremely excited and enthusiastic about virtually all the phenomenal progress being made with CRISPR—save for the advancements on one front.” The energetic writing makes the reader enthused as well, and this section is fun to read, although surely not everybody agrees about what is good and even about what is possible when it comes to gene editing.

The book then moves into what is effectively its third section, focused on uneasiness regarding CRISPR’s use. Doudna acknowledges that other people have different points of view, but she is confident in her own. She feels queasy and worried, and therefore suggests that we all ought to be concerned about one particular kind of research: editing human germlines in ways that will affect future generations. Doing such research is wrong, she argues, or at least we aren’t ready for it yet. Many other people agree with her, and they often offer clearly articulated reasons. But that is not the case here. The fact that Doudna does not offer any of her own well-developed reasons, or ground the discussion in the considerable professional scholarship that exists, is a missed opportunity.

Instead of providing reasons for being concerned about germline editing, Doudna gives us more stories about her own leadership in promoting public discussion. She surely does not mean to suggest that she single-handedly led the way to meetings of the National Academies or other smaller conferences that explored these issues, but an ungenerous reader might read it that way. “Had I created a monster?” she asks, and then answers with reassurance that she has also created a moral response. Yet here is another lost opportunity to do more than exhort scientists to engage in public discussions of the kind that she is organizing. She could help explain what would count as successful communication and engagement, and how to know whether the effort is succeeding.

Doudna does not have a clear framework, and does not seem to feel it necessary to seek one, for help in guiding effective communication with the public. Are scientists supposed to talk at citizens, and then let the highly flawed process of policy-making prevail, no matter what? Are researchers supposed to try to guide the public to the “right answers,” which presumably are the ones they hold themselves? Or should they somehow listen to public concerns and try to take those into account when deciding what research to do? Does public engagement about science work in both directions? What is the nature of this communication and the “public involvement” that she considers critical?

Doudna points to the work of the historian and bioethicist J. Benjamin Hurlbut, whom she quotes as saying that “Imagining what is right and appropriate for our world—and what threatens its moral foundations—is a task for democracy, not for science.” But Doudna worries that society does not understand science, and therefore it is up to scientists to educate the public. She, as a leader of CRISPR science, feels compelled to become the communicator about it to the public. To do this effectively and reflectively takes more than good intentions, however. And her failure to understand or develop a framework for what should count as thoughtful public deliberation about how, when, where, and under what conditions to carry out discussions at the intersection of science and society feels like yet another lost opportunity.

We can applaud Doudna for her willingness to step out of her lab and engage the public. We might wish that she had been more forthright about her and her coauthor’s financial interests in several companies that may well benefit from expanded public awareness and discussion. But we can be glad that Doudna wants to help shape public discussion and engage in deliberation about CRISPR science and its applications.

This is not, as the book jacket suggests, the first time since the atomic bomb that scientists have been concerned about the social consequences of their work. Recombinant DNA, which Doudna mentions, cloning, and human embryonic stem cell research all provide recent examples. CRISPR is not an atomic bomb, Doudna did not invent the tool by herself, and she will not be the one to develop most of the innovations built on CRISPR technology; nor will she be the only one to lead continued public discussions about its uses. But aside from the hype and some missed opportunities, her book offers a beautifully written and compelling story. It presents the science in a wonderfully clear way and will surely be required reading in the public discussion for which it rightly calls.

Helping Fathers be Parents

Don’t Sweat the Small Stuff

How worried should you really be about whether eating red meat or drinking coffee will cause cancer? Will using your cell phone too much result in brain tumors? Can the use of plastic water bottles containing the chemical bisphenol A, commonly known as BPA, cause male infertility or reproductive disorders?

Faced with the continued proliferation of online news outlets and scientific journals, all under the same pressure to publish new, eye-catching findings, sorting the truly alarming revelations about health and environmental risks from those that may be interesting but don’t pose actual danger is increasingly difficult. This challenge of understanding and prioritizing risks is compounded by the media’s tendency to conflate hazard and risk, particularly when advocating for regulatory change. (Hazard is simply a measure of the potential for something, such as a chemical, to cause harm, while risk is the probability that a hazard will actually result in some adverse effect under a specific use scenario.)

Geoffrey C. Kabat addresses these issues in his new book, Getting Risk Right: Understanding the Science of Elusive Health Risks. This is the latest of several books and magazine articles that he has written in an effort to educate both scientists and the general public about the practice of risk assessment, hoping to alleviate growing anxieties about low-level environmental risks that are likely to have minimal or no health effects but consume large amounts of regulatory agencies’ scarce resources.

The book begins by introducing the concept of critical thinking when looking for patterns and associations in large datasets and how thinking outside the box—in his words, “being open to new ways of seeing to look for answers in places where we might not immediately think to look”—can lead to breakthroughs in understanding causal relationships. Next comes an examination of how correlation is often misinterpreted as causality, resulting in misleading and erroneous conclusions. Kabat asserts that such behavior results from scientists interpreting data within what he calls a “biased worldview,” coupled with the need to find alarming associations between environmental hazards and adverse health outcomes to ensure publication of results and a steady funding stream.

Kabat provides multiple examples of how a biased worldview can constrain scientific progress. These include the belief that environmental chemicals are causing global declines in fertility, even after the initiating study was repudiated; believing in an endogenous or dietary cause for all cancers, which limited the search for viral etiologies that proved to be the case in cervical cancers; and continuing to think that electromagnetic fields cause adverse health effects, despite two decades of research showing such effects to be nonexistent. To avoid these biases, Kabat warns strongly against the dangers of using the “court of public opinion” to gain support for a particular hypothesis rather than using the totality of the evidence to subject a favored hypothesis to rigorous testing, a theme found throughout the book.

Though Kabat provides a good discussion of the pitfalls of observational epidemiological studies, how to differentiate a good study design from a faulty one, and key concepts of statistical power and significance, sample size, and preconceived biases, the writing is uneven. At times, he seems to assume the reader is totally new to the subject, while at other times he assumes at least a basic level of scientific understanding. For example, he carefully explains the difference between observational and case-control epidemiological studies, but spends very little time describing the process required for quantitative risk analysis, which is central to his arguments about wrongful assumptions of causality. And sometimes Kabat discusses the same topics in multiple locations within a section, causing the reader to circle back in an attempt to put the pieces together rather than following a linear argumentative approach.

Kabat explores how societal pressures influence research in the field of biomedicine. He highlights the power of the media to amplify scientific findings, resulting in regulatory action furthering political or advocacy agendas—often before a true causal association between exposure and health is determined. For example, a high level of public concern over the use of BPA in thermal printing paper or food containers resulted in its ban in certain consumer products by various states. The ban was implemented despite the US Food and Drug Administration’s conclusion, based on extensive research, that BPA is safe at the current levels occurring in food. Kabat asserts that this kind of premature regulatory action due to fears of endocrine disruption or of cancer is becoming increasingly common. This in turn feeds back into the scientific community and influences funding streams and research pressures in certain fields.

Interspersed throughout these topics is a continuing discourse on why positive associations are published much more frequently than negative (or nonexistent) associations, why even small risks can become magnified when taken out of context, and how certain types of risks raise greater fears than others. Kabat touches on the role of science in the sphere of public policy, arguing that scientists must be aware of how their advocacy of a particular policy approach can pose a serious threat to their ability to conduct truly independent and unbiased research. He cautions that the peer-review system is susceptible to public pressures and biases, and that researchers need to be aware of their tendency to want to find any plausible explanation for observed phenomena. He admonishes scientists to focus on the next experiment or observation needed to critically examine a causal hypothesis, rather than conducting a study with a high level of public appeal.

The meat of the book comprises four in-depth case studies, each examining an instance where research has been applied to identify factors affecting health and disease. The first two studies explore examples where researchers had hypothesized significant health effects resulting from miniscule environmental exposures, yet little progress had been made in studying them in spite of abundant public attention and large amounts of funding. The next two studies turn to examples where scientifically rigorous approaches determined the causes of seemingly intractable health problems. But even as all four chapters are well written and thoroughly documented, they lack sufficient introductory or conclusory material to inform the reader of how they illustrate the primary thesis of the book: that scientists should avoid the biases and pitfalls that arise from becoming vocal public advocates of a hypothesized risk before it has been subjected to rigorous testing or an in-depth analysis of prior work.

The first case study provides a thorough review of the epidemiological research conducted on the safety of mobile telephones and their hypothesized connection to brain cancers. These studies, mostly conducted in the 1990s and early 2000s and summarized in a multiauthored review paper in 2010, stem largely from the very public concerns raised by one vocal man whose wife talked incessantly on her cell phone while holding it primarily against one ear, and ultimately died of a brain tumor on the same side of her head. Kabat documents the numerous shortcomings of research on this topic, including significant recall bias between brain tumor cases and controls, selective and slanted presentation of results, and failure to report on weaknesses and limitations of the studies. He notes that data correlating the (small) number of brain tumors with the (huge) number of people using cell phones, along with animal studies that showed the lack of health effects associated with cell phones, were underrepresented in the public and scientific debate of this issue. Kabat summarizes his use of this example by stating that it “has provided ample occasions for the play of bias in many forms and at many different levels in the interpretation of the results of scientific studies and the translation of these results to the public.”

The next case study presents the lack of scientific consensus on endocrine disruption, a research topic on which large amounts of research dollars have been spent. Kabat describes the research and subsequent significant regulatory pressure brought to bear on identifying chemicals that cause—or have the potential to cause—measurable changes in the human endocrine system that result in disease, reproductive impairment, or both. This work was based on an observational paper published by Danish researchers in 1995. The paper purported to find decreased quality of semen in men during the prior 50 years, and expressed concern that this might be due to “estrogens or compounds with estrogen-like activity or other environmental factors.” Kabat meticulously describes the ensuing research that eventually discredited the initial report and revealed regional differences in male fertility over time without being able to document a widespread causal factor. He illustrates the role that public pressure played, following the publication in 1996 of the sensationalist book Our Stolen Future: Are We Threatening Our Fertility, Intelligence, and Survival? in pushing federal governments and regulatory agencies to identify “endocrine disrupting chemicals” as a special regulatory class. Kabat points out that how a science-based question is posed can significantly affect the type of research and public discourse conducted. Asking the question now enshrined in regulatory language—“What chemicals cause deleterious endocrine-based health effects?”—leads to the study of a multiplicity of weakly active compounds present in the environment. Asking, instead, “What exogenous factors might impact reproductive development via hormonal perturbations?” would focus research more usefully on the potential reproductive effects of potent pharmaceuticals.

The next two case studies describe how researchers build on each other’s successes to understand mechanisms of disease, which can lead to major breakthroughs in diagnosis and prevention. This is interesting, but it is not clear how they illustrate the principles of risk analysis, public awareness, and potential bias that are the focus of the book. In the chapter “Deadly Remedy: A Mysterious Disease, a Medicinal Herb, and the Recognition of a Worldwide Public Health Threat,” Kabat documents the detailed epidemiological work that identified the cause of terminal renal failure in Belgian women during the 1990s and in rural families in the Balkan countries from the 1950s to the 1990s. Both episodes resulted from exposure to aristolochic acid, a powerful kidney toxin and carcinogen found in various Aristolochia plant species. Aristolochia fangchi was mistakenly substituted for the herb Stephania tetrandra and used as a supplement in Belgian weight-loss clinics. Aristolochia clematitis grows wild within wheat fields in the Balkans in a just-sufficient amount to be considered not worthy of removal but toxic if ingested repeatedly over 10 years or more.

The book’s final case study, of viruses and cancer, illustrates how researchers followed their intuition and built on each other’s work to describe the epidemiology of various cancers, identify the causal viral agents, and develop diagnostic and preventative approaches. This is the story of how, in the late 1950s, the surgeon Denis P. Burkitt established the distribution of facial (Burkitt’s) lymphoma in equatorial Africa, from which he concluded that an infectious agent was implicated in its transmission. The pathologist Anthony Epstein at the Bland Sutton Institute in London subsequently isolated the inciting virus, now known as the Epstein-Barr virus. Others, most notably the virologist Harald zur Hausen, followed in the 1970s with similar research to identify the specific human papilloma viruses that cause cervical cancer. These findings radically changed the medical view of cancers, broadening the potential causal agents to include a wide range of viruses and resulting in better diagnostics and life-saving vaccines.

Kabat concludes his book with a short chapter reiterating points made earlier and admonishing readers to realize that important advances in the biomedical field occur through the persistence and collaboration of researchers building objectively on previous work. He argues that such advances are rarely achieved by “following some fashionable but ill-defined idea” based on “over-stated claims, implausible findings, and appeals to irrational fear.” This book presents important topics for consideration and four fascinating and well-documented epidemiological case studies. It is well suited for use in an introductory epidemiology class, where sections or chapters could be assigned as introductory reading followed by in-depth discussion, additional readings on particular topics, or both. And general readers are likely to find the four detailed case studies to be an interesting read.

Reconceptualizing Infrastructure in the Anthropocene

A fundamental shift is afoot in the relationship between human and natural systems. It requires a new understanding of what we mean by infrastructure, and thus dramatic changes in the ways we educate the people who will build and manage that infrastructure. Similar shifts have occurred in the past, as when humanity transitioned from building based on empirical methods developed from trial-and-error experience, which was sufficient to construct the pyramids and European cathedrals, to the science-based formal engineering design methods and processes necessary for the electric grid and jet aircraft. But just as the empirical methods were inadequate to meet the challenges of nineteenth- and twentieth-century developments, today’s methods are inadequate to meet the needs of the twenty-first century. Now that we have entered the Anthropocene period in which human activities affect natural systems such as climate, engineers face far more complex design challenges.

The durability of the pyramids and cathedrals is evidence that the inherited wisdom of experience can be adequate for many tasks, but we can also see the disadvantages of working without the formal design tools based on advanced mathematics and engineering science. The choir vault and several buttresses of the Beauvais Cathedral in France collapsed in 1284, only 12 years after its partial completion. Retrospective analysis suggests that the problem was resonance under wind load, the type of stress that could be anticipated only with the formal design methods that were yet to be developed. And when the first iron bridge was built at Coalbrookdale, England, in 1781, the engineers used the same design they would have used for a masonry bridge. Within a hundred years, however, structures such as James Eads’s 1867 bridge over the Mississippi River at St. Louis and Gustave Eiffel’s 1884 Garabit Viaduct were being formally designed using scientific methods and quantitative design processes that made it possible to predict the performance of the new materials. Personal experience and historical heuristics were replaced by scientific data and quantitative models. Opportunities for infrastructure evolution that would have been impossible without scientific understanding suddenly became available, leading to more efficient, effective, and safer infrastructure, and eventually to entirely new forms of infrastructure such as air travel and electrification.

The engineering tools, methods, and practices necessary for a practicing professional; the engineering education necessary to prepare such a professional; and the way engineering is performed institutionally have all changed fundamentally throughout history as the context within which humans live and prosper has changed. We are now in a period of rapid, unpredictable, and fundamental change that is literally planetary in scope. To date, neither our ideas regarding infrastructure, nor our educational and institutional systems for conducting engineering, have reflected this striking evolution in context. We are well past the point where adaptation is pragmatically, and ethically, required.

It is important to remember that this doesn’t mean that all current practice and educational methods are suddenly obsolete. Rather, it means that they are increasingly inadequate, especially when applied to projects that involve complex social, economic, and cultural domains. Designing a water treatment plant will remain well within the capability of today’s field of environmental engineering; designing the nitrogen and phosphorus cycles and flows in the Mississippi River basin to reduce the life-depleted “dead zone” in the Gulf of Mexico cannot be done so simply.

The planet as infrastructure

That we now live on a terraformed planet is not a new idea. The term “Anthropocene” was popularized in an article in 2000 by Paul Crutzen and Eugene Stoermer, but as early as 1873 the Italian priest, geologist, and paleontologist Antonio Stoppani used the term “anthropozoic era,” and others have used similar language throughout the twentieth century. The force driving the birth of a new era is obvious. A planet that supported roughly 450 million humans in 1500 now supports over 7.5 billion, and because of industrial progress and economic growth, each of those humans has a vastly more consequential impact.

In a blink of geologic time, then, humans left a planet where they were but one species among many and built a world increasingly shaped by the activities of, and for the purposes of, a single species—themselves. It is not that the planet has been deliberately designed by humans, but many human-built systems from transportation to communications have resulted in global changes that were not consciously designed by anyone. Moreover, that human activity is clearly a significant contributor to current dynamics of such natural phenomenon as climate change, biodiversity, nitrogen and phosphorous cycles, microbial evolution, and regional ecosystem parks such as the Everglades does not imply deliberate design in such cases either. Rather, the world we find ourselves in is one where deliberate human activity, itself a complex network of functions, interacts with a highly complex planetary substrate to create unpredicted, often challenging, emergent behaviors, of which climate change is but one example.

These emergent properties would not exist but for humans; the design process involved, however, is not the explicitly rational and quantitative method that we are used to. Rather, a planet characterized by rapidly increasing integration of human culture, built environments, and natural systems producing novel, highly complex, rapidly changing emergent behaviors challenges all our existing ideas about design, operation, and management of infrastructure. Indeed, it challenges the very idea of infrastructure as limited to local, highly engineered systems. For in such a world, “natural” cycles and systems transmute from exogenous conditions into infrastructure components, a process implicitly recognized by, for example, the substantial literature on ecosystem services and earth systems engineering and management.

Consider urban water supply. Replacing a water pipe in New York City, albeit expensive and complex, is well within current professional and institutional capabilities. On the other hand, to maintain that water supply means designing, through engineering and land use regulation, a continuously monitored network of 19 reservoirs in a roughly 2,000-square-mile watershed. Similarly, most people know that Arizona’s population is concentrated in several large urban areas in semiarid desert regions of the state. Contrary to popular belief, however, Arizona’s water supply is robust. It is very diverse, with 17% from in-state rivers and associated reservoirs, 40% from groundwater, 3% reclaimed, and 40% imported from the Colorado River, which channels water from seven states. Flexibility and resilience in operations is enhanced because imported water is largely banked in underground reservoirs, while wastewater is highly managed and either returned to the ground where it can be accessed later or sent to power generation facilities. Similarly, water in California is tightly designed. Mountain snow releases water as it melts, and that runoff is heavily managed for multiple uses as it moves toward the ocean. As in many other regions, natural flows from precipitation are now highly managed and there is increasing recognition that historical flow patterns are becoming less and less relevant for predicting future flows. In short, watersheds are now so highly managed that the environment has in effect become the water infrastructure.

As these examples suggest, the advent of the Anthropocene requires the development of new institutions and frameworks that are beyond today’s traditional disciplinary structures and reductionist approaches. Continued stability of the information-dense, highly integrated, rapidly evolving human, natural, and built systems that characterize the anthropogenic Earth requires development of new abilities to rationally analyze, design, engineer and construct, maintain and manage, and reconstruct and evolve such systems at local, regional, and even planetary scales. It is not that current institutions and disciplinary boundaries are completely obsolete; rather, it is that increasing complexity and rates of system evolution require a new and more integrative level of sophistication in infrastructure conceptualization, design, and management. This can draw on a number of relatively new fields of study, including sustainable engineering, industrial ecology and associated methodologies such as life-cycle assessment and materials flow analysis, systems engineering, adaptive management, and parts of the urban planning and sociology of technology literatures.

The boundaries reflected in current engineering institutions, disciplines, and educational practices and structures are still appropriate for many problems. But today, at the dawn of the Anthropocene, they increasingly fail at the level of the complex, integrated systems and behaviors that characterize the anthropogenic Earth. No disciplinary field in either the physical or social sciences addresses these emergent behaviors, and very few even provide an adequate intellectual basis for perceiving, much less parsing, such complex adaptive systems. Engineering cannot continue to rest on traditional curricula and strengths, which are increasingly inadequate given today’s social, economic, environmental, and technological demands. A road built into a rain forest to support petroleum extraction operations cannot be well designed unless it is understood that such infrastructure becomes a corridor of development and environmental degradation. Similarly, a new airport planned in Peru in the Inca Sacred Valley cannot be adequately designed unless its role in dramatically increasing tourism and development in a fragile ecosystem and historic cultural region is understood, and the numerous effects evaluated. Planning for urban transportation infrastructure is inadequate unless it includes consideration of the implications of such varied developments as hybrid, electric, and autonomous vehicle technologies, potential climate change effects on mobility, aging populations, and terrorism reduction and mitigation. We cannot stop building infrastructure, but especially as many forms of the built environment outlast the context within which they were built—think of coal-fired power plants being built today that will be operating for decades—we can try to design infrastructure to be more resilient to change even when we cannot predict the particulars of such change, to be more adaptable and agile as those changes occur, and to be more aware of obvious effects of our designs that may have heretofore been ignored.

In a world where natural systems are for many purposes becoming simply another element of regional and global infrastructure systems, society in general and engineers in particular are challenged to rapidly develop tools, methods, and intellectual frameworks that enable reasoned responses. The reductionism inherent in pretending the challenge is just climate change or new transportation systems or lower-carbon energy production and distribution systems or water quality and quantity, although appropriate in certain cases, can no longer be defended as an acceptable disciplinary or institutional approach to infrastructure design, operation, and management. Rather, engineers, engineering educators, and institutions that support and manage infrastructure need to create solutions in the real world, and to embrace the implications of the Anthropocene. Bluntly, the world is a design space, and absent catastrophic collapse, there is no going back.

New engineering for a novel planet

The reality of the terraformed Earth, whatever one wants to call it, requires some significant changes in the way technology and infrastructure are designed, implemented, understood, and managed. Although the obvious target disciplines are in engineering, such a focus would be inadequate. Increasingly, private firms develop and manage emerging technologies that shape the integrated human/built/natural infrastructures that are the defining characteristics of the Anthropocene, so business education writ broadly must be part of the shift. Moreover, as systems that were previously considered external to human design, such as biology and material cycles, increasingly become design products, many aspects of science segue into engineering fields, requiring different mental models and tools. This implies that many fields of science also require some grounding in technology, design, and engineering. Against this background, and recognizing that any suggestions for reform at today’s early stage will necessarily be partial and somewhat arbitrary, we can nevertheless make some recommendations for change. For simplicity, we will group these suggestions into three categories: short term, medium term, and long term.

Short term. In all relevant disciplines, including at least all engineering and business programs, at least three areas of learning should be mandatory for all students, undergraduate and graduate, and professional accreditation institutions such as the Accreditation Board for Engineering and Technology (ABET), which approves engineering programs, should set an explicit timetable for achieving this. Whether these areas of expertise are introduced as modules in existing curricula or as independent courses is not as important as ensuring that some level of substantive knowledge is acquired by all students. In order to reinforce this, professional exams, such as the Fundamentals of Engineering, and Principles and Practice of Engineering exams given to professional engineers, should be expanded to include these subject areas. These broad areas of competence are:

1. Technological, social, and sustainable systems theory and principles, including concepts such as wicked complexity and satisficing versus optimizing system performance. This category of knowledge includes understanding the difference between simple and complex system behavior, and designing and managing integrated human/built/natural systems.
2. Big data/analytics/artificial intelligence (AI) functions and systems. Any business or engineering student who graduates without knowing something about these fields is verging on incompetent.
3. Cybersecurity and cyberwar operations from a defensive and offensive perspective. Today, business managers and engineers are trained to rely on, and design and deploy, “smart” systems at all scales; indeed, in many cases this is a performance and business imperative. On the other hand, virtually none of these professionals are taught anything about cybersecurity or cyberwar, a gaping competency chasm when both Russian and Chinese strategic military doctrines have shifted toward deployment of unrestricted warfare/hybrid warfare/cyber warfare initiatives. Inviting an adversary to subvert all your critical infrastructure and systems, which the US educational system now does, is simply stupid, especially since these adversaries are very engaged, and quite competent, in such activities.

These changes in curricula are a one-time improvement and by themselves will quickly become inadequate. At least as regards engineering, two further steps should immediately be taken. It is, for example, sadly the case today that too many engineering students are still being graduated with skill sets that prime them for replacement by expert systems. Thus, the first step should be to review the entire engineering curriculum for each discipline, identifying those elements of existing courses that software, AI systems, and changes in technology have made obsolete. A thorough analysis of the content of existing courses, coupled to modernization of ABET criteria, will help create educational programs that prepare students for the future, not the 1950s.

This should be combined with implementation of a change that engineering schools and those who hire engineers have been talking about for many years but have yet to implement: making engineering a graduate-level professional degree. Uniquely among professions, engineering remains, at least in theory, an undergraduate professional degree: neither law nor medicine nor any other major profession does so. This anachronism today works only by impoverishing the scope and scale of the education provided to undergraduate engineering students, who are then thrust into a world that demands ever more operational sophistication from them—something they cannot provide because there’s no room for anything but engineering-related courses in their undergraduate programs.

Medium term. The Anthropocene is characterized by unpredictable change with far shorter time cycles than those common to most infrastructure systems. An important emphasis, therefore, must be on developing frameworks, tools, and methods that enable more agile, adaptable, and resilient infrastructure design and operation even in the face of fundamental uncertainty about virtually every important challenge infrastructure systems will face.

Although specifics are difficult and remain to be worked out, there are a number of ideas and models across existing engineering systems that can be generalized toward the goal of agile, adaptive, and resilient infrastructure. For example, the entire domain of consumer computational technology is characterized by many different technologies that must converge in working systems that are robust and simple enough in operation for average consumers to understand. One of the basic mechanisms for this process is “roadmapping,” where modularity of technology—printers are different from data storage devices are different from input devices—is combined with robust module interfaces to enable both module and overall product evolution that is at once unpredictable, functional, and economic. Another example from consumer electronics is the substitution of software for hardware functionality whenever possible. Assuming appropriate hardware design, software can easily be upgraded, whereas hardware is far more difficult to change out (imagine if every improvement in software security or function required a new motherboard!). And programming offers a third model: the core of C++, for example, has remained fairly stable for years, but software evolution based on that core has been explosive, unpredictable, and very rapid. This suggests a “hub-and-spoke” framework: those elements of an infrastructure system that are not likely to change rapidly can be built with longer time frames and be more robust, whereas the spokes and edge elements, connected with stable interfaces, can evolve unpredictably and much more rapidly. Other characteristics might include planned obsolescence, resilience thinking that includes safe-to-fail and extensible design that can be easily updated, and increased compatibility and connectivity of hardware. These and similar models should be applied to infrastructure design generally.

Because of their importance to society, infrastructure systems tend to be designed to be robust and long-lasting. But especially as “natural” systems increasingly become a part of large infrastructure design and planning, it may be necessary to identify infrastructure where context may be changing rapidly, and “down-design” it: that is, design it to be less robust, more temporary, and capable of being replaced inexpensively and much more quickly. This could be combined with a major hubs/minor hubs/links design, where robust major hubs last for long times, minor hubs are down-designed, and structural links are robust in the face of unpredictable challenge.

Long term. The longer-term situation is more inchoate and more complex. Three fundamental shifts, one conceptual, one systemic, and one educational, are required, but none of them can be planned with any specificity. Accordingly, an overriding requirement is to learn how to implement continuous dialog and learning with the many systems within which we are embedded so that we can adjust our mental models, institutions, and educational practices in response to real time changes in system state. Policy and practice become processes of adaptive evolution rather than static responses to a stable world, and technocrats, managers, and engineers must learn to focus on ethical and responsible processes, and less on content, which will be continually and unpredictably changing.

The conceptual shift sounds simple. The fundamental implication of the Anthropocene is that everything from the planetary to the human body itself become design spaces: the world has become infrastructure. Especially given the complexity of the systems involved, this doesn’t mean design in the sense of a completely understood controlled system such as a toaster. The world of simple, optimizable design is gone. Rather, it means that in a world of radical human life extension, integrated AI/human cognition, ecosystems as infrastructure, and re-wilding using previously extinct species in accord with wilderness chic design aesthetics, everything is subject to human intervention and intentionality. The Everglades is a design choice and exists today in its current state because we choose to have it exist in that state, even if it obviously has biological and ecological complexity beyond our current understanding. The simple truth of a terraformed planet is difficult for many people to accept, for reasons ranging from fear of the responsibility to ideological commitment to archaic visions of the sanctity of nature, but it is a necessary basis for technocratic and policy professionals if they wish to act ethically and responsibly under today’s extremely challenging conditions.

Systemically, institutions need to evolve to mirror the complexity and interdisciplinarity of the networks with which they are interacting. Importantly, this means that institutions must become multi-ideological if they want to manage regional and planetary infrastructure. If they do not, they will be ineffectual. Thus, for example, the United Nations Framework Convention on Climate Change, signed in 1992, prioritized environmental ideology over other values, with the result that implementation has been sporadic, contentious, and ultimately unsuccessful. The shift involved here is also conceptual, and deceptively simple: institutions driven by activist stakeholder and dominant ideological demands must instead begin to frame themselves as problem-solvers, with the desired goal being not the advance of a particular agenda, but the creation of a politically and culturally stable solution to a particular challenge. Moreover, the institutions that typically govern infrastructure are mechanistic, characterized by hierarchical structure, centralized authority, large numbers of difficult to change rules and procedures, precise division of labor, narrow spans on control, and formal means of coordination. In contrast, industrial organizations in sectors exposed to rapid and unpredictable change, such as Silicon Valley technology firms, are far more organic, with the emphasis on moving fast rather than the lumbering pace imposed by formal structure. Whether current infrastructure institutions are capable of transitioning their management structures is an open question, given their legalistic and bureaucratic incentive structures; if they cannot, they must be replaced.

The educational shift is not unlike the institutional shift. Engineering education in particular combines quantitatively complicated but structured models and learning with creativity, design, operation, and problem-solving capability. The former is rapidly becoming mechanized, and the latter is increasingly done in a rapidly changing social, economic, and technological environment. Because of this, it will not be enough to simply reform engineering education in the short term: rather, engineering education must become a constantly evolving process and curriculum that changes at the same pace as its context. Engineers, who today often resemble expert systems—highly competent in one specific area such as chemical or civil engineering, but with knowledge that rapidly scales off at the edge of that domain because they don’t have time to gain much education outside their particular discipline—must instead become constantly learning quantitative problem solvers, able to integrate across many different domains, as well as the still yawning gap between the science and technology fields and the social sciences and humanities.

These changes sound challenging, and indeed they are. But the Anthropocene is not a choice that we can reject; it is a reality that we have often unintentionally been creating since the Neolithic. We are now at the point where it must be embraced in all its complexity if we intend to respond rationally, ethically, and responsibly to its challenges.

FDA Overreach on Genetically Engineered Animals

The oft-quoted quip “To a man with a hammer, everything looks like a nail” anticipated the practice of federal agencies expanding their mandate by shoehorning policy initiatives into regulatory regimes for which they were never intended. This practice seems to run contrary to Article I of the US Constitution, which vests all legislative power in the Congress, yet it can happen because Congress has delegated broad rule-making authority to executive branch agencies, which in turn interpret their rules to their own advantage.

Congress has established some controls over runaway agency rule-making, notably in the 1946 Administrative Procedure Act, which created a public notice-and-comment process for the orderly and public promulgation of regulations and provided for judicial review of regulations. Increasingly, however, agencies have ducked the formal requirements by issuing flexible, nonbinding “guidance documents” that interpret existing rules in a manner that can greatly distort or expand the scope of regulations to entirely new purposes. That’s how the Food and Drug Administration (FDA) has managed to craft inventive but feckless regulation of animals modified using molecular genetic engineering techniques.

Although a typical FDA guidance document asserts that it “does not create or confer any rights for or on any person and does not bind FDA or the public,” anyone ignoring its provisions would likely face an agency compliance action or be denied approval of a product or activity for which approval was being sought. Moreover, because the power of a gatekeeper agency such as the FDA—whose approval is required for a product to be marketed—is so great, manufacturers will not commonly challenge the FDA’s approach, requirements, or approval conditions because to do so could risk future delays, disapprovals, and difficulties with other products regulated by the agency. Thus, for practical purposes, guidance is equivalent to a rule—except that it’s not legally binding on regulators.

In an FDA guidance document titled “Regulation of Genetically Engineered Animals Containing Heritable rDNA Constructs,” issued in 2008, the agency announced that “recombinant DNA” when introduced into the DNA of an animal is a “new animal drug” and thereby requires the animal to be reviewed as a drug under the Federal Food, Drug, and Cosmetic (FD&C) Act. Accordingly, in order to be sold, the animal must undergo government review and approval, the same as a veterinary drug such as an antibiotic or pain reliever. The guidance focuses inappropriately on the use of a single precise breeding technique among the spectrum of techniques used by animal breeders, but without identifying a demonstrable, rather than speculative, risk as the basis for imposing the high evidential standard of drug “safety and effectiveness” required by the FDA’s new animal drug regulations.

There is no hint anywhere in the FD&C Act, the FDA’s primary enabling statute, that animals could be, in effect, regulated as a drug. Nor was such an interpretation necessary for the safe sale and consumption of genetically engineered animals. A more apposite model is the way that another FDA component, the Center for Food Safety and Nutrition, regulates other foods. The law places the burden of ensuring the safety of foods and food ingredients on those who produce them. It prohibits the adulteration (contamination) or misbranding (mislabeling) of food, but does not require the inspection or evaluation of all food before its sale in shops, supermarkets, or restaurants. Rather, the FDA’s oversight, encompassing all food except meat, poultry, and egg products, which are regulated by the US Department of Agriculture (USDA), relies on market surveillance or post-marketing regulation, and the FDA takes action only if there is an apparent problem.

The law does require a premarketing review for certain food-related products. These include most food additives—a class of ingredients that includes preservatives, emulsifiers, spices and sweeteners, and natural and synthetic flavors or colors, among others. In general, a food additive must be approved if it becomes a component of or otherwise affects the characteristics of a food and it is “not generally recognized as safe (GRAS) by qualified experts for its intended use.”

GRAS is an important concept, especially in the context of genetically engineered animals. Before a new food additive is marketed, it is the responsibility of the producer to determine whether or not the substance is GRAS. The agency routinely reviews food additive applications for safety only when the substance in question has been determined not to be GRAS by the producer. If the producer determines that a substance is GRAS, then the FDA requires only that it receive a notification, which is then subject to agency review.

Another important aspect of the GRAS concept is that multiple GRAS substances that have been combined are still considered GRAS. Similarly, because adding a GRAS gene to a GRAS organism is likely to yield a GRAS outcome, a lengthy FDA premarketing review should not be necessary for genetic constructions.

It should come as no surprise that the resulting practically unworkable FDA approach has been a disaster for research and development (R&D) advances in the entire, once-promising sector. Two examples will illustrate the FDA’s dysfunction, which is the reason there is meager R&D on genetically engineered animals in the United States.

The first is a genetically engineered Atlantic salmon that reaches maturity 40% faster than its unmodified cohorts. The genetic changes confer no detectable difference in the fish’s appearance, ultimate size, taste, or nutritional value; it just grows faster and consumes less food over its lifetime. Also, because the fish are all sterile females and farmed inland in closed systems, there is negligible possibility of any sort of genetic contamination of the wild fish gene pool or other environmental effects.

More than a decade before the FDA issued its guidance in 2008, its officials had told the developer to submit a marketing approval application to the agency, without identifying a clear regulatory rationale or pathway. The FDA held up the application for approval of the salmon for almost 13 years before even reaching a decision on how this fish should be reviewed. Review of the salmon as a “new animal drug” required several more years. At the end of a two-decades-long regulatory process, the FDA concluded what should have been obvious long before: that no health or environmental risks or food quality concerns existed. The salmon is not yet available in the United States because of congressional delays over labeling. However, in Canada, the faster-growing salmon was approved without difficulty, is available without special labeling in supermarkets, and is selling well.

A delay in the availability of cheaper salmon isn’t the end of the world, of course, but the FDA also unnecessarily and inexplicably delayed small-scale field trials of an innovative method to reduce the population of Aedes aegypti mosquitoes that transmit Zika virus, yellow fever, dengue fever, and chikungunya. The method uses a genetically engineered male mosquito constructed with a genetic defect that causes it to require a certain growth supplement for survival. When released in the absence of the supplement, the mosquitoes survive only long enough to mate with wild females and pass the lethal gene to their progeny, which soon die. Because male mosquitoes don’t bite, they present no health risk, and because the progeny die before they can reproduce, none should persist in the environment. This approach has been successfully tested in several countries.

The FDA took an unconscionable five years (2011-16) to approve a single small-scale field test of this mosquito, and that came only after mounting pressure from the growing Zika threat and the consequent need to control A. aegypti. In August 2016, the agency finally approved a field trial at one site in the Florida Keys, some 160 miles from the Zika outbreak in Miami, but that trial has yet to begin.

The use of the new animal drug regulatory pathway for the mosquito presented an insoluble legal conundrum for the FDA. In order to approve it for marketing as a drug, regulators would have to conclude that the genetic material that causes a male mosquito to self-destruct after producing defective, doomed offspring is safe and effective for the mosquito, the requirements for approval specified in the FD&C Act. The FDA would have found itself tied up in legal knots if its ultimate approval of the insect were challenged in court by environmentalists and anti-genetic-engineering activists, as would have been inevitable. After we first pointed out the “safe and effective” impossibility in the Wall Street Journal in 2016, the FDA in January 2017 ceded the regulation of mosquitoes to the Environmental Protection Agency (EPA), an agency that does have the statutory authority to regulate insecticides.

The FDA had previously confronted the question of how to regulate the new and increasing number of lines of genetically engineered laboratory animals created for medical research. The agency’s solution was simply to exempt them from the excruciating approval process by magisterially invoking “enforcement discretion,” meaning the agency would not enforce requirements under the FD&C Act. The FDA also invoked enforcement discretion to obviate the need for review and approval of GloFish, genetically engineered varieties of aquarium fish. Regulators’ rationale was that the fish posed no threat to the food supply, and that there was no evidence that it posed any more threat to the environment or to public health than their unmodified counterparts. Thus, as viewed through the FDA’s distorted lens, a small-scale field trial of a suicidal mosquito with only non-reproducing offspring that would reduce the mosquito population for a clear public health benefit poses a greater risk than the unregulated presence of unlimited numbers of reproducing aquarium fish and many lines of genetically engineered laboratory animals.

The illogic of the FDA’s use of the new animal drug regulatory pathway demonstrates how the agency’s contortions enabled it to arrogate regulatory jurisdiction over animals modified with recombinant DNA techniques. Far from having seen the error of its ways, in January 2017 the FDA doubled down and sought to further expand its regulatory turf. Anticipating that the incoming Trump administration would not agree with its intention to regulate all emerging animal breeding technology, in the last hours of the Obama administration the FDA rushed to publish proposed guidance that would encompass all molecular genetic modification techniques (including gene-editing techniques such as CRISPR-Cas9) that lay outside the definition in the 2008 guidance. Thereby, the FDA proposed to adopt a scientifically unwarranted precautionary policy that would require the agency to review and approve all new molecular animal breeding innovations. The FDA’s approach is flawed in several ways:

• Its rationale is not supported, and in fact is repudiated, by the history of innovative and safe animal breeding using various modern techniques to improve the characteristics of food animals, including by means of cloning and interspecies hybrids that are overseen by breeders, ranchers, farm organizations, states, and other federal agencies.
• It is inconsistent with the federal government’s overarching principle of biotechnology regulation, laid out in 1986 in the Coordinated Framework for the Regulation of Biotechnology. This principle holds that the regulation of recombinant DNA-modified organisms should be risk-based and avoid requiring process-based regulation and case-by-case review of all products, regardless of whether they are of high, moderate, low, or trivial risk.
• It is unnecessary, a solution in search of a problem. The FDA’s existing oversight of foods and food additives by its Center for Food Safety and Nutrition is sufficient to ensure the safe consumption of genetically improved animals. Other federal agencies already have jurisdiction over nonfood animals.

The FDA’s expanded proposed guidance has not gone unnoticed on Capitol Hill. On October 17, 2017, scores of members signed a letter to the secretary of agriculture, the FDA commissioner, and the EPA administrator raising serious concerns about the contradictions between the FDA’s approach and the USDA’s “thoughtful and science-based” regulatory approach, which did not expand its regulation to include all genetic modification techniques (as did the FDA), and complaining about the resulting negative impact of the inconsistencies domestically and internationally.

The letter affirms that once Congress has delegated authority to regulatory agencies, it is difficult for it to pull back the authority or even to oversee the myriad intricacies of daily regulatory activities. Although the passage of corrective legislation and the placing of limits on appropriated funds and personnel slots can address specific regulatory excesses, and the cumbersome and limited Congressional Review Act can provide ex post facto redress of executive branch actions, these processes are not easy to use. The current Congress is considering new proposals, such as the Regulations from the Executive in Need of Scrutiny (REINS) Act and the Regulatory Accountability Act, that call for greater congressional and Office of Management and Budget oversight.

Until Congress takes more control, it will fall to the White House to bring rationality to the oversight of biotechnology products. So far the Trump administration has not focused on biotechnology in its efforts to roll back and rationalize regulation. In the short term, the White House should direct the FDA to cease its efforts to regulate animals via the new animal drug framework and to clarify to what extent and how genetically engineered animals, whatever the techniques employed, will be overseen by the FDA’s Center for Food Safety and Nutrition, the USDA, and other federal and state agencies with oversight over animals.

Time is of the essence if the United States is to regain its competitive position in a field where other countries such as China are taking advantage of the void created by overbearing and dysfunctional US regulation.

What We Mean When We Talk about Workforce Skills

Labor Secretary Alexander Acosta stated recently that the United States “has more than six million open jobs, but some employers can’t find workers with the skills to fill them.” Before the 2008 recession, employers complained repeatedly of skill gaps and mismatches. With low unemployment, around 4.5%, they could not find the workers they needed. Yet the narrative did not change even as unemployment peaked, with almost 15 million people out of work in 2010. Since mid-2017 unemployment has been beneath postrecession levels and the refrain continues. Too many Americans lack skills suited to today’s economy—technology-intensive and postindustrial, exposed to global competition in services as well as manufacturing. In late 2017, the chief executive of Microsoft, Satya Nadella, told Bloomberg Businessweek “There are 500,000 jobs today in the tech sector that are open.” Some commentators go so far as to argue that skill gaps foster automation: if employers cannot find humans with the capabilities they want, they will buy robots and artificial intelligence systems.

These complaints lack context. Skill is a murky concept. It is hard to measure or observe; employers, political figures, and journalists use the term in confusing ways; and skill is too often associated with education, as if what people learn in school is all that matters. When employers speak of skills, they sometimes mean the ability to perform a particular task or job: using off-the-shelf or customized software for recordkeeping; selling enough ads to keep a struggling newspaper afloat; working as a nurse in an operating room. At least as often, employers and their HR (human resources) staff mean something quite different: social and interpersonal skills, and not just the obvious sort, the ability to get along with coworkers and deal with customers. In HR jargon, soft skills begin with showing up on time every day, doing as one is told, exercising ordinary courtesy—and also, if sometimes left unsaid, passing a drug test. As parents and schoolteachers know, these are far from universal among young people, or for that matter on Wall Street, in corporate boardrooms, and among elected and appointed officials.

Job seekers, on the other hand, claim to have been turned away even though perfectly capable of doing the work or quickly learning whatever they need to know. Some find the pay they’re offered less than merited by the job description and their experience. Others believe that employers think they’re too old or otherwise discriminate. In need of a paycheck, some take minimum wage, dead-end jobs. After the 2008 recession, many stopped looking and exited the labor force in frustration.

In bad times and good, millions of those counted as unemployed are simply in process of moving from one job to another, whether as a bartender or a biomedical engineer, part of the churn that is one of the constants of the labor market. According to the Bureau of Labor Statistics, some 5.2 million people left a job in December 2017, voluntarily or involuntarily, and 5.5 million took a new position. The bureau estimated the number of unemployed as 6.7 million, yet the long-term unemployed, those without a job for six months or more, totaled only 1.4 million, less than 1% of a labor force numbering more than 160 million people (by definition, the labor force includes those actively looking for work as well as those employed). People switch jobs for all kinds of reasons, after all, including better hours. Was it a surprise when a firm surveyed by the Minnesota Department of Employment and Economic Development responded, “It’s hard to get someone with that level of experience to work on a weekend”?

What employers want in the way of skills ranges widely. Companies ramp up their expectations when labor markets slacken, with many applicants for each vacancy, and relax them when supply tightens. Another company surveyed in Minnesota replied, “We were looking for experience in Atomic Force Microscopy.” Almost regardless of the job, employers expect basic literacy and numeracy. Employees may need to understand written instructions, if only safety manuals; they may have to compose emails or make change at cash registers. Beyond the basics, skills come in uncountable variety. Those of a dental hygienist are nothing like the skills and knowledge of an accountant, the skills of baseball players unlike those of snowboarders, the work of police officers resembles that of high school counselors in some respects and firefighters in others. Though some engineering tasks could be called “applied science,” engineering and science differ fundamentally, since the core objective in science is to understand nature, most commonly through reductive analysis, whereas engineering centers on conceptualization and design, culminating in the realization of some sort of human creation, a matter of synthesis, with science-based analysis, if present, in a supporting role.

It is easy enough to say that skills are multidimensional and mysterious. How many dimensions might there be? How should skills be subdivided? The questions are nonsensical. Skills overlap, combine, and melt together invisibly, defying measurement except in simple cases such as athletic performance or chess playing. Piece-rate wages, whether in sweatshops or in selling fur coats on commission, reward both effort and skill. Software firms can count the number of error-free lines of code generated by programmers. Otherwise companies are mostly guessing when they claim to pay workers, whether at the top or bottom of the salary scale, according to their contributions to productivity and profits.

Consider math skills. These range from addition and subtraction to differential equations and set theory. Many people find themselves counting things at work, but when manufacturers talk of statistical quality control, they do not mean that factory workers analyze statistical data; that is a task for specialists. For management, instruction in quality control serves as a motivational tool: do the job right or it may disappear, perhaps to Mexico or Vietnam. What is certain is that each of the many millions of people in the labor force embodies, in HR parlance, a unique “skill set.” Some of these skills will be universal and for practical purposes indistinguishable, others (C++ programming, German grammar) distinct and perhaps arcane. When published findings in molecular biology cannot be reproduced, did replication fail for lack of some exquisitely refined laboratory skill? Or did the original report err because of deficient practices? There is no way to tell until further attempts either succeed or fail and consensus develops, consensus among those accepted by their peers as proficient in experimental technique—or perhaps proficient in argument.

Useful tests exist for academic skills such as reading and writing, for the sort of manual dexterity needed at one time in cigar-rolling and still needed in brain surgery, and for more-or-less innate traits such as IQ. The Program for the International Assessment of Adult Competencies, developed under the auspices of the Organization for Economic Cooperation and Development to assess and compare cognitive and workplace skills, includes measures of “problem-solving in technology-rich environments” (meaning “computer-rich environments”)—a desirable step but a modest one, since the test questions mostly have to do with fluency in computer usage, as in navigating the internet and interpreting on-screen information, rather than addressing more complex sorts of problems that demand judgment. Beyond these sorts of test-defined capabilities, conceptual clarity recedes. Taxonomies go little beyond mental/manual/interpersonal categories, and it is one thing to include leadership among interpersonal skills, quite another even to catalog the ways people demonstrate leadership.

Distinguishing between skills and knowledge helps a bit. Anyone with even rudimentary skill in welding can join two pieces of low-carbon steel. And anyone with knowledge of metallurgy knows that although low-carbon steel can be easily welded, steels with high carbon content cannot; phase transitions on cooling cause the weld region to crack. Welders know not to attempt to join high-carbon steel, but few know much about phase diagrams. Few metallurgists, on the other hand, know how to weld.

Much of human knowledge has been written down somewhere, codified for accessibility. The storehouse is vast and continually expanding, yet at least in principle knowledge can be subdivided, marked off in bits and pieces, cataloged for maintenance in databases and updated as needed. Heat-treatment charts for steels complement phase diagrams and serve as guides to practice. If you do not know the average molecular weight of gasoline and need it for your work, you can find the value on the internet. Know-how also comes in tacit forms, meaning things we know, or think we know, but cannot articulate. As the philosopher Gilbert Ryle put it many years ago, “The wit … knows how to make good jokes … but cannot tell us or himself any recipes for them.” Leaving aside tacit knowledge, which as in stand-up comedy often merges with skill—a good joke badly told gets few laughs—the big mysteries then concern skills rather than knowledge. In principle if not in practice we might assess the domain-specific knowledge that people carry around in their heads; we could hardly imagine doing such a thing for their full complement of skills.

This has not stopped theoretically inclined economists. They divide skills into two categories, general and firm-specific. General or generic skills are those that people take with them as they move from one job or occupation to another. Firm-specific skills are assumed to have value, in the sense of contributing to productive output, only in a particular employment setting. In this way economists explain the reluctance of employers to train their employees except in narrow skills unique to the firm; investments in general training will be lost when workers leave.

Economists who labor in the subculture of “skill-biased technological change” make another assumption. Seeking to explain growing disparity in wages, with better-educated workers in the upper tiers of the wage distribution gaining while those in the lower tiers see static or declining pay, they link wages with educational attainment. In tribal shorthand, years of schooling become “skill.” This assumption reduces human skills to a unidimensional construct and sets aside as immaterial whatever people have studied, even at graduate and professional levels. On this basis economists conclude that those with more education earn higher wages because technological change since the 1970s has reduced demand for relatively routine work, jobs of the sort commonly taken by high school graduates, and, less obviously, that technological change—likewise taken to be a unitary construct—has increased demand for those with more schooling. The results leave much unexplained, yet appeal because parsimonious and because wages and education do, after all, correlate.

Unfortunately, policy-makers and the public have come to believe that more and better schooling can therefore be counted on to push wages higher, increase mobility, and wash away other labor market ills. With educators reinforcing the message, college has come to be seen as the road, perhaps the only open road, to a decent job, one with a future. The focus on education deflects attention from broader and deeper institutional problems in the labor market. There are plenty of reasons why schools should be improved and plenty of reasons for attending college. But more schooling for more people, rather than boosting wages and occupational mobility, could simply boost the numbers of “overeducated” Americans competing for whatever jobs the future may bring, a future that is bound to include further waves of technological and economic change, jobs paying whatever wages employers find they must offer to attract enough people to staff their organizations.

Learning, like skill, remains a considerable mystery, notwithstanding advances in psychology, neurosciences, and related fields, and a great deal of what we know and can do, beginning in infancy and continuing long past any terminal credential, whether GED or PhD, stems from learning that takes place outside the confines of formal education. Children learn quickly from raw experience. At first they grapple uncertainly with their surroundings, accumulating understanding more or less haphazardly. Once they grasp the meanings of spoken language and begin to talk, they go on to learn from other children, from parents and teachers, from the media. Adults of all ages likewise learn through mostly unconscious processes, often slowly and sometimes painfully.

What people learn, what they study in school, how long they continue in formal education, and how they fare in the labor market depend in part on intelligence as measured by IQ and other tests (there are many), and on much else as well. Most psychologists take cognitive ability to be inherent rather than learned. Cognitive tests, appropriately age-adjusted, show modest increases on average through the teenage years and modest declines in older adults, patterns quite unlike those for skills and knowledge, which increase rapidly from infancy, stabilize later, and at least for physically demanding skills diminish with advancing age. Many psychologists also believe that the so-called Big Five personality traits—extraversion; agreeableness; conscientiousness; emotional stability; openness to learning (each with their opposites, introversion to closed-mindedness)—capture the bulk of interpersonal differences. Although categories that are expansive in ways that IQ is not, they nonetheless can be related to intelligence; indeed, the last of the Big Five, openness to learning, has sometimes been called intellect (and alternatively, openness to experience and openness to intellectual inquiry, among others). Research further suggests that the third trait, conscientious, also called persistence, has the most to do with whether someone maximizes his or her potential, whether in scientific achievement or playing the trumpet.

Intelligence tests that come to be accepted based on large-sample statistical analyses correlate reasonably well with one another, and the measure that shows the closest correlations with the others has been labeled g, for “general intelligence.” Research shows, unsurprisingly, that g also exhibits the closest correlations with educational outcomes and with available measures of workplace competencies and labor market outcomes, whether subjective (supervisor ratings) or more nearly objective (wages, advancement along career trajectories). But IQ, g, and other measures put more weight on what people know than on how well they make use of what they know. Plenty of smart people, after all, do dumb things. Thus young people who understand full well that success in school matters for finding a good job may still drop out. In the end, measures of intelligence take us only a little ways in grappling with notions of skill.

The simple answer is from experience, at school and on the playground, at home and at work. A gifted few, very few, can learn just about anything they want or need on their own. For the most accomplished scientists, after all, education is simply the starting point in careers marked by continued learning, discovery, and invention. For others, school is little more than a place to make friends, pick up basic skills, and become accustomed to the sort of regimented discipline found in many workplaces.

Schools seek first to convey codified knowledge and skills, treating these as standardized and formulaic even in many college courses. Postsecondary occupational preparation likewise begins in the classroom, then may follow an apprentice-like model, familiar from art classes and the construction trades, that combines formal instruction with less structured practice. Medical students learn the basics of physiology and anatomy, of suturing and bone-setting, of listening to patients and evaluating symptoms. Suturing is mechanical. Diagnosis is not, and thus far more difficult to teach or to evaluate. And once physicians leave school, some, like their counterparts in other occupations, continue to learn while others do not. Studies of continuing medical education find little or no change, on average, in clinical practice. Physicians have few incentives to keep up with advances in medicine since customers cannot evaluate their skills. After all, who can really tell whether the family doctor falls in the 40th percentile of skills or the 90th percentile? Or for that matter their child’s third-grade teacher? Much the same holds for auto mechanics and financial advisers.

Schools get much blame for workforce skill deficiencies. Since the US Department of Education’s 1983 report A Nation at Risk, the public has repeatedly been told that primary and secondary schools fail to prepare young people for work (as if that was their chief task), that too many leave high school unready for college, and that shortages of STEM (science, technology, engineering, and mathematics) graduates threaten the nation’s economic future. US students rank in the middle or below in international surveys of educational achievement, and with calls for school reform going back generations, the reasons, and the appropriate responses, continue to be argued over. Recent popular proposals include secondary-school courses in what’s usually called computer science. In some cases this seems little more than training in off-the-shelf software (high school students in North Carolina can earn certification in Microsoft Office) and in others programming. Apple’s chief executive, Tim Cook, told Fortune magazine “we think that coding is the sort of the second language for everyone in the world. And that’s regardless of whether they’re in technology or not. I think that you don’t have to be in technology for coding to be very important.” Few of the four million or so people with jobs in what the Bureau of Labor Statistics considers one of the computer occupations will ever write or even look at code, and computer science as a discipline is more about systems and problem-solving than line-by-line programming, just as writing an essay is more about what is said than spelling and grammar. Even so, learning to code in Java or Python can be a useful step toward rigorous thinking, much like algebra, and for some people a point of entry into a rapidly growing field full of opportunities to learn more and advance, a field in which experiential learning will almost certainly remain as important as in the past. Information technology simply could not have expanded as it did if dependent on academic training: employment has increased much faster than the number of graduates from specialized programs, and people continue to pick up skills informally, in the workplace and outside it, from peers, through trade and professional societies, from vendors. These forms of learning should be encouraged, systemized, incentivized, and supported, and not just in information technology. This is, after all, how postdocs are supposed to get their start in research and how established scientists stay at the frontiers of their fields.

Employers interview job candidates and expect letters of reference for some positions, but they rely more heavily on educational attainment and credentials, which as screening devices have proven far more reliable. Like supervisor evaluations of incumbent workers, interviews and references mean little. Educational attainment, on the other hand, reflects thousands of hours spent in ordered and supervised settings not unlike those of typical workplaces. When hiring, then, firms place more weight on a high school diploma than a supposedly comparable GED, on a two-year degree from a reputable community college than a seemingly comparable credential from a for-profit school (the more so if known for shady practices and underperforming graduates), on a four-year degree from a “good” university than from one with a lesser reputation. More than 20 million people enroll in US colleges and universities each year, and around 60% of them eventually graduate, according to the Department of Education. Respectable grades from a respectable school signal ability to learn and also discipline (showing up), persistence (turning in work on time), agreeableness (please the teacher, please the boss), and social skills sufficient to stay out of too much trouble.

Although STEM occupations make up only a small slice of the workforce, around 6%-7% by most accounts, since the late 1940s handwringing over the numbers of graduates has been a “central, at times hysterical concern,” in the words of physicist and science historian David Kaiser. During the Cold War, educators and government officials claimed the United States risked falling behind the Soviet Union in engineering and science. With détente and the later collapse of the Soviet empire, economic competition, first with Japan and then with China, became the new challenge, to be won through innovation and productivity growth, hence research and development and STEM graduates.

STEM training has undeniable value across wide swaths of the economy, as demonstrated by the jobs held by those with engineering and science degrees, over half of which fall into non-STEM occupational categories. Although occasionally portrayed as underemployment, or a misallocation of human capital, this is better seen as an indicator of demand for employees with good quantitative and analytical skills, regardless of field of application. Indeed, it would seem hard to argue that there could be an oversupply of STEM graduates, and for such reasons the acronym seems unobjectionable as an exercise in branding. The attention recently focused on STEM might even be taken to suggest some sort of inchoate yearning for a citizenry better equipped to respect specialized knowledge and hard-headed empiricism, rationality in a word. At the same time, many Americans—around one-fifth, according to the most recent surveys—continue to find their way into STEM occupations without relevant academic credentials. Mobility of this sort has always been a strength of the US economy, likewise to be encouraged.

Despite the wage premiums that come with degrees from well-regarded colleges, Fortune tells us that “In its early years, [Google] had recruited from elite schools like Stanford and MIT. But when Google examined its internal evidence, it found that grades, test scores, and a school’s pedigree weren’t a good predictor of job success. A significant number of executives had graduated from state schools or hadn’t completed college at all.” Indeed, “the proportion of people without any college education increased over time.” Meanwhile a September 2017 report from the Government Accountability Office titled Low-Wage Workers found that “the percentage of workers with college degrees earning $12.01 to $16 per hour increased from 16% in 1995 to 22% in 2016.”

Recognition of these sorts of dynamics has spread, with pundits, politicians, and some employers pointing to apprenticeships and occupationally oriented community college programs (often called career and technical education) as underappreciated alternatives to four-year degrees. Yet in the United States apprenticeships have never spread much beyond construction, where poaching of trained workers by nonunion contractors has curbed participation by unionized firms, and companies in other industries that once trained employees or subsidized tuition no longer do so. In most parts of the country, community colleges have been underfunded from their beginnings, and states have been cutting support for four-year public institutions. As the Wharton School’s Peter Capelli argues, businesses have succeeded in pushing off more and more of the costs of workforce preparation onto the public sector. Leave it to the schools, seems to be the message. The response from government: Leave it to student loans. As policy, this amounts to viewing human capital as a purely private good rather than a national resource. It also assumes that young people and their parents are fully capable of assessing the content and quality of yet-to-be-delivered educational services, even though these are hard-to-judge intangibles, the value of which will depend on their suitability for careers extending decades into the future. This is like treating purchases of education as equivalent to buying an off-the-shelf smartphone or an insurance policy.

Put simply, and a bit crudely, they want it all: high skills with low wages; big pools of well-educated job candidates from which to pick and choose potential superstars—engineers, computer scientists, and entrepreneurial managers who just might seed a billion-dollar line of business (the reason that Silicon Valley, like Hollywood, obsesses over “talent”); obedient employees, whether counter workers or data scientists, who when asked can nonetheless bring creativity and independent initiative to the workplace; freedom to hire and fire at will. And they’d like the public to pay the costs of preparing workers to fit their job openings. These expectations underlie many of the complaints over skill gaps, mismatches, and shortages.

Much of what employers want they get. Among wealthy nations, US employment law has always been uniquely friendly to business and hostile to workers. Into the twentieth century the courts typically accepted, in the words of historian Melvin Urofsky, “the legal fiction that employer and employee stood as equals in the bargaining process.” Labor-friendly legislation, whether barring child labor or wages paid in scrip accepted nowhere but the company store, might make it through state legislatures only to be struck down as interfering with freedom of contract. Workers, said the courts, could always say no (and shouldn’t a father be free to contract out his offspring to work in the mine or the mill?). Then as now, employers also used implicit or overt threats to move production out of state or out of the country to hold down wages, salaries, and benefits. Other regulations governing business conduct, such as antitrust policy and enforcement, might wax and wane; the general rule in labor law has always been weak standards, weakly enforced. Firms that take differing positions in Washington on trade policy and compete vigorously for favors from local and state governments find their interests aligned when it comes to employment law and policy.

Following something of an interregnum during the Great Depression and World War II, the 1947 Taft-Hartley Act marked the beginning of resurgence in employer prerogatives, and when the Reagan administration further deregulated labor markets, “right to work” laws—resurrected and refurbished on the model of union-busting strategies devised a century earlier—resumed their spread. Despite legal prohibitions enacted since the 1960s against racial and other forms of discrimination, employers remain largely free to fire employees with or without cause. Low minimum wages hold entire families beneath the poverty line. Since the late 1970s the share of productivity gains delivered to employees in the form of compensation has fallen and the share flowing to owners in the form of profits and dividends or held as retained earnings has risen; while economists search, so far in vain, for technical explanations, it seems at least as likely that the turn away from what had seemed a well-established pattern of wages increasing in step with productivity growth reflects a still-continuing rise in the political power of business interests and weakening of labor. Underlying these trends since the 1980s has been the fraying of social safety nets for individuals and families, but not for businesses.

Without too much risk of overgeneralization, we can conclude that many—not all—claims of skill gaps and mismatches reflect misunderstandings of the nature of skills, the wishes and expectations of employers and their hiring practices, and the purposes of public education.

• Skills, unlike knowledge, cannot in general be observed or measured except in the simplest cases, as for basic literacy and numeracy.
• Companies often state that they cannot find workers with needed skills, yet at the same time decline either to raise wages or provide training.
• To employers, educational attainment signals potential. They evaluate applicants on the basis of credentials, not necessarily because of what job candidates may have learned in school (beyond basic skills, and with exceptions as for STEM occupations and professional degrees), but because a diploma or degree signifies, first of all, willingness to work and to conform. The weight given to credentials such as degrees from Ivy League schools also helps preserve social hierarchies.
• Many firms, and their supporters in policy circles, behave as if private interests coincide with public interests and the chief task of education is to prepare young people to slide seamlessly into available jobs—in today’s workplaces, which will not be tomorrow’s.

Viewing education and training as public goods, which they are, would shift our perspective to preparation for the future. Some policy-makers do recognize this. In February 2018, David Long, president pro tempore of the Indiana State Senate, told The Hill that “Fifty years from now, half the jobs that we know of today will be gone.” This may prove an understatement. Yet policy-makers continue to tinker at the margins of schooling rather than seeking to build platforms for continuing learning of the sort that many Americans in computer occupations had to improvise. And we seem to have let educators highjack “lifelong learning,” now taken almost reflexively as meaning “go back to school” (and then do it again), which is hardly realistic for most of those already in the workforce.

Rising inequality in wages, benefits, and wealth has been one of the most pronounced shifts in US society over the past half-century. Higher wages accompanied by employer-supported training would dissolve most of the concerns over skill shortages in the labor market and would begin to counter, if not necessarily reverse, economic inequality. Basic literacy and numeracy, willingness to learn, and flexible access to high-quality training on a continuing basis comprise the necessary starting points. Tax revenues are the proper source of support. Studies of actual skills deployed in actual workplaces, based on careful observation rather than HR talk, show that relatively few workers need high-level specialized skills. What everyone needs are opportunities to learn and to advance in accord with their wishes and motivations without sacrificing already meager paychecks.

What is “Fair”? Algorithms in Criminal Justice

On rare occasions, new technologies open up straightforward routes to a better world. But on many other occasions, the mirage of a simple path forward fades quickly. In the case of social policy algorithms, the promise was that systems from hiring to criminal justice can be improved through “objective” mathematical predictions. Today, however, communities of scholars and practitioners are calling for a closer look before we leap.

A “Governing Algorithms” conference at the University of California, Berkeley, a few years ago explored the “recent rise of algorithms as an object of interest in scholarship, policy, and practice.” A group called Fairness, Accountability, and Transparency in Machine Learning (FAT-ML) formed soon after and now holds meetings each year. And New York University recently established the AI Now Institute, with scholars dedicated to investigating the “social implications of artificial intelligence.” These conferences and communities aim to bring together researchers and practitioners to think through the use of algorithms in society. A common theme is the concern that predictive algorithms, far from leaving bias in the past, carry both the risk of bias and increased threats of unaccountability and opacity.

Take the use of algorithms in the criminal justice system. Although algorithms have been used in some form in criminal justice decision-making since the 1920s, they are gaining wider use in areas such as pretrial decision-making. The algorithmic tools take in a variety of inputs, ranging from just a few variables to over a hundred, and assign defendants a risk score based on probability of rearrest, failure to appear in court, or both. The score is then often shown to judges, who may choose to release defendants with low risk scores on their own recognizance or under some form of limited supervision. The others are held in jail or assigned bail (which often means remaining in jail because they lack the money to pay bail). For some criminal justice reformers, the hope is that the use of the tools will help to reduce jail populations. And in some places, including New Jersey, the jail population has decreased after adopting pretrial risk assessment algorithms such as the Arnold Foundation’s Public Safety Assessment (PSA) and abandoning use of bail. But there has also been criticism of the PSA and other algorithms.

In 2016, journalists at ProPublica criticized as unfair the risk assessment tool called COMPAS, developed by the company Northpointe (later renamed Equivant). After analyzing data they obtained from a jurisdiction in Florida that uses the algorithm, the reporters concluded that the algorithm is racially biased. They found that among defendants who are not rearrested within two years, 45% of those who are black—compared with 24% of the whites—had been assigned high risk scores. Yet when Northpointe responded to the critique, they pointed to a different statistic, supporting a different sort of fairness: within each risk category, black and white defendants had the same rearrest rate.

Could the algorithm have satisfied both conditions of fairness? A number of academic researchers have argued otherwise, purely on mathematical grounds. The reason is that there are different base rates of rearrest among the racial subgroups. Because the groups are different in that regard, achieving one kind of fairness automatically means that the other will not be achieved. To return to Northpointe’s kind of fairness: when a judge sees a particular risk score, he or she can infer that this indicates the same chance of rearrest, regardless of the race of the person given the score. That is, arguably, a kind of fairness.

But achieving this kind of fairness creates another kind of unfairness. Because black defendants are arrested at higher rates, and because criminal history is a significant input into risk-assessment tools, a greater percentage are assigned high risk scores than white defendants. This affects other metrics, including the percentage of those who are not rearrested and yet are deemed high risk, which was the bias that ProPublica pointed out. Since being deemed higher risk makes it more likely that someone will end up in jail, this means there is a disparate impact for black defendants.

However, remedying the kind of unfairness that ProPublica points to has its own problems. It involves a trade-off with the first kind of fairness (i.e., risk scores that mean the same risk of rearrest, regardless of race). After exploring the trade-off between these two kinds of unfairness, a team of Stanford University researchers wrote in the Washington Post, “Imagine that…we systematically assigned whites higher risk scores than equally risky black defendants with the goal of mitigating ProPublica’s criticism. We would consider that a violation of the fundamental tenet of equal treatment.” In other words: it is a difficult and open question whether it would be ethical or constitutional to design an algorithm that treats people differently on the basis of race, in order to achieve the kind of fairness to which ProPublica points.

Richard Berk, a statistician at the University of Pennsylvania who has developed criminal justice algorithms, emphasizes the trade-offs involved in their design. In a paper on fairness in assessment algorithms, Berk and coauthors write, “These are matters of values and law… They are not matters of science.” They discuss kinds of fairness and note that increasing one kind of fairness can often involve decreasing other kinds of fairness, as well as reducing the predictive accuracy. They argue that it is mathematically impossible to achieve what they call “total fairness” in one algorithm. The authors conclude that “it will fall to stakeholders—not criminologists, not statisticians, and not computer scientists—to determine the tradeoffs.”

Arvind Narayanan, a computer scientist at Princeton University, added in a recent talk that input from philosophers is needed. There is something incoherent, he said, in laying out technical definitions of fairness without bringing to bear the long history of work on justice and fairness in ethics. “It would be really helpful,” he suggested, “to have scholars from philosophy talk about these trade-offs [in algorithms] and give us guidelines about how to go about resolving them.”

Another significant normative challenge occurs when decisions are made about which variables (or “features”) are included as inputs in the model. Although no algorithm includes race itself an input, some include features that are highly correlated with race and socioeconomic status. Including these variables will run the risk of increasing the disparities in how different communities are treated. By being more likely to deem people as high risk, for instance, they may increase their rate of being held in jail. And there is strong evidence that people who are held in jail as they await court hearings plead guilty at considerably higher rates than do people who are released. The resulting conviction would then serve as an additional data point held against them the next time they are arrested, leading to a vicious circle. To add to that, a criminal record can severely limit housing and job opportunities. Human Rights Watch has expressed this concern and has also pointed out that the use of algorithms does not guarantee a reduction in jail populations.

A third consideration in algorithm design involves weighing the trade-off between the harm done when someone is released and commits a crime and the harm done by holding in jail people who would not have committed any crime if released. In other words, there’s a key question of how much to prioritize safety risks over harming people and their families through incarceration.

Finally, a fourth consideration, discussed by legal scholars such as Megan Stevenson and Sandra Mayson, is the question of which outcomes to try to predict. Many tools predict rearrest for any offense, including for low-level crimes and drug offenses. Focusing on the outcome of rearrest for a violent charge might make more sense, scholars suggest, if the primary concern is violent crime.

Connected to all of the normative questions is an (applied) epistemological one: How can we know what decisions were made in designing the algorithms?

Critics of algorithms have also pointed to lack of transparency as a major problem. In 2016, a defendant in Wisconsin who pleaded guilty to eluding the police and operating a vehicle without its owner’s consent sued the state, claiming that reliance on an algorithm (Northpointe’s COMPAS) violated his rights to due process, in part because he was not able to see the algorithm’s code. The company refused to share the algorithm, claiming that the tool’s workings are proprietary. The Wisconsin Supreme Court eventually ruled against the defendant, and an appeal to the US Supreme Court was rejected. It also should be noted that the inputs and weights of several other algorithms, including the PSA, are publicly available.

Still, several big questions remain when it comes to transparency. Should individual defendants be notified about what their scores are, and should they have a right to see how the scores were calculated? Beyond that, should other people—for instance, independent researchers—have the access required to check a given tool for computational reproducibility, or whether they arrive at the same results, using the original data and code? In many scientific domains, there’s been increasing concern about the reliability of research. This has been a major topic of discussion in fields such as psychology and biomedicine. Recently it’s come up in criminal justice as well, with researcher David Roodman at the Open Philanthropy Project finding major methodological problems in seven of the eight criminal justice studies that he reanalyzed. Who, if anyone, is able to check the reproducibility of algorithms, in terms of the original calculations of the designer and whether they are relevant to the population to which they are being applied?

On the accountability of algorithms, there are major outstanding questions as well. How can stakeholders weigh in on values they’d like to see the algorithm represent? As we saw above, there are different concepts of fairness, and these concepts interact in complex ways with each other and with accuracy. As more and more jurisdictions adopt assessment algorithms, will not only policy-makers but also communities affected by the tools be able to provide input? After all, these are not objective decisions, but nuanced normative decisions with the potential to affect many thousands of lives.

The challenges are daunting. Yet at the same time, as many commentators have pointed out, the baseline is the current US criminal justice system, which on any given day holds in jail nearly half a million people who haven’t been convicted of a crime. Some defenders of algorithms also point out that judges conduct their own internal algorithmic decision-making, the workings of which are not transparent and can be affected by factors as irrelevant as whether the judges have eaten lunch. As long as risk-assessment algorithms stand a good chance of improving the status quo, proponents argue, it’s worth the effort to work through the challenges.

This argument seems to be a good one, but at the same time, pretrial risk assessment should be used only under certain conditions. First, a tool should be made transparent. At the very least, it should be made public what inputs, outputs, and weights are used. Second, a tool should be regularly audited by independent researchers to check its design and its real-world impact on jail populations and racial inequity (the AI Now Institute has called this kind of check “algorithmic impact assessment”). In order to check effects, data need to be regularly collected and shared with auditors. Third, the wider community of stakeholders beyond algorithm designers should be included in deciding what an algorithm should look like: What kind of fairness is important to prioritize? What threshold of risk should be used for release, and what kind of risk makes sense to measure in this context? (As discussed above, risk sounds as if it’s referring to public safety, but the algorithm measures the chance of any kind of rearrest, including many activities that pose no danger to the community). And finally, the tools should be used only to reduce jail populations, not to lock up more people unjustly as they await court hearings.

Many jurisdictions have adopted one pretrial tool or another, and as the legal scholar Sonya Starr writes, “It is an understatement to refer to risk assessment as a criminal justice trend. Rather we are already in the risk assessment era.” Yet at the same time, it’s important to recognize that risk assessment is only as good as the decisions that go into designing the tool. We should hold the tools and their designers accountable, and should use the tools only if they demonstrably serve the goal of dramatically reducing pretrial incarceration.

Economic Policy in the Time of Reactionary Populism

Addressing the nation’s political unrest will require a rejection of some cherished economic dogmas.

Political developments in the past year—notably the 2016 electoral campaign in the United States, and Brexit in England—have revealed a depth of anxiety and resentment among citizens in many rich nations that go well beyond the economic environment and probably cannot be addressed by economic policy alone. Nonetheless, the economy appears to have been the key to the political upheaval through which we are now living, and though solutions to these problems may no longer be enough to stabilize the political and social environment, it is difficult to imagine how we can restore a sense of order without addressing the economic concerns.

Those concerns are the product of the pressures for structural change and adjustment that have battered the economy over the course of at least the past 30 years. Pressures for structural change and adjustment are of course inherent in any dynamic economy; indeed, they are the engines of economic growth and development. One can argue about whether the recent pressures have been greater than those that the economy has absorbed in the past, but in the United States, at least, there is no question that whatever their absolute magnitude, their costs have been very concentrated in the old industrial heartland of the Midwest, where the communities in which people’s identities were embedded have been undermined and to some extent abandoned. These communities were a key constituency of the Democratic Party, and their desertion of the party was the determining factor in the electoral victory of Donald Trump.

The principle forces producing the structural changes against which the Midwest electorate was reacting were globalization and technological change. But they have been aggravated by institutional changes in corporate governance associated with financialization. Most important, in my view, the country has been guided in its response to these pressures by an approach to economic analysis that makes the forces producing these changes seem beyond the control of politics and policy, and thus cripples our ability to anticipate the problems that globalization and technological change have engendered, and to conceive of alternative solutions.

The analytical approach I single out can be understood in terms of what might be called policy paradigms, the broad frameworks through which policy-makers tend to think about the economy, judge its performance, and attempt to influence its direction. In the post-World War II period, policy has been guided by three such paradigms: a Keynesian paradigm in the immediate postwar decades; the so-called Washington consensus, emerging in the late 1970s and continuing through the 1990s and into the new millennium; and more recently what might be called the Silicon Valley consensus encapsulated by a mantra along the lines of “innovation and entrepreneurship in the knowledge economy.” The Silicon Valley and Washington consensuses, in turn, are each linked to globalization in a way that almost amounts to an additional paradigm. In the Silicon Valley consensus, globalization is seen as the product of innovations in communication and transportation. In the Washington consensus, it is promoted by trade treaties and innovations in regional and international governance conceived as an expression of the efficiency of a market economy as understood in terms of standard economic theory.

That such paradigms exist and that they vary over time is difficult to deny. Where they come from and what role they actually play in the evolution of the economy is on the other hand unclear: Do they reflect social and economic reality, or do they actually influence and direct its evolution? Are they, to borrow a phrase from the University of Edinburgh sociologist Donald MacKenzie, a camera or an engine?

The difficulties that conservatives in Britain and Republicans in the United States are having in translating the political reaction that brought them to power into a coherent program bring this question to the fore and suggest the intellectual vacuum in which the political reaction is taking place. These are ominous developments, given the way communism, fascism, and two world wars grew out of the collapse of what my MIT colleague Suzanne Berger calls the “first globalization” in the early twentieth century. And the parallels and dangers were reinforced in the summer of 2017, at least in the United States, in Charlottesville, Virginia, by the marching of youths shouting Nazi slogans, and seemingly encouraged by President Trump. But this is not the 1930s. Today, the reaction against globalization is occurring without the kind of alternative paradigm that communism and fascism offered in the earlier period, and this is a good thing, because it creates an opportunity to meet the challenge with a new series of ideas.

It would be premature to try to specify what an alternative policy paradigm might look like. But at least with respect to issues surrounding income distribution and human resource policy, one can, I think, identify some of the limits of the prevailing paradigm that any alternative would have to take into account.

The country has been guided in its response to these pressures by an approach to economic analysis that makes the forces producing these changes seem beyond the control of politics and policy, and thus cripples our ability to anticipate the problems that globalization and technological change have engendered.

Two pillars of the Silicon Valley consensus are “innovation” and the “knowledge economy.” Considered together, they imply the evolution of labor toward sophisticated technology that must be created and managed by highly skilled workers. When additionally combined with globalization, this evolution implies that as legacy industries become increasingly dependent on low-cost, low-skilled, uneducated labor, they will migrate to the developing world, meanwhile abandoning that part of the US labor force that cannot be absorbed into the high-tech sectors.

Implicit here are views of both technological change and knowledge that are highly suspect, and that distort policy and investment decisions in public and private sectors alike.

The view that technological change must always require a more skilled workforce is particularly suspect given that it is not attached to any real theory or strong empirical evidence about the direction of such change. Rather, it is promulgated in a world in which there is a belief, reinforced by the Silicon Valley consensus, that pursuing a broadly predictable direction of technological change is the key to economic prosperity, for individual actors and for cities, regions, and nation-states alike. Such a deterministic view about knowledge and technology is bound to play a role in what projects inventors choose to work on and which ideas entrepreneurs and financiers choose to develop. Its role in this respect seems to have been enhanced by the changes in corporate governance associated with financialization. As business has become increasingly dependent on outside financing, outsiders have become increasingly influential in business decisions. And management is called upon to justify decisions that depart from fad and fashion to people who are not in a position to form an independent judgment about what the business is doing.

But the commitment to the development of Silicon Valley technology goes well beyond a diffuse consensus that influences private decision-making. In the United States, the federal government plays a pivotal role in the evolution of technology. Though the government finances less than a third of national expenditures on research and development (R&D), it has been responsible for key innovations since World War II in areas as diverse as communications, materials, aviation, and biomedicine. Its role remains important today, for example in financing and promoting robotics technology, most prominently through the DARPA Robotics Challenge, a competition funded by the US Defense Advanced Research Projects Agency. As the loss of manufacturing jobs became a major concern of public policy in the Obama administration, the federal role included a new program to promote advanced technology in a way that—ironically—sought to preserve manufacturing by reducing the employment requirements and increasing the educational requirements of the jobs that remain. To the extent that this program sought to address the gap between worker qualifications and job requirements, the focus was on raising worker qualifications, rather than promoting technological developments that could lower job requirements and hence bridge the distance between the existing labor force and employment requirements of the labor market.

At the same time, the consensus that technology must evolve toward an economy that demands a more technically qualified workforce leads policy-makers to tilt investments in education and training toward the formation of engineers and scientists and, more broadly, toward institutions of higher education, rather than, for example, primary and secondary education or vocational training, or training on the job. Developing economies such as China, India, and the Philippines have also bought into this deterministic view of knowledge and technology, and have overinvested in higher education, resulting in the emigration of a part of their educated labor force to North America and Europe, and facilitating the movement toward the advanced technologies embodied in their skill sets. The widely shared belief in the inevitability of this kind of technological change has led to a virtual panic about the availability of skilled and highly trained workers, perhaps best exemplified by the influential 2007 National Academies report Rising Above the Gathering Storm. This despite the fact that half of the graduates in science, technology, engineering, and mathematics—the STEM fields—trained in the United States are working in non-STEM jobs and occupations. Thus, there is an interaction between the policies of the United States and those of India and China that is leading to increasing immigration of highly educated workers from abroad, as opposed to adjustments in job design and recruitment practices at home in which business might otherwise be forced to engage.

The second problem with the Silicon Valley consensus is that the “knowledge” that it so valorizes through tropes such as “the knowledge society” is exclusively formal knowledge acquired through classroom learning in distinct educational institutions, and carried into the productive sector by students who graduate from these institutions and by their professors working as consultants and entrepreneurs. This view of knowledge does not recognize tacit or clinical knowledge, acquired on the job in the process of production. Clinical knowledge, moreover, appears to evolve informally through practice as less educated workers, working alongside formally trained engineers and managers, gradually take over many of their tasks, actually inventing other ways of doing the job and understanding the work. The relationship between formal and clinical knowledge is unclear in large part because clinical knowledge is seldom explicitly recognized, and because it goes unrecognized it is understudied. Recognition is complicated by the fact that tacit knowledge is by definition immeasurable and its nature, even existence, draws on anecdotal evidence that is easily dismissed as atypical or anachronistic and that is destined to be replaced by the kind of formal “scientific” analysis that we think of as characteristic of modernity.

Given that we think of information technology as emblematic of contemporary modernity, the role of tacit knowledge in the development of software is especially relevant to my point here. Efforts to standardize and formalize software development have proven particularly frustrating, and instead rapid, efficient development depends on the tacit understanding embedded in a community of practice that grows up through direct, personal interaction among a team of developers and the architects and designers of the programs they are attempting to write. Thus, for example, General Electric, when it began offshore development in India, found that it needed at least 30% of the Indian workforce to be physically present in the United States at any given time to engage in face-to-face contact with their US contractors. And Frederick Brooks in his famous 1975 treatise on software development, The Mythical Man-Month, argues that adding new people to a development project as it falls behind schedule actually slows down the development process even more because the newcomers do not share the tacit understanding of the architecture, an understanding that can be developed only through interaction with experienced members of the team on the job. For certain processes, clinical knowledge simply seems to be indispensable. But even where clinical and formal knowledge can be substitutes for each other, clinical know-how should provide opportunities for upward mobility to workers for whom a lack of resources or formal educational preparation bars access to formal education. But clinical knowledge goes largely unrecognized and certainly undeveloped when employers are able to recruit formally trained labor abroad, a bias reinforced by public policy.

The third problem with the Silicon Valley consensus emerges from a widely recognized and almost universally ignored intersection with globalization. Both technology and trade policies in recent years imply fundamental structural changes in the economy. As professors are quick to point out in elementary economics courses, such changes typically generate both gains and losses. The structural changes are desirable and the public policies that encourage them justifiable if the gains outweigh the losses—if, in other words, there are net social gains. Where this is the case, the gainers can compensate the losers. But in fact net gains are not enough to justify such policies. In the conventional theoretical framework, the structural changes are justified only if the compensation is actually paid. In practice, compensation is almost never actually paid. Nor is this surprising: No institutional mechanism is in place to ensure that compensation will be paid. Indeed, no institutional mechanism exists for systematically weighing the gains against the losses to determine whether there is a net social benefit. The people who make the critical decisions and reap the gains are not generally linked to the people who experience the losses. In fact, it is not usually possible to trace worker displacement to particular causal factors, and certainly not the displacement of a particular worker. If, as I and many others have argued, technological change and globalization have been imposed together on the same communities so that alternative employment opportunities have been limited and ancillary economic activities in these communities destroyed, those costs cannot be directly related to any identifiable gain, so policies that provide for direct compensation are not even theoretically possible.

The key decisions that affect structural change—in technologies and in trade—are made by institutional actors who reap the benefits of these changes but escape accountability for the social cost they have created.

The major exceptions here are programs designed to provide training (or rather retraining) to workers displaced by globalization. Such programs exist in virtually all advanced developed countries. And in the trade debate ignited by the US presidential campaign, the only concrete policy that the Democratic candidate, Hillary Clinton, proposed in response to the criticism of her earlier support of trade was an expansion of adjustment assistance of this kind for displaced workers. But such programs are everywhere also very limited in scope, and their success has been limited as well, even for those displaced workers who actually get to participate. Studies suggest that the returns to participation in such programs relative to control groups of similarly displaced workers who do not participate are barely enough to yield a positive return on the investment, let alone actually compensate workers for the loss of previous jobs. The reasons for those failures are complex, but the basic problem is the institutional difference between the schools that run the retraining programs and the businesses that would have to hire the graduates of the programs. Schools and productive enterprises have different missions and face different constraints and different incentives.

To take one example, schools face a hard budget constraint that leads them to minimize the wear and tear on the equipment and wastage of material used in the teaching process, whereas businesses are willing to tolerate equipment damage and material scrap in order to meet tight delivery schedules. Thus, schools train workers to be, from a business perspective, overly solicitous of equipment and material consumption, and in the eyes of business, graduates from school training need to be retrained; often it is easier to hire untrained workers than to break what business views as the bad habits cultivated by the schools. Unless the enterprises take an active interest in the schools and intervene to mold the programs to their needs, the schools do not produce graduates who are useful to employers.

That enterprise participation is important is now widely recognized, particularly in the literature on vocational education and community colleges. What is not recognized is that getting the enterprise to take an interest in the schools is itself an institutional problem: The firms have no incentive to do so if they can find trained workers more easily by poaching from other enterprises or recruiting immigrants from other countries. Most programs try to recruit business participation by appealing to civic responsibility. But businesses are getting appeals for help from the Boy Scouts, breast cancer patients, and homeless advocates too, so it is unclear why they should put displaced workers at the top of their civic responsibility list.

Ideally, the pace of employment displacement would be held to the rate of natural employee attrition; for a variety of reasons, such a goal would be very costly and difficult institutionally to achieve (although government restrictions upon layoff and discharge do work in this direction). But policy-makers do not even know how the impact of the different treaties and technical innovations they have promoted in recent years relate to each other. In retrospect, it seems crazy that in the waning years of the Obama administration, when the political reactions to job dislocation in the form of the Trump and Bernie Sanders candidacies were already gaining momentum, policy-makers were promoting two new major trade treaties, one with the countries of the Pacific Rim and the other with Europe. (The “fast track” process through which trade treaties are reviewed by Congress encourages the negotiation of multiple treaties at the same time.) What could have been presented as a question of the pace and timing of globalization was instead made an issue of being either for or against globalization itself. Moderation in the pace of change could in fact have been built into the treaties, and their impacts spread out over time, but this was never considered.

Indeed, the analytical frameworks implicit in the Silicon Valley and the Washington consensuses would never lead to these considerations. In this sense, the failure to consider time and geographic dimensions of trade are basically symptoms of the general problem with the paradigms in which policy has been conceived in recent years, especially in their treatment of structural change. The paradigms all imply major changes in the structure of the economy; they promote and celebrate such changes, viewing them generally as producing desirable increases in social welfare, at times arguing that such changes are inevitable. But they are focused on the end point of the changes they advocate, and have very little to say about the process of change, about alternative paths of adjustment. They foresee an increase in overall social welfare, but offer no insight into the costs as well as the benefits, nor as to how those costs and benefits will be distributed. Thus, the whole debate around the North American Free Trade Agreement, arguably the pivotal point in the US turn toward globalization, was conducted in terms of what economists call “computable general equilibrium models,” focused as the name implies on a comparison of equilibria under the new and old trading regimes, with no capacity to consider the political and social costs of the transition from one regime to the next. But such costs may be very real indeed. Consider the sudden end in 2005 of the Multi-Fibre Arrangement, which had governed the world trade in textiles and garments over the preceding 30 years and imposed quotas on the amounts that developing countries could export to developed countries. Its end led directly to an abrupt concentration of production in Bangladesh, which the real estate market there did not have time to anticipate. Production facilities moved into hazardous buildings, one of which collapsed, causing thousands of fatalities. Other tenants had moved out of the building when it was condemned, but the garment manufacturer remained, fearing that if it moved it would miss the tight production deadlines imposed by the international brands and be blacklisted as a result, denied any contracts in the future.

The Silicon Valley consensus, and the Washington consensus that preceded it, have thus had the effect of divorcing the process of adjusting workers to jobs and to new technologies from the productive process itself. The result is that forms of learning and understanding that would facilitate worker adjustment are neglected. Most importantly, the key decisions that affect structural change—in technologies and in trade—are made by institutional actors who reap the benefits of these changes but escape accountability for the social cost they have created. As a result, the costs are not only uncompensated; they easily go unrecognized as well and are not taken into account in the key decisions that determine the direction in which the economy evolves.

This was not true of institutional structures that emerged out of the first of the postwar policy paradigms, the Keynesian consensus. In the early postwar decades, businesses were not free to lay off workers when they introduced new technologies or developed new patterns of trade. The restrictions on their ability to do so varied from country to country, but for the most part layoffs required the consent of government or worker representatives or both, and typically compensation was required as well. Nor were companies free to adjust wages to attract better trained substitutes and thereby avoid providing training themselves or adjusting new technologies so that jobs were accessible to the existing labor force. The institutional structures that the policy paradigm sustained forced adjustment to take place within the productive sector, closely linked to the production process.

Such structures operated not only in Western Europe and Latin America, but also in the United States where, in the early decades of the postwar period, union seniority rules imposed restrictions that made layoff and discharge costly and adjustments in the wage structure virtually impossible. The threat of union organization imposed these restraints even on nonunion firms. Wage adjustments were further inhibited by federal government incomes policy: The policy involved statutory wage controls during the WWII and Korean War periods that left a legacy that persisted in the 1950s. After formal controls were lifted, in the 1960s, the wage-price guidelines were promulgated by President John F. Kennedy and enforced by, among other things, public shaming and regulatory and tax harassment, which had a similar effect. Statutory controls were reinstituted once again in 1971 by President Richard Nixon, and only finally eliminated at the end of the decade—at which point, probably not coincidently, the income shares at the top of the distribution began to diverge sharply and progressively from those that had prevailed in the earlier postwar decades.

Yet restrictions of this kind are not an ideal way to force companies to take into account the social costs of structural adjustment. Where individual firms compete with new companies that do not have any institutional obligations to the legacy of older forms of production, restrictions that force the enterprise to absorb the costs of structural adjustment jeopardize the efficiency and competiveness of the economy. Obviously this problem is more serious in an open, global economy than in the relatively closed economies in which the Keynesian paradigm was conceived. But as Andrew Schrank and I have argued elsewhere, the general systems of labor inspection in Southern Europe and Latin America have the administrative flexibility to adjust the regulations to accommodate competitive pressures of this kind. And a similar flexibility was introduced into the US system by the collective bargaining that generated the restrictions in the first place.

Thus, my point here is not to promote a revival of the Keynesian paradigm nor of the particular institutions to which it gave rise and sustained. It is to use the contrast between Keynesian and the prevailing policy paradigms to overcome the limits of the frameworks in which policy is currently conceived, and to widen the range of approaches with which we can respond to the political pressures created by existing policy. The institutions of the Keynesian period are not irrelevant here, but the world has changed and they cannot be uncritically re-created.

What is more relevant is the critical spirit that Keynes brought to the policy process at a time when other mainstream economists were reluctant to question the ruling economic paradigms of that earlier era. Such a spirit would be worthwhile at any time, but seems particularly important at the current moment, which in so many ways resembles the interwar period where public policy was caught by surprise, unprepared and ill-equipped to respond to the political reaction against globalization.

In that earlier period, economic policy was paralyzed by an intellectual impasse between market liberalism and Marxist historical materialism, diametrically opposed to each other, but each deterministic in a way that left little room for policy innovations to address the crisis. What is most unsettling today is the way in which this dichotomy is reproduced in the contrast between the Silicon Valley consensus and the Washington consensus. Indeed the former is technologically deterministic in the way that Marxism was—however different the technological trajectory that it thinks we are forced to accommodate—while the latter is basically a revival of the deterministic market liberalism of the interwar period. In this context, the great contribution of the Keynesian paradigm is the notion that there is room for action. And it is that aspect of Keynes that we need to recover today—a sense of option and opportunity, rather than passive acceptance.

We have subscribed to a kind of technological determinism, and ignored the role of the federal government through its financing of R&D that contributes to the upskilling of jobs, and then facilitates an institutional environment that dampens pressures to prepare the labor force to meet the jobs that public policy has promoted.

After all, as Keynes famously quipped, “In the long run we are all dead.” The point in this context is that the dominant policy paradigms, particularly that of globalization, have focused on the far horizon, where we reach a new, long-run equilibrium without recognizing the process through which we get there or the path that we follow in doing so. The paradigms do not allow for the possibility that there may be alternative adjustment trajectories or indeed that the end point might not be independent of the adjustment process. Nor do they recognize that the process may be more or less rapid, more or less spread out over time in ways that are critical to the welfare impact of change and, not incidentally given the present moment, to the political tolerance for the policies that promote it.

Economists and other analysts, not to mention policy-makers and business leaders, who have stood behind and rationalized the Silicon Valley and Washington consensuses have failed to recognize and take responsibility for the role that public policy has played in putting us where we now are. We have subscribed to a kind of technological determinism, and ignored the role of the federal government through its financing of R&D that contributes to the upskilling of jobs, and then facilitates an institutional environment that dampens pressures to prepare the labor force to meet the jobs that public policy has promoted. In reaction, democratic politics has taken matters into its own hands. As President Trump proposes an increasing array of tariffs aimed at redressing the dislocations of the past several decades, it may seem incredible that he is resorting to such an antiquated, blunt, and potentially counterproductive instrument. Yet those of us in the business of providing new policy options and opportunities have been asleep at the wheel. We need to wake up.

We Need New Rules for Self-Driving Cars

There is a sign on the campus of the National Transportation Safety Board (NTSB) in Virginia that reads “From tragedy we draw knowledge to improve the safety of us all.” The NTSB is obsessive about learning from mistakes. Its job is to work out the causes of accidents, mostly those that involve airplanes. The lessons it takes from each crash are one reason why air safety has improved so dramatically. 2017 was the first year in which, for the first time since the dawn of the jet age, not a single person was killed in a commercial airline passenger jet crash.

The NTSB also looks into other major transport malfunctions, particularly those involving new technologies. In June 2016, the board’s chair, Christopher Hart, was invited to speak to the National Press Club in Washington, DC, about self-driving cars. Hart—a lawyer, engineer, and pilot—gave a speech in which he said, “Rather than waiting for accidents to happen with driverless cars, the NTSB has already engaged with the industry and regulatory agencies to help inform how driverless cars can be safely introduced into America’s transportation system.” On finishing the speech, Hart was asked, “Is there a worry that the first fatal crash involving a self-driving car may bring the whole enterprise down?” He replied, “The first fatal crash will certainly get a lot of attention … There will be fatal crashes, that’s for sure.”

Hart didn’t know it at the time, but seven weeks earlier a first fatal crash had already happened. The same day that he gave his speech, the electric carmaker Tesla announced in a blog post that one of its customers had died while his car was using the company’s Autopilot software. Joshua Brown was killed instantly when his car hit a truck that was crossing his lane. The bottom half of Brown’s Tesla passed under the truck, shearing off the car’s roof. The only witness to speak to the NTSB said that the crash looked like “A white cloud, like just a big white explosion … and the car came out from under that trailer and it was bouncing … I didn’t even know … it was a Tesla until the highway patrol lady interviewed me two weeks later … She said it’s a Tesla and it has Autopilot, and I didn’t know they had that in those cars.”

The process of learning from this crash was as messy as the crash itself. The first investigation, by the Florida Highway Patrol, placed the blame squarely with the truck driver, who should not have had his truck across the road. Once it had become clear that the car was in Autopilot mode, a second investigation, this time by the National Highway Traffic Safety Administration, concluded that Brown was also at fault. Had he been looking, he would have seen the truck and been able to react.

It took the NTSB to appreciate the novelty and importance of what had transpired. The board’s initial report was a matter-of-fact reconstruction of events. At 4:40 p.m. on a clear, dry day, a large truck carrying blueberries crossed US Highway 27A in front of the Tesla, which failed to stop. The car hit the truck at 74 mph. The collision cut power to the car’s wheels and it then coasted off the road for 297 feet before hitting and breaking a pole, turning sideways and coming to a stop.

In May 2017, the NTSB released its full docket of reports on the crash. Tesla had by this point switched from coy to enthusiastically cooperative. A Tesla engineer who had formerly been a crash investigator joined the NTSB team to extract and make sense of the copious data collected by the car’s sensors.

The data revealed that Brown’s 40-minute journey consisted of two and a half minutes of conventional driving followed by 37 and a half minutes on Autopilot, during which his hands were off the steering wheel for 37 minutes. He touched the wheel eight times in response to warnings from the car. The longest time between touches was six minutes.

Despite a mass of information about what the car’s machinery did in the minutes before the crash, the car’s brain remained largely off-limits to investigators. At an NTSB meeting in September 2017, one staff member explained: “The data we obtained was sufficient to let us know the [detection of the truck] did not occur, but it was not sufficient to let us know why.”

Another NTSB slogan is that “Anybody’s accident is everybody’s accident.” It would be easy to treat the death of Joshua Brown as a mere aberration and to claim that, as his car was not a fully self-driving one, we can learn nothing of relevance regarding their performance. This would be convenient for some people, but it would be a mistake. Brown was one of many drivers lured into behaving, if only for a moment, as if their cars were self-driving. The hype surrounding the promise of self-driving cars demands our attention. As well as marveling at the new powers of machine learning to take over driving, we should look to the history of transport to build new rules for these new technologies.

Henry Bliss has the dubious honor of being the first person killed by a car in the United States. In the final weeks of the nineteenth century, an electric taxi hit Bliss as he was getting off a trolley car on the corner of 74th Street and Central Park West in New York City. The report posted on September 14, 1899, in the New York Times was brutally frank: “Bliss was knocked to the pavement, and two wheels of the cab passed over his head and body. His skull and chest were crushed … The place where the accident happened is known to the motormen on the trolley line as ‘Dangerous Stretch,’ on account of the many accidents which have occurred there during the past Summer.”

The driver was charged with manslaughter but later acquitted.

Over the twentieth century, as the internal combustion engine replaced the electric motor and car use exploded, the number of US car deaths grew, peaking in the 1970s, when the average was more than 50,000 per year. Individual tragedies stopped becoming newsworthy as the public grew used to risk as the price of freedom and mobility. Improvements in technology and political pressure from safety campaigners meant that even as the number of miles traveled kept climbing, road deaths declined until the end of the century. Only recently has this trend reversed. The years 2015 and 2016 both saw jumps in fatalities, almost certainly due to increased distraction from smartphones.

Proponents of self-driving cars claim that machines will be far better than humans at following the rules of the road. Humans get distracted; they get drunk; they get tired; they drive too fast. The global annual death toll from cars is more than a million people. At least 90% of crashes can be blamed on human error. If fallible human drivers could all be replaced by obedient computers, the public health benefit would surely be enormous.

However, technologies do not just follow rules. They also write new ones. In 1988, the sociologist Brian Wynne, looking back at recent calamities such as the Chernobyl nuclear disaster and the Challenger space shuttle crash, argued that the reality of technology was far messier than normally assumed by experts. Technology, for Wynne, was “a form of large-scale, real time experiment,” the implications of which could never be fully understood in advance. Technological societies could kid themselves that things were under control, but there would always be moments in which they would need to work things out as they were going along. Even exquisitely complex sociotechnical systems such as nuclear power stations were inherently unruly.

The United States’ early experience with automobiles is a cautionary tale of how, if society does not pay attention, technologies can emerge so that their flaws become apparent only in hindsight. The car did not just alter how we moved. It also reshaped our lives and our cities. Twentieth-century urban development took place at the behest of the internal combustion engine. Cities are still trying to disentangle themselves from this dependence.

It’s a story that the historian Peter Norton has narrated in detail. In the 1920s, as cars were becoming increasingly common, the automotive industry successfully claimed that the extraordinary social benefits of its creations justified the wholesale modernization of US cities. In the name of efficiency and safety, streets were reorganized in favor of cars. Led by the American Automobile Association, children learned the new rules of road safety in school. Ordinary citizens were recast as “pedestrians” or, if they broke the new rules, “jaywalkers.” By the 1930s, people were clear on how the privileges of access to streets were organized. The technology brought huge benefits from increased mobility, but also enormous risks. In addition to what the author J. G. Ballard called the “pandemic cataclysm” of road deaths, the nation’s enthusiasm for cars also made it harder to support alternative modes of transport. The conveniences of cars trumped other concerns and allowed for the reshaping of landscapes. Vast freeways and flyovers were built right into the hearts of cities, while the network of passenger railroads was allowed to wither. Around the cities’ edges, sprawl made possible by two-car families leaked outwards. By the 1950s, the United States—and much of the world—had been reshaped in the car’s image. The car and its new ways of life had created a new set of rules.

In the twentieth century, the unruliness of technology has become a brand. Silicon Valley sells “disruption,” a social media-friendly remix of what the economist Joseph Schumpeter called “creative destruction.” The idea, established by picking through the wreckage of once-powerful companies such as Kodak and Sears, is that incumbent companies will be brought down by upstarts bearing new technologies that change the rules. Among its banal slogans, the disruptive innovation movement proclaims that “doing the right thing is the wrong thing.” The disruptors are nimble and constantly experimenting. Disruptive innovation is intentionally reckless, looking for opportunities in challenging or evading regulatory rules. Facebook chief executive Mark Zuckerberg’s motto was, until recently, “Move fast and break things.” It’s a message that is easy to live with if you are benefiting from the changes and the stakes are low. But high stakes sometimes emerge unexpectedly. In the past couple of years, we have seen how software systems such as Facebook can challenge not just social interactions but also the functioning of democratic institutions.

For software, constant upgrades and occasional crashes are a fact of life; only rarely are they life-and-death. In the material world, when software is controlling two tons of metal at 70 mph, the collateral damage of disruption becomes more obvious. Software engineers encountering the challenge of driving appreciate the complexity of the task. It is infuriatingly unlike chess, a complicated game that takes human geniuses a lifetime to master but which computers now find rather easy. Driving, as self-driving car engineers regularly point out, doesn’t take a genius. Indeed, cars can be, and are, controlled by flawed and error-prone human brains. It is a process that, like identifying images and recognizing speech, has only recently become amenable to artificial intelligence. The approach to tasks of this complexity is not to try to work out how a human brain does what it does and mimic it, but rather to throw huge quantities of labeled data at a deep neural network (a layered software system) and let the computer work out patterns that, for example, tell a cat from a dog. In “deep learning,” as this approach is called, the aim of the game is to work out the rules. In some areas, the achievements have been remarkable. One of these systems—AlphaGo Zero, from Google DeepMind—allowed a computer to become the world’s best Go player in 40 days in 2017, working out strategy and tactics for itself using nothing but first principles and millions of practice games against itself. Deep learning can be extraordinarily powerful, but it is still learning.

A Tesla’s software is in a process of constant improvement. Fueled by data from millions of miles of other Teslas’ experiences in a process called “fleet learning,” the brains of these cars are being regularly upgraded. Tesla’s chief executive, Elon Musk, was so optimistic about the speed of this process that when he produced a new generation of the Tesla Model S in October 2016, he described it as having “full self-driving hardware.” All that would be required was for the car’s brain to catch up with its body, and for lawmakers to get out of the way, allowing hands-free driving on US roads.

The Tesla blog post that brought the May 2016 crash to light referred to the car’s software as being “in a public beta phase.” This was a reminder to Tesla owners that their cars were still not self-driving cars. The software that was driving their cars was not yet artificially intelligent. Its algorithms were not the driving equivalent of AlphaGo. They were, in the words of Toby Walsh, a leading researcher in artificial intelligence, “not very smart.” As the NTSB found, not only was the machine not smart enough to distinguish between a truck and the sky, it was also not smart enough to explain itself.

Elon Musk is relaxed about his car’s brain being a black box. In an email response to one business journalist’s critical investigation, Musk responded: “If anyone bothered to do the math (obviously, you did not) they would realize that of the over 1M auto deaths per year worldwide, approximately half a million people would have been saved if the Tesla autopilot was universally available. Please, take 5 mins and do the bloody math before you write an article that misleads the public.”

Similarly, when Consumer Reports called for a moratorium on Autopilot, Tesla replied, “While we appreciate well-meaning advice from any individual or group, we make our decisions on the basis of real-world data, not speculation by media.”

For Tesla, “doing the math” means that if self-driving cars end up safer on average than other cars, then citizens have no reason to worry. This relaxed view of disruption has a patina of rationality, but ignores Brian Wynne’s insight that technological change is always a real-time social experiment. Musk’s blithe arithmetic optimism fails to take account of a range of legitimate public concerns about technology.

First, the transition will not be smooth. It is not merely a matter of replacing a human driver with a computer. The futures on offer would involve changing the world as well as the car. The transformations will be unpredictable and intrinsically political.

Second, some risks are qualitatively different from others. When an airplane crashes, we don’t just shrug and say, “Oh well, they’re safer than cars.” We rely on accident investigators to dig out the flight data, work out what went wrong and why, and take steps to prevent it happening again. If passenger jets were anywhere near as deadly as passenger cars, there would be no commercial airline industry. Citizens are legitimately concerned about the risks of complex, centralized technological systems in which they must place their trust. (And not nearly concerned enough about the risks from cars, it has to be said.)

Third, a technology’s effects are not just related to the lives it takes or the lives it saves. Technologies distribute risks and benefits unevenly. They create winners and losers. Autopilot will never become universally available. All the current signs suggest that self-driving car technology is set to benefit the same people who have benefitted most from past technological change—people who are already well-off. Traffic, we might think, is the archetypical example of us all being in it together. But poor people typically spend more time commuting in traffic (and often in older, less fuel-efficient cars) than do the well-off who can afford to live closer to their workplaces. Well-designed transport systems can enable social as well as physical mobility to at least partly redress such inequities. Bad ones can be bad for commuters, bad for the environment, and especially bad for those who are already economically marginalized.

The efficiencies of algorithms should not be used as an excuse to disrupt time-honored principles of governance such as the need to explain decisions and hold individuals responsible. Concerns about algorithmic accountability have grown in volume as it has become clear that some advanced decision-making software has revealed implicit biases and, when questioned, its creators have been unable, unwilling, or both, to say where the biases came from. ProPublica’s investigation into the use of a risk-assessment algorithm in policing revealed an encoded bias against black people. The issue was not just the bias, but also the inscrutability of the algorithm and its owner, a company called Northpointe. With deep learning, these problems are magnified. When Google’s algorithms began mislabeling images, the company’s own engineers could not work out the source of the problem. One Google engineer took to Twitter to explain to an African-American man whose photo of himself and his friend had been tagged “gorillas” that “machine learning is hard.”

The inscrutability of machine learning, like technological inequality, is not inevitable. Controversies such as these are starting to convince machine learning companies that some form of transparency might be important. Toyota is currently working on an algorithmic transparency project called “the car can explain,” but such activities are only recently starting to move in from the fringes. Redressing the balance requires the engagement of governments and civil society as well as scientists. In some places, the lawyers have stolen a march on the innovators. The European General Data Protection Regulation, which comes into force in May 2018, demands what some observers have called a “right to explanation” from automated systems. In the 1990s, the European Union took a similar approach to the regulation of agricultural biotechnology, scrutinizing the processes of genetic modification. The difference of opinion with the United States, which looked only at the products of innovation—the seeds themselves and the traits they exhibited in plants—resulted in a high-profile dispute at the World Trade Organization.

Understanding processes of algorithmic decision-making is vital not just to govern them democratically, but to ensure that they will be able to deal with unfamiliar inputs. When deep learning systems work as designed, they may be exceptionally powerful. When they fail, we may not know why until it is too late. As well as the capabilities of artificial intelligence systems, we must consider their reliability and transparency.

If we expect too much of machine learning for self-driving cars, we will lose sight of everything else that is needed for well-functioning transport systems. The risk is that today’s algorithms become tomorrow’s rules of the road. When Sebastian Thrun ran Google’s self-driving car research, he argued that “the data can make better rules” for driving. As cars start to be tested, we can already see their handlers attempting to write their own rules. In the United States, the federal government has exempted thousands of vehicles from existing safety laws and made few demands in return. Behind the technological dazzle, there is little appreciation of the public cost of upgrading infrastructure to suit the needs of self-driving cars. We can get a sense of how the politics might play out. At the Los Angeles Auto Show in 2015, Volvo executive Lex Kerssemakers took the city’s mayor, Eric Garcetti, on a test drive in a prototype self-driving Volvo XC90. When the car lost its way, Kerssemakers said, “It can’t find the lane markings! … You need to paint the bloody roads here!” He deftly off-loaded responsibility for the failure of his technology onto the public sector. The comment was lighthearted, but the implications for infrastructure will be serious. Our roads have been designed with human perception in mind. When they get rebuilt, at substantial public cost, the pressure will be to do so in ways that suit self-driving cars, and thus benefit those who can afford them. If built without attention to winners and losers, smart infrastructure could easily end up further exacerbating economic and social inequities.

Self-driving cars will change the world. But that doesn’t mean much. The ways in which self-driving cars will change the world are profoundly uncertain. The range of possible sociotechnical futures is vast. All we can say with certainty is that the development of the technology will not be as flawless or as frictionless as the technology’s cheerleaders would imagine. A future in which all cars are computer controlled is relatively easy to imagine. The transitions required to get there could be horrendously complex. The current story of self-driving car innovation is that this complexity can be engineered into the system: machine learning will be able to handle any eventuality. Engineers talk about “edge cases,” in which unusual circumstances push a system outside its design parameters. For self-driving cars it is easy to imagine such situations: a confused driver going the wrong direction on a freeway, a landslide, a street entertainer, an attempted carjacking. Factoring such things into a system will require extensive training and testing. It would mean adding sensors, processing power, and cost to the car itself. The temptations to remove such complexities from the system—for example, by forcing pedestrians away from roads or giving self-driving cars their own lanes—could well prove irresistible. The segregation of different forms of traffic may be efficient, but it is controversial. In Europe, for example, the politics of streets are played out in cities every day. City planners everywhere should not let technologies force their hand.

If handled with care, self-driving cars could save thousands of lives, improve disabled people’s access to transport, and dramatically improve lifestyles and landscapes. If they are developed and governed thoughtlessly, the technology could lead to increases in sprawl, congestion, and the withering of mass transit. At the moment, the story is being led by the United States. It is a story that prioritizes freedom—not just citizens’ freedom to move, but also companies’ freedom from regulation. The story of the ideal “autonomous vehicle” is not just about the capabilities of the robot. It is also about unfettered innovation. It is a story in which new technologies come to the rescue, solving a problem of safety that policy-makers have for decades been unwilling to prioritize. This story will, if allowed to continue, exacerbate many of the inequalities created by our dependence on conventional cars. If we are to realize the potential for self-driving car technology, this story needs to change.

The race for self-driving car innovation is currently causing a privatization of learning. The focus is on proprietary artificial intelligence, fueled by proprietary data. The competition this creates leads to fast innovation, but speed can be bad if pointed in the wrong direction or if there are unseen dangers in the road ahead. If we want innovation that benefits citizens rather than just carmakers or machine-learning companies, we urgently need to recognize that the governance of self-driving cars is a problem of democratic decision-making, not technological determinism. Alongside machine learning, we must create mechanisms for social learning.

The first target should be data-sharing. The Tesla crash revealed an immediate need. In this case, Tesla cooperated with the NTSB to extract the data required to work out what went wrong. We should not have to rely on companies’ goodwill. US regulators have for years tried without success to mandate event data recorders in cars that would, like airplane “black boxes,” provide public data to narrate the last moments before a crash. The arrival of automated decision-making in driving makes this more urgent. Marina Jirotka, a social scientist, and Alan Winfield, a roboticist, recently argued that we need to enforce data sharing in robot systems so that people beyond just roboticists can learn from accidents. The challenge here is to not just relax companies’ grip on data, but also to improve the accountability of artificial intelligence.

In September 2016, the National Highway Traffic Safety Administration, by the request of the Obama administration, issued a call for data-sharing, which it justified using the language of “group learning.” The regulator also suggested that companies should collect and analyze data on “near misses and edge cases,” join an “early warning reporting program,” and find ways for their cars to communicate with one another. The NTSB concluded its investigation into the Joshua Brown wreck with a similar recommendation: “We don’t think each manufacturer of these vehicles need to learn the same lessons independently. We think by sharing that data, better learning and less errors along the way will happen.” Data-sharing is not just important when machines go wrong. If self-driving pioneers are prioritizing machine learning, then we should ask why they can’t learn from one another as well as from their own data sources.

Regulators are right to challenge the story of heroic independence that comes from such a heavy emphasis on artificial intelligence. Getting innovators to work together makes more urgent an inclusive debate on standards. When pushed, self-driving car engineers admit that for the things to work, they cannot be completely autonomous robots. They must be digitally connected to one another and with their surroundings. We must start to work out the real costs of doing this. Smart cars would require smart infrastructure, which would be expensive. It would also mean that the benefits of self-driving cars will be felt by some people far earlier than others. There is fevered discussion of when self-driving cars will be with us. The question is not when, but where and for whom. Cars will be “geofenced”—prevented from operating outside particular places and particular conditions. The dream of complete automotive autonomy and freedom will likely remain a dream.

Connectivity gets less attention than autonomy, but its challenges are just as great. Connected cars bring new risks of cybersecurity, data breaches, and system failure. Ensuring effective, safe transport across entire cities and countries demands early standards-setting. This process would be an ongoing conversation rather than a once-and-for-all. Technologies for self-driving are fluid, and the future of transport is profoundly unpredictable. Governance must therefore adapt. However, the first step, which requires real political leadership, is for governments to reassert their role in shaping the future of transport. Two philosophers at Carnegie Mellon University who study artificial intelligence ethics, David Danks and Alex John London, recommend a regulatory mechanism analogous to the Food and Drug Administration. New technologies would be systematically tested before release and continually monitored once they are out in the wild. In deciding whether self-driving cars were safe, it would also be necessary to ask, Safe enough for what? Addressing such questions would in turn require democratic discussion of the purposes and benefits of the technology.

Governments in the United States and elsewhere have held back from proactive regulation of self-driving cars. Demanding regulatory approval before these technologies hit the market would be a big shift. It would also force governments to rediscover skills that have been allowed to atrophy, such as those of technology assessment. If these technologies are as new and exciting as their proponents say they are, then we should ask what new rules are needed to ensure that they are safe, broadly accessible, and free from problematic unintended consequences. If the public does not have confidence in the future benefits of self-driving cars, the next Autopilot crash may cause far more damage and controversy, jeopardizing the future of the technology.

Indeed, as this article goes to press, the next accident has occurred in Tempe, Arizona: a crash of a self-driving Uber test vehicle, which resulted in a pedestrian death. Details are unclear. The NTSB has begun an investigation. Governance by accident continues.

From the Hill – Winter 2018

The House of Representatives and the Senate spent most of 2017 turning the annual budget process on its head, with the result that at the end of December there is no official budget for fiscal year (FY) 2018, which began October 1.

The annual budget resolution is intended to be the mechanism by which Congress establishes the overall spending framework within which the appropriations committees are supposed to operate when distributing funds to specific agencies and programs. The budget resolution is supposed to be completed by April 15, before the appropriations process really gets going. But this year, for a variety of reasons, the budget committees that draft these resolutions punted the work until much later. The House Budget Committee didn’t approve its budget plan until July, and it wasn’t approved by the full House until October. The Senate approved its significantly different budget resolution in October. Unable to reach final agreement, Congress was forced to approve a series of continuing resolutions that were necessary to keep the government operating in the absence of an approved budget. The most recent continuing resolution was passed December 22 and lasts until January 19, 2018.

The lack of a final budget resolution didn’t stop appropriators from trying to do their work. In fact, the House completed its annual appropriations, approving all 12 bills on schedule before the end of September. The full Senate failed to pass a single spending bill, but the Senate Appropriations Committee did produce its spending recommendations by early December. At this stage, the House appropriations differ significantly from the administration’s requests, and the figures from the Senate Appropriations Committee differ from both the administration and the House levels. But until there is a budget resolution that establishes overall spending levels, neither the House nor the Senate numbers can be considered final, so the two chambers cannot begin reconciling their differences.

The reality is that neither plan does much to resolve the major issue of the day for science and technology (S&T) funding where the current spending caps will end up. Under current law, the statutory caps on both defense and nondefense spending, which have been in place since 2011, are slated to decline by about one-half a percentage point below FY 2017 levels. The Senate budget resolution simply adopts the current caps, but Senate appropriators chose to ignore the scheduled decrease for FY 2018 and wrote their spending bills to the FY 2017 limit instead. The Senate budget numbers are functionally a placeholder, pending a deal to raise the spending caps later. If Congress does not vote to raise the budget ceiling, the Senate spending numbers would be subject to across-the-board reductions under the sequestration rules.

On the House side, appropriators remained within the limits for nondefense spending established in the budget resolution, but they approved defense spending that far exceeds the cap. The defense spending cap is unlikely to be raised because that would require 60 votes in the Senate, and Democrats are unlikely to go along with that. Thus, the House defense appropriations would be reduced according to sequestration rules.

The budget debate will pick up again in mid-January, when the continuing resolution expires, and Congress will again face the necessity of arriving at a budget deal. In the meantime, it is worth reviewing the science funding recommendations that emerged from the House and Senate appropriators.

Defense. The proposed Senate bill includes spending levels for S&T that are roughly equal to the FY 2017 budget and that are higher than what was recommended by the White House or approved by the House. Although the defense subcommittee of Senate Appropriations has been a supporter of Department of Defense (DOD) basic science in the past, this year’s legislation would cut basic research programs by $17 million or 0.8%. About half of this reduction would come at the expense of university partnerships. Naval basic science is actually increased by 5.8% across an array of fields, but this is offset by Army and Air Force reductions.

It’s a somewhat mixed bag for the Defense Advanced Research Projects Agency. Although the agency received a 4.9% increase, with space and electronic technology research boosted, the increase is smaller than that provided by either the House or requested by the administration, with constrained or reduced funding for materials, biotech, aerospace, and sensor technology.

Elsewhere, manufacturing research and development (R&D) is one of the brighter spots in the Senate bill, with several manufacturing S&T programs granted extra funding. An extra $25 million was added to the National Defense Education Program for manufacturing-oriented grants. Senate appropriators also increased DOD’s Defense Innovation Unit-Experimental by $5 million above the request and added $25.5 million for the new National Security Technology Accelerator, a public-private-academic consortium. Peer-reviewed medical research via the Defense Health Program received several hundred million dollars more than the request, though less than last year.

The Senate defense bill would exceed the current spending cap by $70 billion. As Democrats pointed out, that would mean DOD spending would be automatically ratcheted down substantially via sequestration unless Congress decides to roll back the spending caps approved in 2011.

Interior. The Senate Appropriation subcommittee’s draft bill would diminish the Environmental Protection Agency’s (EPA) environmental and climate research activities, while keeping US Geological Survey (USGS) programs funded at last year’s levels. Within EPA’s discretionary budget, the S&T account would drop by 11.2% below FY 2017 levels; most core S&T research programs would be subject to reductions in the order of 10% to 12%. The House approved slightly larger cuts, and the administration had requested even steeper reductions. Sen. Tom Udall of New Mexico, the ranking Democrat on the subcommittee, said that if the Senate bill had received a markup, he would have offered an amendment including $200 million to restore proposed cuts to EPA’s core research and regulatory programs.

Of note, a provision in the Senate’s bill continues to prohibit EPA from using funds to implement a mandatory greenhouse gas reporting system within the agricultural sector. A separate provision would change federal policy to treat forest biomass activities as non-contributors of carbon dioxide, a concept known as “carbon neutrality.”

The USGS would see overall flat funding in the Senate bill, compared with a 4.2% reduction recommended in the House and a 15% cut requested by the administration. USGS climate R&D and Climate Science Centers would be flat-funded rather than reduced as the House and administration wanted. Meanwhile, Senate appropriators joined their House counterparts in rejecting the administration’s proposed elimination of the agency’s earthquake early warning system. Landsat 9 development is fully funded in the Senate bill.

Homeland Security. The Senate’s draft appropriations bill would cut the Department of Homeland Security’s (DHS) research activities and laboratory facilities funding. Still, overall R&D funding would remain significantly higher than either the House version or the administration’s request. S&T laboratory facilities would be subject to a 23.8% cut, whereas the House had recommended flat funding. University programs would also see a moderate reduction under the Senate bill.

For the Domestic Nuclear Detection Office’s R&D account, the Senate prescribes a 7.8% cut, in line with House and administration budget proposals. The bulk of this reduction would come from detection-capability development and assessment programs. The Transformational R&D account, which supports an array of R&D activities to support detection as well as university and Small Business Investment Research activities, would see a small cut. Nuclear forensics would also be trimmed.

Also included in the Senate’s bill is a provision that would provide $1.6 billion for President Trump’s proposed border wall with Mexico. Senate Appropriations Committee Vice Chairman Patrick Leahy (D-VT) ridiculed the border wall proposal as “bumper sticker budgeting” and said he would have offered an amendment that would have blocked funding for the wall unless it was paid for by Mexico, as Trump promised during his election campaign.

National Science Foundation. The total NSF budget would decrease by $161 million or 2.2% below FY 2017 under the Senate legislation. Senate appropriators prioritized $105 million for construction of three Regional Class Research Vessels (RCRVs), continuing recent Senate efforts to shore up funding for the RCRV project, whereas the House offered no funding for RCRVs in its bill. Conversely, the Senate bill would cut NSF’s primary research account by 1.9% and NSF’s Education Directorate by 2% below last year.

National Aeronautics and Space Administration. Within NASA, Senate appropriators would impose a $234 million or 12.7% cut to Planetary Science, while keeping Earth Science funded at last year’s levels—the opposite of the approach taken by the House. The Senate’s move would allow the agency to continue funding for several Earth Science missions slated for elimination in the administration’s request, including the Orbiting Carbon Observatory-3 (OCO-3); the Plankton, Aerosols, Clouds, ocean Ecosystem (PACE) mission, and the Climate Absolute Radiance and Refractivity Observatory (CLARREO) Pathfinder mission.

On the exploration front, the Space Launch System and the Orion Crew Vehicle would be funded at FY 2017 levels, the same as in the House bill, as opposed to the cuts requested by the administration. The Space Technology directorate would see a 2% increase, including $130 million to preserve the RESTORE-L (a robotic spacecraft equipped with the tools, technologies, and techniques needed to extend the lifespans of satellites) from its requested elimination; the House provided only $45.3 million for RESTORE-L. NASA’s Office of Education, which was slated for termination in the administration’s request, was also protected in the Senate bill, with flat funding recommended for FY 2018.

National Institute of Standards and Technology. At the Commerce Department, NIST’s laboratories would see a small 0.7% uptick instead of the cut recommended by the administration. Notably, Senate appropriators rejected the administration’s proposed elimination of the Hollings Manufacturing Extension Partnership (MEP); the House proposed a $30 million cut to MEP. Funding for NIST’s coordinating role within Manufacturing USA, the interagency network of public-private manufacturing innovation institutes, would be equal to the president’s request of $15 million in FY 2018, a $10 million decrease from last year.

National Oceanic and Atmospheric Administration. Elsewhere in the Commerce Department, the Senate would maintain NOAA’s Office of Oceanic and Atmospheric Research at last year’s level, avoiding the 19% cut recommended by the House and the administration. Senate appropriators did side with the House in rejecting the administration’s proposed elimination of the National Sea Grant College Program. The Senate also sidelined the administration’s proposal to terminate federal funding for the National Estuarine Research Reserves System, as did the House. NOAA’s flagship weather satellites, the Joint Polar Satellite System and the Geostationary Operational Environmental Satellite system, would receive full funding for FY 2018. Notably, the Polar Follow On mission, which was targeted for big cuts by the House and administration, would instead gain $90.1 million above its FY 2017 amount of $328.9 million.

Transportation. The Senate committee took a similarly mixed approach for Department of Transportation (DOT) technology programs The committee granted a moderate funding increase above FY 2017 levels for the Federal Aviation Administration’s (FAA) NextGen initiative to modernize the nation’s air traffic control system. NextGen would receive $1.1 billion, whereas the administration had requested only $988 million in its FY 2018 budget proposal. Senate appropriators also granted a small increase for FAA’s research, development, and engineering program, mostly for advanced materials research for commercial aviation.

Grad students heave sigh of relief on tuition waivers

The final version of the tax reform bill passed in late December did not include a provision that was in the House version of the bill that would have required graduate students to pay taxes on certain tuition allowances, a change that would have significantly increased the tax burden for some students.

House considers five science bills on “Science Day”

Rep. Lamar Smith (R-TX) named Monday, December 18, “Science Day” in Congress to acknowledge the five bipartisan bills from the Science, Space, and Technology Committee that were considered on the House floor that afternoon. The bills included H.R. 4375, the STEM Research and Education Effectiveness and Transparency Act; H.R. 4323, the Supporting Veterans in STEM Careers Act; H.R. 4254, the Women in Aerospace Education Act; H.R. 1159, the United States and Israel Space Cooperation Act; and H.R. 4661, the United States Fire Administration, AFG, and SAFER Program Reauthorization Act of 2017.

Senate committee revisits law that may hinder battle against opioids

On December 12, the Senate Judiciary Committee met to discuss whether the 2016 Ensuring Patient Access and Effective Drug Enforcement Act should be repealed or amended. The House and Senate had passed the bill unanimously, but critics charge that it strips the Drug Enforcement Administration (DEA) of its most effective tool in combating the flow of opioids. Demetra Ashley, the acting assistant administrator of DEA, testified that under the law, DEA investigators face a greater challenge in proving that a company’s conduct poses an immediate danger of death or harm, making it more difficult to shut down shipments of painkillers from a distributor to a pharmacy for these cases. She noted that Congress should choose between repealing and amending the law. Senators split largely along party lines in opinions on what changes Congress should make to the bill.

Reprieve for ARPA-E

In December, the Government Accountability Office (GAO), an independent federal agency, officially found that the Department of Energy violated federal law when it withheld previously approved funding for the Advanced Research Projects Agency-Energy (ARPA-E) earlier in 2017. According to the GAO, the move violated the Impoundment Control Act, a federal law that requires the executive branch to obligate funding appropriated by Congress. The Trump administration proposed eliminating ARPA-E in its FY 2018 budget request and also began to freeze funding before Congress had given its consent. The funding in question has since been released.

Assessing NASA progress

On December 15, the National Academies of Sciences, Engineering, and Medicine released a report assessing the progress that NASA has made on scientific priorities outlined in the Academies’ 2011 decadal survey. A Midterm Assessment of Implementation of the Decadal Survey on Life and Physical Sciences Research at NASA concludes that NASA should “raise the priority of scientific research that addresses the risks and unknowns of human space exploration” and stresses that it should develop a US strategy for the International Space Station beyond 2024.

Improving undergraduate STEM education

On December 12, the National Academies released Indicators for Monitoring Undergraduate STEM Education, which offers a framework to track progress toward assessing the quality and impact of undergraduate education in science, technology, engineering, and mathematics. The report identifies three overarching goals to improve undergraduate STEM education: increase students’ mastery of STEM concepts and skills; strive for equity, diversity, and inclusion of STEM students and instructors; and ensure adequate numbers of STEM professionals.

Talk to the Hand

Friends, colleagues, and relatives from across the globe all seem to have the same questions: What is it like to be in Washington, DC, in the age of policy mayhem? What role is there for policy wonkery and expert advice when evidence, facts, intellectual consistency, and honesty are decomposing on the compost heap?

Grizzled veterans of many presidential administrations might answer that there never was any golden age when policy-makers consistently acted on the basis of strong evidence and careful reasoning, when legislators spoke the truth and were free of the influence of wealthy donors, when public policy was aligned with the public good rather than the narrow goals of special interests. But even the most jaded old hands among Republicans and Democrats alike concede that these are special days—yes, these are the most special days in the entire history of the country as we are regularly reminded in tweet storms.

Curious to gauge the attitude of Washington insiders, I devoted a few days in mid-December to attending a variety of events at key DC organizations. My overall impression is that most Beltway denizens have flipped DC to CD—cognitive dissonance. They continue to do research, conduct analyses, and provide expert guidance as if the nation’s leaders would give them thoughtful consideration.

My first visit was to the American Enterprise Institute, a respected bastion of the conservative establishment, for a discussion of agricultural research. Vincent Smith of the institute and Philip Pardey of the University of Minnesota summarized the findings of their white paper, which warned that by reducing the share of the US Department of Agriculture’s budget devoted to research, the federal government was weakening the competitive position of US farmers at a time when other countries are boosting agricultural research. Smith and Pardey encouraged federal investment not only in basic science but also in applied research of more immediate use to farmers, which is a departure from the standard conservative position that federal investment should be limited to basic science.

My next stop was at the Computing Research Association for an all-day meeting—organized with support from Rice University, Google, Microsoft Research, and several trade and professional organizations—on technology and jobs, with particular emphasis on the role of robotics and artificial intelligence. Andrew McAfee of the Massachusetts Institute of Technology reviewed the history of US wage growth and stagnation, David Blustein of Boston College provided a ground-level view of the psychological and sociological stress created by the rapid evolution of the labor market, Edward Felton of Princeton University described the analysis he contributed when working at the White House Office of Science and Technology Policy in the Obama administration, Philip Longman of the Open Markets Institute argued that the primary threat to workers is not technological advances but the concentration of power in a few giant companies, and the day ended with a lively panel discussion in which experts from a range of ideological perspectives addressed policy challenges.

Then it was off to a meeting at the Brookings Institution, the well-known left-of-center think tank. Although the primary purpose of the meeting was to examine whether the patent system was operating effectively to nurture scientific and technological innovation, the day began with a wide-ranging discussion between Wall Street Journal reporter Greg Ip and Kevin Hassett, the recently appointed chair of the Council of Economic Advisers. Hassett is a prominent spokesperson for the administration’s contention that its tax reform legislation will create many new jobs and raise wages. This is a hard sell at Brookings, which is a cosponsor of the Tax Policy Center, whose analysts project that the tax changes will have a relatively weak effect on economic growth.  The patent discussion focused on the need to decrease the number of invalid patents that are being approved. Although there has been a general feeling that this is a problem, Michael Frakes of Duke University and Melissa Wasserman of the University of Texas provided empirical evidence that identified practices in the US Patent and Trademark Office that too strongly favored the granting of patents. The discussion was a refreshing instance when data and analysis were more important than ideological posturing.

Indeed, what was striking about this sample of Washington wonkery was how much it differed from the take-no-prisoners, accept-no-compromise, make-your-own-facts brouhaha that is presented on cable news channels. Although there is no doubt that political discourse has become coarser and more polarized, it is a mistake to see Washington as the source of the problem. It’s just the stage. The voters choose the firebrand legislators who come to DC to drain the swamp, but professional Washington is more drawing room than bog. The majority of people who work in government and the think tanks are far more pragmatic than ideological. Sure, there are screaming single-issue lobbyists and monomaniacal ideologues who make for good TV drama, but most of the discussions that take place off camera are boringly civil, well-informed, and respectful.

Of course, that does not mean that greed, moral righteousness, intolerance, fear, and lust for power are not widespread in Washington. These forces are prevalent throughout the nation—indeed, throughout humanity—and Washington provides a forum where they can be displayed. The tribe of experts who work in and around government are not unaware of this maelstrom of conflict—Toto, we’re not in Sweden anymore—and understand that it cannot be ignored in crafting practical policy prescriptions. But when the wild-eyed combatants eventually recognize that the failure to do anything in a world that is rapidly changing is a formula for failure, it will be helpful to have something practical at the ready to provide a catalyst for compromise.

I suppose this can be laughed away as pointy-headed idealism, but if the experts do their work honestly and rigorously, in the end reality will side with them because policies based on fantasy will fail. The economy will not grow on false facts, public health will not improve with false beliefs, public confidence will not improve with unfulfilled promises, and global peace will not be achieved by ignoring the interests of other countries.

But what about the core issues of science and technology? Who is providing expert guidance to the White House and cabinet secretaries? In many cases, no one. In advance of each new presidential administration the National Academies prepares a list of the key political positions to be filled by people with expertise in science, technology, and health, and in previous administrations the Academies have played an informal role in suggesting qualified candidates. A central goal of this effort was to see these positions filled quickly so that appropriate expertise would be available from the outset of the administration’s planning and policy development. It was cause for celebration when President Obama nominated John Holdren to be his science adviser a month before taking office.

As the current administration completes its first year, no one has been nominated to be science adviser or to fill any of the other key positions in the White House Office of Science and Technology Policy. No one has been proposed to serve on the President’s Council of Advisors on Science and Technology. Other positions waiting to be filled include director of the Bureau of the Census, commissioner of the Bureau of Labor Statistics, assistant secretary of state for oceans and international environment and scientific affairs, under secretary for health at the Department of Veterans Affairs, assistant administrator for research and development at the Environmental Protection Agency, director of the US Geological Survey, assistant secretary for fish and wildlife and parks at the Department of the Interior, and director of the Education Department’s Institute of Education Sciences. At the Department of Energy, there are no nominees for assistant secretary for energy efficiency and renewable energy, assistant secretary for environmental management, assistant secretary for nuclear energy, under secretary for nuclear security, director of the Office of Science, and director of the Advanced Research Projects Agency-Energy.

But is this evidence of malign neglect necessarily a bad thing? When viewing the actions of the people who have been appointed to key positions in the Departments of Energy or Interior or at the Environmental Protection Agency, perhaps the Robert Browning phrase “less is more,” which Mies van der Rohe applied to modern architecture, might be an apt slogan for the next few years in policy. We might be grateful to let the practical, faceless, slow-moving career federal deckhands keep the ship of state more or less on a steady course. And the nongovernment policy wonks looking for a receptive ear could take the advice I used to hear from my smartass teenage kids: Talk to the hand, the head’s not listening.

What Did I Just Buy?

We all do it. Click on that little button—the one that says “buy”—to purchase digital books, music, movies, and software online. What many of us don’t realize is that US courts have been concluding for some time that the simple act of clicking that button means you’ve entered into a contract, you have “manifested assent” under the jargon of contract law, which is a necessary step in forming a binding contract. That contract you’ve just agreed to purports to govern what it is you’ve actually bought. In other words, forget your assumptions about ownership: read the contract if you want to know what you’re getting when you click “buy.” And, frankly, who does that?

We buy digital goods with these contracts all the time. Increasingly we also are buying physical goods that have digital components embedded in them (think: the internet of things), and these purchases also have contracts associated with them, along with technological capabilities that make ownership of these things quite different from traditional concepts of ownership. With the expansion of broadband, we are also streaming content over the internet without ever owning a copy of the work that is streamed. The nature of our relationship to the products we buy is undergoing a tectonic shift that few realize is occurring. That shift carries with it a host of unsettling consequences.

Aaron Perzanowski and Jason Schultz’s new work, The End of Ownership, masterfully explores this new reality. That thing we think we just bought: often we don’t own it, but are merely licensing it under the terms set out in the terms of use. We are no longer in the old world, in which the limits on what we can do with the things that we own were governed by laws, decided on collectively by individuals elected by the population at large. Instead, we have entered a world in which the limits on what we can do with the things we buy are defined by a contract drafted by those who are selling us access and use rights—but not ownership. Without ownership rights, our ability to resell purchased objects, or even just lend them, is being eliminated. Add in the technological protection measures embedded in products, and our lack of control over the objects we buy doesn’t resemble ownership at all.

After explaining the differences between the four types of property (real, personal, intellectual, and intangible), Perzanowski and Schultz recount the transition from the analog era to the digital era, highlighting the benefits to consumers that made the transition appealing. At the same time, the authors clearly explain the legal doctrines that have allowed producers to use their copyrights to define in the purchase agreement the rights obtained when someone “buys” something. And it’s not just that we don’t bother to read these agreements, it’s that the agreements are hard to read, filled with legalize, and long. After all, what iTunes users in their right mind would read a 19,000-word agreement before purchasing a 99-cent song? We just keep buying and manifesting our assent to these terms without really understanding what we are buying.

Perzanowski and Schultz offer a sensible test that courts should use to determine when a purchase amounts to a sale, rather than a license. They suggest courts examine three considerations: “(1) the duration of the consumer possession or access; (2) the payment structure of the transaction; and (3) the characterization of the transaction communicated to the public.” If the court determines that a sale has occurred, any attempt by the seller at shaping the permissions granted to the consumer through a contract would be ignored. Instead, purchasers would own the items that they purchased with the permissions generally associated with ownership and defined by law.

The authors spend a chapter exploring the third consideration: the characterization of the transaction. Recounting their previous studies of how to clearly communicate what is really being sold, insisting on clear notices would certainly be an improvement over the current state of the law that allows a “buy now” button to really mean “license now.” This informative exploration of the role clear notices play is instructive and demonstrates just how easy it would be for sellers to do a much better job of clearly communicating what it is they are offering for sale.

The authors also interestingly trace how streaming content is another step down the path to a world without ownership. As they experience explosive growth, streaming services such as Netflix and Spotify have tapped into consumer demand that values access over ownership. A major benefit of the access model is much-reduced prices for consumers, and Perzanowski and Schultz acknowledge the positive distributional effects of reduced prices. They point out, however, that if everything is streamed and no one owns copies anymore, there is a risk that the physical and legal infrastructure necessary to support ownership will disappear. If that happens, works can disappear as well.

For example, if a publisher decides that a work has become too controversial and therefore a liability, the publisher can just remove it from the database of available content. Effectively the work would no longer exist: there would be no untethered, freestanding copies in the world preserving what once was. Though libraries might be thought of as a bulwark against such erasure of collective memory, increasingly libraries are buying into subscription-based purchasing models as well. The authors’ chapter exploring “The Promise and Perils of Digital Libraries” is both intriguing and scary.

The End of Ownership uses the tools of law and economics to highlight transaction cost problems, such as reading the 19,000-word license for a 99-cent sale, and the negative externalities that have yet to be internalized through regulation. But it also uses insights gained from behavioral economics to emphasize the genuine lack of understanding that consumers have of what it is they are actually buying. Perzanowski and Schultz persuasively articulate the consequences people may face in a world without ownership. The book is most valuable in pulling together the related strands of legal doctrines to sound the alarm so that more consumers and regulators will pay attention to what is happening in today’s purchasing transactions.

As the authors explore, it is not just the legal details of whether we own or merely license the things that we pay for that are causing this enormous shift in our relationship with those things. The other driving factor in the end of ownership involves the technologies of control embedded in everything from ebooks to coffeemakers, from printer cartridges to tractors. Yep, that’s right: tractors. Digital rights management (DRM) permits manufacturers to physically constrain what uses are allowed with the products you purchase, and federal law makes it illegal to manipulate the DRM. Though there is a way to obtain three-year exemptions from that federal law, it involves a lengthy process overseen by the Librarian of Congress. For example, in 2015 farmers sought a DRM exemption to be able to repair, modify, and upgrade their tractors without violating federal law. The exemption was granted, but its implementation was delayed for 12 months, effectively consuming one of the three years of the period of exemption. In 2018, farmers will again need to make their case that an exemption is needed. As Perzanowski and Schultz point out, these exemptions are hardly a permanent response to the problems that DRM technologies cause.

Perzanowski and Schultz conclude their book with suggestions for reform that could safeguard ownership. They are not suggesting a ban on streaming services or a ban on licensing models. Far from it. Instead, they suggest some sensible reforms to intellectual property laws that would ensure the survival of ownership and the benefits that go along with it. Implementing the reforms necessary to secure the legal infrastructure that facilitates real choice in the marketplace, including the choice to own what you buy, is an important goal. Legal reforms alone, however, are unlikely to be sufficient; changes in technology are also necessary. Perzanowski and Schultz explore the possibility that blockchain technology (a distributed digital ledger that is a core component of cryptocurrencies such as bitcoin) could be employed as a reliable ownership recording system for transferring copies of owned digital works. Though their technological optimism in pointing to blockchain technology may seem far-fetched to some people, this approach is provided to demonstrate that it may be possible to create a technological infrastructure that can support an ownership model in digital goods.

The book is filled with illuminating and memorable stories and examples. Beginning with the infamous (and highly ironic) time that Amazon remotely deleted Kindle copies of George Orwell’s dystopian classic Nineteen Eighty-Four, Perzanowski and Schultz go on to recount less well known cases such as the Hello Barbie that engages in conversations with children. Unbeknownst to most parents, the doll’s end user license agreement grants wide latitude to the managers of a cloud-based speech recognition platform to use information about the children’s conversations. You get the real sense that the proverbial tip of the iceberg has been spotted; the iceberg itself is on a collision course with our core understandings of ownership.

In the opening chapter the authors assert that “The most fundamental value at stake in the choice between ownership and licensing is autonomy.” Ultimately what Perzanowski and Schultz have done is provide a tour de force account of the current state of our relationships with the products we buy and how laws evolved to allow these relationships to be shaped this way. The authors have also teed up the conversation that we all must have about how these new relationships affect not only simple economic choices, but the ability of individuals to access and share information and knowledge and, ultimately, to facilitate human flourishing.

Is the Bayh-Dole Act Stifling Biomedical Innovation?

The productivity of biomedical research has fallen dramatically during the past three decades. For example, Nicholas Bloom and colleagues, in a 2017 National Bureau of Economic Research working paper, found that productivity in cancer research (measured as the ratio between years of life saved per one hundred thousand people and the number of publications reporting on clinical trials for cancer therapies) fell at an annual rate of 5.1% from 1975 to 2006 for all cancers. For breast cancer the annual decline was 10.1%. The researchers also looked at heart disease, and they found that between 1968 and 2011 research productivity fell at an annual rate of 7.2%. Another measure of productivity, the ratio of new medical products approved by the Food and Drug Administration to the number of researchers, fell at 3.5% per year from 1970 through 2014. Using yet another measure, Jack Scannell and colleagues, writing in Nature in 2012, documented an 80-fold decrease in the number of new drugs approved per dollar over the period 1950 to 2010.

Can this downward trend be reversed?

Public funding in the United States for biomedical science follows two paths. The direct path is typically through the National Institutes of Health’s (NIH) intramural research efforts and its extramural support of university research. The indirect path is through the purchases of pharmaceutical products and services by Medicare and Medicaid, the Veterans Administration, and the Affordable Care Act. Portions of the revenue generated from those sales are then invested in industrial pharmaceutical research and development (R&D).

One policy approach to improving productivity along these two pathways has been to encourage university-industry collaborations in the belief that the basic research focus of universities and the applied focus of industrial R&D are complementary. For example, the Bayh-Dole Act of 1980 required universities to pursue patents based on commercializable discoveries made by their scientists. Other policy interventions aimed, at least in part, at increasing collaboration between industry and universities include the National Cooperative Research Act (NCRA) of 1984; the National Cooperative Research and Production Act (NCRPA) of 1993; and the Advanced Technology Program (ATP), created through the Omnibus Trade and Competitiveness Act of 1988 (though terminated in 2007). Approximately 15% of research joint ventures created in response to the NCRA and the NCRPA had a university as a research partner, and about one-half of ATP-funded joint ventures had a university as a research partner.

More recently, NIH pronouncements on the importance of translational medicine, and its formation in 2012 of the National Center for Advancing Translational Sciences, suggest that the agency recognizes the need for a convergence of universities and industry research domains. Indeed, industrial and NIH R&D priority areas for new therapies are generally the same, focusing on oncology and immunomodulators, nervous system disorders, and anti-infectives. NIH is also trying to fund the development of new drugs that target common diseases that are also being addressed by pharmaceutical companies. NIH’s Active Requests for Applications include an emphasis on diabetes, Alzheimer’s disease, cancer detection, and stem cell products.

Pursuing stronger links between universities and firms conducting biomedical research may seem like a sensible way to reverse declining productivity. But empirical studies we have conducted on university involvement in private-sector industrial research, especially company research funded by NIH through its Small Business Innovation Research (SBIR) program, tell a different story. Our research finds that the R&D productivity of companies is thwarted by university partnerships. Could university involvement in industrial R&D actually be one cause of the productivity decline in biomedical innovation, actually stifling research and technological progress?

Our analyses focused on three research performance metrics: whether or not the NIH-funded SBIR project resulted in a commercialized technology; the number of new project-related employees retained by the company performing the SBIR-funded project at the completion of Phase II (pre-commercialization) studies; and the overall growth in employment after the completion of Phase II studies directly attributable to the funded SBIR project. We found that firms that received SBIR-funding and partnered with a university were, compared with similar firms that did not partner with a university, less likely to commercialize their technology, less likely to retain employees who were hired to help with the funded project, and less likely to realize employment growth beyond what would have been predicted in the absence of the award.

What might explain these counterintuitive findings? We believe the culprit is the Bayh-Dole Act. The role of Bayh-Dole in distorting the innovation system is apparent. Pursuit of intellectual property has generated conflicts between universities over precedence and restricted the use of knowledge by both university and industry researchers, as exemplified by the competing claims of the University of California, Berkeley, and the Massachusetts Institute of Technology for patents on the CRISPR-Cas9 genome-editing system. There are also patent infringement lawsuits between universities and industry over alleged unauthorized use of university ideas, as with the well-known case of the University of California, San Francisco’s suit over the development of human growth hormone by Genentech. When university collaborations are with new companies, rather than established ones, as is often the case with the SBIR projects, conflicts and possibly insurmountable barriers often arise in the negotiations about how to share any profits that result from a collaborative R&D project. Some industry partners in such projects report that universities overestimate the market value of their research contribution, and this often creates a conflict over how to divide any resulting intellectual property and profits; such conflicts translate into inefficient research partnerships. Most intellectual property negotiations between universities and companies involve technology transfer officers, and case-based research has shown that such individuals are often less well trained in evaluating the commercial prospects for potential technologies than are their company counterparts.

By allowing universities to secure intellectual property rights for findings from publicly funded research, the Bayh-Dole Act inhibits the open flow of knowledge. The exclusive licensing arrangements typically used by universities give established companies rights to use the university’s intellectual property, at standard terms (with modest up-front fees, milestone payments, royalties, and sublicense payments) that provide revenue for the university while leaving generous amounts of expected returns to industry. Exclusive licensing agreements tie up new basic knowledge so that other university and industry scientists can’t use it. Though nonexclusive licensing arrangements can partly solve this problem, they are likely to generate less revenue for the university. Moreover, universities’ technology transfer offices may have little bargaining power with a large company capable of developing ideas generated by academic scientists, and find it hard to resist granting an exclusive license to a company then willing to pay for the right to develop the technology and commercialize it. Bayh Dole requires universities to seek patents for potentially commercializable discoveries. Universities, often increasingly strapped financially, are incentivized to seek maximum returns on these patents. Knowledge created at universities that would otherwise be freely available to innovative companies is instead locked up.

In our view, Bayh-Dole has pushed patenting too far upstream. More effective innovation policies would provide incentives for universities to make their commercializable discoveries freely available and to disseminate them widely. Firms, sometimes working with university partners, could then develop patentable, commercializable biomedical products, and the industrial R&D would in turn provide feedback stimulating new directions for university research and ideas. For example, patents protect valuable intellectual property underlying the great commercial success of drug-eluting stents used in life-saving cardiac interventions. The pharmaceutical company Johnson & Johnson’s Cordis unit holds a very valuable patent for drug-eluting stents, and that patent cites university research. Subsequent university research and patent filings in turn were stimulated; such research develops new types of stents, new drugs to coat the stents, and new delivery systems for implantation of the stents with the prospects for better patient outcomes.

In an era of decreasing state and federal support to higher education, universities are acting rationally as they seek to identify and protect commercially valuable intellectual property under the mandate provided by the Bayh-Dole Act. But freely sharing the knowledge created by university research has social benefits, and when that knowledge is the result of research that has been publicly funded, making the knowledge freely available would seem to be appropriate. Indeed, the history and theory of invention and innovation teach that discoveries and innovations occur most readily when knowledge is shared freely among diverse organizations that are competing to discover critical invention insights.

Though the problem of declining biomedical research productivity almost certainly has many interrelated causes, the possibility that Bayh-Dole is a significant contributing factor has not, to our knowledge, been recognized. Mostly, Bayh-Dole has been celebrated as advancing the nation’s innovation system by greatly increasing the number of university-based patents. Our research suggests that such increases may, on the contrary, be contributing to the slow-down of biomedical innovation. We recommend that Congress revisit the Bayh-Dole Act, and amend it to require that NIH extramural research recipients make public their research results. Open innovation should thus be asserted by NIH as a condition for receiving public monies. Intellectual property would be taken when the freely available ideas created with public funding at US universities are incorporated into developed biomedical technology.

The Second Coming of UK Industrial Strategy

Industrial strategy, as a strand of economic management, was killed forever by the turn to market liberalism in the 1980s. At least, that’s how it seemed in the United Kingdom, where the government of Margaret Thatcher regarded industrial strategy as a central part of the failed post-war consensus that its mission was to overturn. The rhetoric was about uncompetitive industries producing poor-quality products, kept afloat by oceans of taxpayers’ cash. The British automobile industry was the leading exhibit, not at all implausibly, for those of us who remember those dreadful vehicles, perhaps most notoriously exemplified by the Austin Allegro.

Meanwhile, such things as the Anglo-French supersonic passenger aircraft Concorde and the Advanced Gas-cooled Reactor program (the flagship of the state-controlled and -owned civil nuclear industry) were subjected to serious academic critique and deemed technical successes but economic disasters. They exemplified, it was argued, the outcomes of technical overreach in the absence of market discipline. With these grim examples in mind, over the next three decades the British state consciously withdrew from direct sponsorship of technological innovation.

In this new consensus, which coincided with a rapid shift in the shape of the British economy away from manufacturing and toward services, technological innovation was to be left to the market. The role of the state was to support “basic science,” carried out largely in academic contexts. Rather than an industrial strategy, there was a science policy. This focused on the supply side—given a strong academic research base, a supply of trained people, and some support for technology transfer, good science, it was thought, would translate automatically into economic growth and prosperity.

And yet today, the term industrial strategy has once again become speakable. The current Conservative government has published a white paper—a major policy statement—on industrial strategy, and the opposition Labour Party presses it to go further and faster.

This new mood has been a while developing. It began with the 2007-8 financial crisis. The economic recovery following that crisis has been the slowest in a century; a decade on, with historically low productivity growth, stagnant wage growth, and no change to profound regional economic inequalities, coupled with souring politics and the dislocation of the United Kingdom’s withdrawal from the European Union, many people now sense that the UK economic model is broken.

Given this picture, several questions are worth asking. How did we get here? How have views about industrial strategy and science and innovation policy changed, and to what effect? Going forward, what might a modern UK industrial strategy look like? And what might other industrialized nations experiencing similar political and economic challenges learn from these experiences?

Changing views about industrial strategy and science policy have accompanied wider changes in political economy. The United Kingdom in 1979 was one of the most research-intensive economies in the world. A very significant industrial research and development (R&D) base, driven by conglomerates such as BAC (in aerospace), ICI (in chemicals and pharmaceuticals), and GEC (in electronics and electrical engineering), was accompanied by a major government commitment to strategic science.

In common with other developed nations at the time, the UK’s extensive infrastructure of state-run research establishments developed new defense technologies, as part of what the historian David Edgerton called the “warfare state.” Civil strategic science was not neglected either; nationalized industries such as the Central Electricity Generating Board and the General Post Office (later to become British Telecommunications) ran their own laboratories and research establishments in areas such as telecommunications and energy. The Atomic Energy Authority carried out both military and civil nuclear research.

This situation was the product of a particular consensus established following the Second World War. From the left wing of the science-and-technology establishment there was a pre-war enthusiasm for central planning most coherently and vocally expressed by the Marxist crystallographer J. D. Bernal. From the right, there were the military engineers and capitalist chemists who built the Cold War state. From the left side of politics, there was the 1962 government of Harold Wilson proclaiming “the White Heat of Technology” as the mechanism by which the United Kingdom would be modernized. From the right, there was the determination, in the face of the UK’s relative geopolitical decline and economic difficulties, to remain a front-rank military power, with the accompanying decision to develop and maintain an independent nuclear weapons capability.

The ideological basis for an attack on this consensus was developing in the 1950s and 1960s. The leading figure here was Friedrich Hayek, an Austrian-British economist and philosopher and author of forceful critiques of the notion of central planning in general. His friend and intellectual ally, the chemist Michael Polanyi, adapted this argument specifically to oppose the case for planning and direction in science. Polanyi insisted on a strict division between pure science and applied science, introducing the idea of an independent “republic of science” that should remain free of any external direction. This idea was, and remains, very attractive to the world of elite academic science, though it is debatable whether this powerful myth ever described an actual, or indeed a desirable, situation.

Margaret Thatcher was the critical individual through whom these ideas became translated into policy. The influence of Hayek on Thatcher’s general thinking about economics and policy is well known. But Thatcher was also a scientist, whose practical experience was in the commercial world, as an industrial chemist. In a 2017 article in Notes and Records, the Royal Society journal of the history of science, the historian Jon Agar traced the influence of Thatcher’s own experience as a scientist on the evolution of science and innovation policy in her governments. In short, nothing in her experience, or in the experience of those who advised her, would persuade her that there was any special status for science that should exclude it from the market mechanisms to which she believed the whole economy should be subject.

Since the market turn, a key feature of science policy initiated by the Thatcher governments has been the decline of state-sponsored strategic science. By strategic science, I mean science that directly supports what the state regards as strategically important. The outstanding category here is of course the science directly motivated by defense needs. However, strategic science also includes science that supports the infrastructure of the market, for standards and regulation. It could also include science that supports environmental protection, communications infrastructure, medical advance, and the supply of energy.

The obvious point here is that the boundaries of what the state defines as strategic may change with time. Given that the Thatcher government had an explicit goal of shrinking the state, it is unsurprising that the state withdrew support for R&D in areas formerly thought of as strategic. The program of privatization took industries such as steel and telecommunications out of state control and left decisions about the appropriate degree of support for R&D to the market.

This had the largest effect in the area of energy. The privatized energy companies aimed to maximize returns from the assets they inherited, and levels of R&D fell dramatically. What had been a large-scale civil nuclear program was wound down. Even in the core area of defense, there was significant retrenchment, given extra impetus by the end of the Cold War. All but the most sensitive R&D capacity was privatized, most notably in the company Qinetiq. As the historian Agar has emphasized, none of this was an accident, but should all be considered part of a conscious policy of withdrawing state support from any near-market science.

The withdrawal of the UK state from much strategic R&D provided a test of the notion favored by some free market ideologues that state spending on R&D crowds out private-sector spending. In fact the reverse happened; the intensity of private-sector R&D investment fell in parallel with that of the state’s. The relationship between the two may not be straightforward, however, as the market turn in UK politics led to significant changes in the way companies were run. A new focus on maximizing shareholder value and an enthusiasm for merger and acquisition activity in the corporate sector resulted in the loss of industrial research capacity.

The fate of the chemicals conglomerate ICI provides a salutary example. A hostile takeover bid from the corporate raider James Hanson in 1991 prompted ICI to “demerge” by separating its bulk chemicals and plastics business from its pharmaceuticals and agrochemicals businesses. The company housing pharmaceutical and agrochemical operations—Zeneca—underwent further divestments and mergers to produce the pharmaceutical company AstraZeneca and the agrochemical company Syngenta. The residual rump of ICI, attempting to pivot toward higher-value specialty chemicals, made an ill-timed, debt-financed purchase of National Starch. A series of divestments failed to lift the debt burden, and what was left of the company was sold to the Dutch company Akzo-Nobel in 2007.

The story of the electronics and electrical engineering conglomerate GEC offers some parallels to the ICI story. In the 1990s, GEC sold its less exciting businesses in electrical engineering and electronics in order to make acquisitions in the booming telecom sector. Renamed Marconi, the company had to restructure after the bursting of the dot-com bubble, and finally collapsed in 2005.

These corporate misadventures resulted in a loss of a significant amount of the UK’s private-sector R&D capacity across a wide range of areas of technology. The common factor was a belief that the route to corporate success was through corporate reorganization, mergers, acquisitions, and divestments rather than through researching and developing innovative new products. There are parallels here with the decline of long-term, strategic R&D in some big corporations in the United States, such as General Electric, AT&T Bell Laboratories, Xerox, Kodak, and IBM, though in the United Kingdom the loss of capacity was significantly greater and took place with no compensating new entrants at the scale, for example, of the US company Google.

It is also possible to interpret these stories as highlighting different beliefs about information and the power of markets. In the old industrial conglomerates such as ICI and GEC, long-term investments in R&D were made by managers and paid for by the retained profits of the existing businesses (which for companies such as GEC were substantially boosted by government defense contracts). A newer view emphasizes the role of the market as a more effective device for processing information; in this view, money locked up in the conglomerates would have been better returned to shareholders, who would have invested it in innovative, new companies.

There are arguments on both sides here. On one hand, questions can clearly be asked about the motivations and effectiveness of the managers of the conglomerates. They may seek to protect the incumbent position of existing technologies, they may be too reluctant to adopt new technologies developed outside their organization, and they may be inhibited by the scale and bureaucracy of their companies. On the other hand, one result of the turn to the markets has been a sequence of investment bubbles resulting in substantial misallocation of capital, together with a pervasive short-termism. Whatever the mechanisms at work, the outcome is not in doubt: a significant loss of private-sector R&D capacity in the United Kingdom since the Thatcher era.

The obverse of the ideological determination of Thatcher and her advisers to withdraw support from near-market research was a new valorization of “curiosity-driven” science. The result was a new, rather long-lasting consensus about the role and purpose of state-supported science that emphasized economic growth as its primary goal. But its tacit assumption was that innovation could be driven entirely from the supply side. In this view, the best way to make sure that state-supported science could contribute to a strong economy was by creating a strong underpinning of basic research, developing a supply of skilled people, and removing the frictions believed to inhibit knowledge transfer from the science base to the users of research.

The supply-side view of science policy was first clearly articulated in 1993, in a white paper introduced by the Conservative science minister William Waldegrave. This influential document halted a pattern of decline in research funding in the academic sector, using the classical market failure justification to call for the state to fund basic research. It reasserted the role of the private sector as the key funder of applied research, and with a continued program of privatization of government research establishments ensured a further withdrawal of the government from strategic research.

The advent of a Labour government in 1997 did not change matters. In line with the general acceptance of the post-Thatcher settlement, there was considerable policy continuity. A major policy statement in 2004, under the sponsorship of an influential and long-serving science minister, Lord Sainsbury, restated the principles of supply-side science policy.

The Sainsbury approach included new elements that reflected the changing corporate R&D landscape: more emphasis on spin-out companies based on protectable intellectual property and funded by venture capitalists, and on the aspiration to attract overseas investment. A sense that there was now too little private-sector research underpinned an explicit target for increasing business R&D over the next 10 years, to 1.7% of gross domestic product (a target that was conspicuously missed, as the figure currently stands at 1.1%).

The main practical effect of the 10-year investment framework was a series of real-term increases in spending on academic research. This was accompanied by a further run-down of strategic research, with R&D spending by government departments continuing to decrease.

Meanwhile, policy-makers displayed a growing sense that the academic research base, now benefitting from a more generous funding settlement, should be pressed harder to make sure it delivered economic growth. This expectation manifested itself in a heightened rhetoric about “impact,” with various bureaucratic measures to incentivize and reward activities that produced such economic effects, whether through the formation of spin-out companies or through collaboration with established businesses. These measures culminated in the 2014 Research Excellence Framework, which included impact as a criterion to be assessed in university research, and whose results directly determine university research funding.

The emphasis on impact produced the paradoxical effect that even as the overall balance in the UK’s research system in fact shifted from strategic research toward undirected research, many people in the academic part of the system felt that they were being pressured to make their own research more applied.

The industrial policy of the Conservative governments between 1979 and 1997 was to not have an industrial policy. The New Labour government of 1997 broadly accepted this consensus, in particular resisting so-called vertical industrial policy—that is, specific measures in support of particular industrial sectors.

Yet absolute opposition to industrial policy was at times also honored in the breach. The government’s policy of partial devolution to Scottish and Welsh assemblies gave an economic development function to these administrations and to agencies in the English regions. In 2007 an innovation agency—the Technology Strategy Board—was given free-standing status, empowered to award collaborative R&D grants to industry and to oversee some cross-sector networking activities, mostly between industrial partners.

But it took the global financial crisis of 2007-8 to bring about a change in mood. A new, powerful business minister in Gordon Brown’s Labour government, Peter Mandelson, emphasized the need to rebalance the economy away from the financial sector and toward manufacturing. The automobile sector was singled out for a series of interventions. Most strikingly, plans called for the government to form a new class of translational research centers, modeled on the successful and much-envied centers developed by the Fraunhofer Society, a major German research organization.

In 2010, the new Conservative-Liberal Democrat coalition government accepted the research center plan, continued the support for the automobile sector, and began to speak of industrial policy again. In practice, policy consisted of a mixture of sector-based support and the championing of selected technology areas, and it could be argued that many of the interventions were inadequate in scale. But perhaps the most important significance of this development was that after 30 years in which the very words industrial strategy were essentially unspeakable in the British state, there was now an acceptance, even in polite political circles, that support for industry was a proper role for government.

What does the innovation landscape in the United Kingdom now look like, after the dramatic shifts of the past three decades? The overall R&D intensity of the UK economy, which 30 years ago was among the highest in the world, is now low compared not only with traditional competitor economies, such as France, Germany, and the United States, but with the fast-growing economies of the far East, such as Korea and China.

Within the United Kingdom’s R&D enterprise, there is an academic science base that is very high performing when measured by academic metrics such as citations. But there are some notable problems on the industrial side. Uniquely for a developed economy of the UK’s size, more than half of industrial R&D is conducted by foreign-owned companies. This industrial R&D is concentrated in a few sectors, dominated by the pharmaceutical industry, with other major contributions in aerospace and computing. The biggest change in recent years has been seen in automobiles, where industrial R&D more than doubled since 2010, perhaps reflecting its status as the test-bed of the new wave of industrial strategy.

State-supported translational research is, with a very few exceptions, weak. The new Fraunhofer-inspired “Catapult Centres,” established post-2010, are finding their feet. Two of the most successful centers were built around preexisting initiatives, and they are worth considering in more detail as demonstrations of how new translational research capacity can be created. These are the Warwick Manufacturing Group (WMG) at the University of Warwick and the Advanced Manufacturing Research Centre (AMRC) at the University of Sheffield. Both are the creations of individual, highly entrepreneurial academics (Lord Kumar Bhattacharyya at WMG and Keith Ridgway at AMRC), and both began with a strong sector focus (automotive at WMG and aerospace at AMRC).

Although both institutions have grown out of conventional research universities and remain associated with them, their success arises from a mode of operation very different from university-based science, even in applied and technical subjects. AMRC began as a collaboration with the aircraft manufacturer Boeing, soon joined by the aero-engine manufacturer Rolls-Royce. Much of the research is focused on process optimization, and it is carried out at industrial scale so that new processes can rapidly be transferred into manufacturing production.

A key feature of such translational research centers is the way that the large companies that form their core partners—Boeing and Rolls-Royce in the case of AMRC, and Jaguar Land Rover for WMG—can bring in smaller companies that are part of, or aspire to be part of, their supply chains, involving them in joint research projects. Another way in which these translational research centers extend the mission of the traditional research university is through a greater involvement in skills development at all levels, including the technical skills typical of an engineering apprenticeship program. One measure of the success of the institutions is the degree to which they have been able to attract new investment in high-value manufacturing into what since the 1980s had been underperforming regions that had failed to adapt to successive waves of deindustrialization

Meanwhile, economists and policy-makers in the United Kingdom and the United States are increasingly recognizing that the effects of deindustrialization on regional economies have in the past been underestimated. For example, in a 2009 article in Harvard Business Review, Gary Pisano and Willy Shih, both professors of business administration, drew attention to the way in which manufacturing anchors what they called a “manufacturing commons,” the collective resources and knowledge that underpin a successful regional cluster.

These commons are based on the collective knowledge, much of it tacit, that drives innovations in both products and processes. A successful manufacturing commons is rooted in R&D facilities, networks of supplying companies, informal knowledge networks, and formal institutions for training and skills. Pisano and Shih’s key point is that the loss of a manufacturing plant, perhaps through outsourcing, can have a much greater impact than the direct economic impact of the loss of the plant’s jobs, by eroding this larger manufacturing commons.

But stories such as those of the Sheffield Advanced Manufacturing Research Centre suggest that manufacturing commons can be rebuilt. The emerging formula brings together several elements. Research facilities need to have an avowedly translational focus, and they should create strong research partnerships between or among academia, large companies already operating at the technological frontier, and smaller companies wishing to improve their innovation practices, possibly to make them more competitive as suppliers to the large companies. Education institutions need to focus on building skills at all levels. They should be linked with these research centers, creating clear pathways for individuals to progress from intermediate-level technical skills to the highest-level qualifications in technology and management. As these research facilities become successful and recognized, this should lead to a virtuous circle in which further inward investment is attracted and the existing business base grows in capability.

The past decade has seen a new consensus about industrial strategy emerge in the United Kingdom, to this extent at least: the Conservative government has a department with industrial strategy in its title (the Department of Business, Energy and Industrial Strategy) and has published a major policy document on the subject, and the opposition Labour Party advocates an industrial strategy as a major plank of its alternative economic policy.

To what extent is a consensus emerging on the substance of what an industrial strategy looks like? One attempt to articulate a new consensus has recently been made by the Industrial Strategy Commission, an independent initiative supported by the Universities of Sheffield and Manchester, of which I was a member.

In the commission’s view, the beginning of a strategy needs to recognize some of the real weaknesses of the UK economy now. One key issue that has become particularly pressing since the global financial crisis is the very low rate of domestic productivity growth. There is a global context here, in that productivity growth throughout the developed countries has been slowing since the 1980s. But the situation in the United Kingdom is particularly bad: levels of productivity were already significantly below those achieved in the United States, France, and Germany, and the slowdown since the global financial crisis has been dramatic.

The United Kingdom also has gross geographic disparities in economic performance, with an economy dominated by a single city, London. The UK’s second-tier cities underperform, there are many very poor urban areas that have not recovered from 1980s deindustrialization (analogous to the US Rust Belt), and many places in the rural and coastal peripheries have been left behind by economic success elsewhere.

As the commission sees it, an industrial strategy should be framed with a view of the whole economy, not just a few high-technology sectors. It needs to recognize the importance of the state as an actor uniquely able to coordinate activities and create new markets. And if it is to have a long life, the strategy needs to be linked to the broader long-term strategic goals of the state.

One positive aspect of the 1980s turn to free market liberalism has been an increased recognition of the importance of competition in driving innovation. But the wave of privatization that occurred has produced a set of industries (in transport, energy, and water, for example) that are heavily regulated by the state, but whose structure and incentives do not seem to reward new investment or innovation. This needs to be rethought.

The United Kingdom has underinvested in infrastructure for many years. For traditional hard infrastructure—roads and railways—the investment criteria used to assess new investments have rewarded parts of the country where the economy is already strong, and this must change. Of equal importance, investment needs to include the infrastructures underlying newer parts of the economy, such as mobile telephony and fast broadband internet coverage. Nor should the soft infrastructure that makes successful industrial societies function be neglected—in education and health, for example. The commission’s headline recommendation here is for a Universal Basic Infrastructure guarantee to ensure that all parts of the country have in place the infrastructure needed to make economic success possible.

Policy-makers across the political spectrum now seem to realize that the R&D intensity of the UK economy needs to increase. But this needs to be done in a way that considers the whole landscape: public and private-sector, undirected, use-inspired, translational, and strategic. More emphasis is required on the translational part of the picture than we’ve seen before, and the links to skills at all levels need to be made more coherent. Currently the geographical distribution of R&D, in public and private sectors alike, is highly imbalanced, with the biggest investments being made in the most prosperous parts of the country: London and the South-East. This too needs to change; if new R&D institutions are to be set up, the role they can have in catalyzing regional economic growth needs to be explicitly considered when decisions are made on their location.

Above all, the United Kingdom needs to move beyond the supply-side science policy that has dominated innovation thinking for the past three decades. More attention needs to be paid to generating demand for innovation. Here the government can have a central role, buy using its spending power much more purposefully to encourage innovation in the private sector, especially when linked to the strategic goals of the state. In the UK’s case, these include a long-term commitment to reducing the carbon intensity of the energy economy while maintaining the security and affordability of energy to domestic consumers and industry. The United Kingdom also maintains a wide, cross-party consensus in support of universal health care coverage. These goals are unlikely to be deliverable without substantial innovation. Done right, industrial strategy should enable the state to meet its strategic goals while at the same time providing the new business opportunities for the private sector.

In the post-war years, the United Kingdom, like other developed countries, had a warfare state, which did successfully drive innovation. The innovation system associated with the warfare state was dismantled, and what has arisen in its place has not been sufficient to drive economic growth or to meet the long-term challenges UK society faces. This, too, seems to be a difficulty shared by the United States and other industrialized nations.

We should not be nostalgic for the Cold War, but the United Kingdom does now need to rebuild an innovation system appropriate for its current challenges. Rather than attempting to re-create the military-industrial complex of the past, we should aspire to a social-industrial complex that can drive the innovation that is needed to create a sustainable, effective, and humane health and social care system and to place the energy economy on a sustainable, low-carbon footing.