Innovation Policy for Complex Technologies
U.S. technology policy must be revamped to deal with accelerating technological and organizational complexity.
The complexity of the technologies that drive economic performance today is making obsolete the mythic image of the brilliant lone inventor as well as undermining the effectiveness of traditional U.S. technology policy. Innovation in what we define as complex technologies is the work of organizational networks; no single person, not even a Thomas Edison, is capable of understanding them fully enough to be able to explain them in detail. But our mythmaking is not all that needs to be updated. The new processes of innovation are also undermining the effectiveness of traditional U.S. technology policy, which places a heavy focus on research and development (R&D) and unfettered markets as the major sources of the learning involved in innovation, while downplaying human resource development, the generation of technology-specific skills and know-how, and market-enhancing initiatives. This is inconsistent with what we now know about complex technologies. Innovation policies should be reformulated to include a self-conscious learning component.
In 1970, complex technologies made up 43 percent of the 30 most valuable world goods exports. By 1995, their portion had risen to 82 percent. With the rapid growth in the economic importance of complex technologies has come a parallel growth in the importance of complex organizational networks that often include firms, universities, and government agencies. According to a recent survey, more than 20,000 corporate alliances were formed between 1988 and 1992 in the United States; and since 1985, the number of alliances has increased by 25 percent annually. Complex networks coevolve with their complex technologies.
As complexity increases, the rate of growth and the characteristics of organizational networks will be significantly affected by public policy. The most important effects will be on network learning. Technological progress requires that networks repeatedly learn, integrate, and apply a wide variety of knowledge and know-how. The computer industry, for instance, has required repeated syntheses of knowledge from diverse scientific fields (such as solid-state physics, mathematics, and language theory) and a bewildering array of hardware and software capabilities (including architectural design and chip manufacturing). No single organization, not even the largest and most sophisticated firm, can succeed by pursuing a go-it-alone strategy in the arena of complex technologies. Thus, complex technologies depend on self-organizing networks that behave as learning organizations for success in innovation.
Networks have proven especially capable of incorporating tacit knowledge into their learning processes (such as unwritten know-how that often can be understood only with experience). Examples of tacit knowledge include rules of thumb from previous engineering design work, experience in manufacturing operations on the shop floor, or skill in using research instruments. Learning based on tacit knowledge tends to move less easily across organizational and geographical boundaries than more explicit or codified learning. Therefore, tacit learning can be a major source of competitive advantage.
Because of the centrality of network learning, public policy aimed at fostering innovation in complex technologies must give attention to three broad initiatives.
Developing network resources. Networks have at least three sets of resources: existing core capabilities, already internalized complementary assets, and completed organizational learning. A successful network must hold some core capabilities–that is, it must excel in certain aspects of innovation. Among the most important and difficult core capabilities to learn (or to imitate) are those that are essential to systems integration. Because there are many different ways to organize the designing, prototyping, manufacturing, and marketing of a complex technology, it is obvious that the ability to quickly conceptualize the technology as a whole and carry it through to commercialization represents a powerful, often dominant, capability. The design of a modern airplane, for instance, demands the ability to understand the problems and opportunities involved in integrating advanced mechanical techniques, digital information technology, new materials, and other specialized sets of technologies.
Engineering design teams in the aircraft industry typically contain about 100 technical specialties. The design activity requires a systems capability that may constitute a temporary knowledge monopoly built on some of the most complex kinds of organizational learning. In something as complex as the design of an aircraft, systems integration involves the ability to synthesize participation from a range of network partners. There is no way to achieve analytical understanding and control of integration of this type; the capability is, in part, experience-based, experimental, and embodied in the structure and processes of the network. Some have called this integration an organizational technology that “sits in the walls.”
U.S. national innovation policy has paid little attention to network resources. Federal policies relevant to the education and training of the workers essential to network capabilities have been limited, tentative, and contradictory, reflecting an ideological predisposition against a significant government role. However, the realization that broadly based human resource policies are critical to the future of U.S. innovative capacity seems to be gaining ground. Jack Gibbons, former director of the White House Office of Science and Technology Policy, has urged assigning a higher priority to the lifelong learning needs of the future science and technology workforce. And Rep. George Brown, who until his recent death served as the ranking minority member of the House Science Committee, made education and training a key part of his “investment budget” proposal.
But even Gibbons and Brown did not identify the special educational needs that result from the growing importance of networks. In addition to scientific and technical knowledge, successful networks require people who know how to function effectively in groups, teams, and sociotechnical systems that include individuals and organizations with diverse tacit and explicit knowledge. The importance of this kind of social knowledge is underlined by the fact that companies such as Intel spend major training resources on teaching their employees how to function in groups. Nothing would be more useful for evolving innovation networks than a national capacity for appropriate education and training. Appropriate training and retraining of personnel that provides continuous upgrading of needed skills would contribute to the workforce’s ability to adapt with changes in technology. Human resource competencies that include both technical and social knowledge are inseparable from network core capabilities. To the extent that public policy can help provide the needed range of worker skills and know-how, networks will be better able to make rapid adjustments.
Direct U.S. interventions designed to develop capabilities at the firm, network, or sector level have been rare. Governmental support for companies to develop their capabilities in flat-panel displays is an obvious exception. The effort was aimed at recapturing state-of-the-art video capabilities lost when Japan drove the United States out of the television manufacturing business in the 1970s. Advanced video capabilities are widely believed to be essential core technologies in the information society. The justification for government funding was defined largely in terms of defense requirements. But because the need to rebuild core video capabilities in U.S. firms was seen as critical, the civilian sector was included in the initiative.
If technologies and the capabilities they embody diffused rapidly across company boundaries and national borders in a global economy, there would be no need for this type of policy. But because not all relevant know-how is explicitly accessible and because the ability to absorb new capabilities depends in part on an available mix of old capabilities, technology diffusion is a process of learning in which today’s ability to exploit technology grows out of yesterday’s experience and practices. Thus, the demise of the U.S. consumer electronics industry brought with it a corresponding decline in the capability to produce liquid crystal displays in high volume, even though the basic technology was explicitly available. Active resource development policies such as the flat-panel display program have been attacked as industrial policy that picks winners and defended as a dual-use exception to the general rule of no government involvement. This debate illustrates the U.S. preoccupation with the concepts and language of an earlier era. Nonetheless, Richard Nelson of Columbia University is persuasive when he argues that technology policy ought to have a broader industry focus than the relatively narrow flat-panel initiative. Not only are the politics of broader programs more ideologically and politically plausible in the United States, but more effective public-private governance mechanisms appear easier to develop when industry-wide technological issues are being addressed.
Creating learning opportunities. Many of the most important changes needed in U.S. technology policy are related to learning opportunities. The history of any network includes a set of boundaries (sometimes called a path dependency) that both restricts and amplifies the learning possibilities and the potential for accessing new complementary assets (such as sources of knowledge outside the network). The network learning that has taken place in the past is a good indicator of where learning is likely to take place in the future. Most networks tend to learn locally by engaging in search and discovery activities close to their previous learning. Localized learning thus tends to build upon itself and is a major source of positive feedback, increasing returns, and lock-in.
A history of flexible and adaptive learning relationships within a network (with suppliers, customers, and others) provides member organizations with formidable sources of competitive advantage. Alternatively, allowing learning-based linkages to atrophy can lead to costly results. For instance, inadequate emphasis on manufacturing and a lack of cooperation among semiconductor companies contributed to an inability to respond rapidly to the early 1980s Japanese challenge. When the challenge became a crisis, cooperative industry initiatives moved U.S. companies toward closer interactions with government (such as the Semiconductor Trade Arrangement) and eventually to a network (the Sematech consortium) that improved the ability of industry participants to learn in mutually beneficial ways, including collaborative standards setting.
The Sematech experience is an example of efforts to enhance network learning through government-facilitated collaborative activities. Like the flat-panel display initiative, Sematech was justified largely in national security terms, and like so much technology policy, exaggerated emphasis was initially focused on R&D, with too little attention given to other learning opportunities.
R&D funding must continue to be a high government priority, but the primacy given to R&D is a problem for innovation in complex technologies. It is assumed that support for R&D–and sometimes only the “R”–is synonymous with technology policy. Rep. Brown called this an “excessive faith in the creation of new knowledge as an engine of economic growth and a neglect of the processes of knowledge diffusion and application.” Support for R&D certainly creates learning opportunities, but many other learning avenues exist that have little or nothing to do with R&D, and these are especially evident in networks (see sidebar).
The policy overemphasis on R&D skews learning and the generation of capabilities that are often needed for innovation success. R&D support is easy for the government to justify, but it is often not what companies and networks need. What are frequently needed are new or enhanced organizational capabilities that facilitate development of tacit know-how and skills, integrated production process improvements, and ways to synthesize and integrate the talents and expertise of individuals into work groups and teams. These organizational “black arts” are not usually a part of the R&D-dominated policy agenda, but if the challenge of innovating complex technologies is to be met, policy will need to be flexible enough to incorporate them as well as other nontraditional ideas.
Enhancing markets. U.S. innovation policy has tended to emphasize only one set of factors that affect markets: improving competitive incentives to firms by measures such as strengthening intellectual property rights and the R&D tax credit. The U.S. fascination with factors such as patents as stimulators of technological innovation ignores the need for other kinds of market-enhancing policies.
Markets left unfettered except for incentives to compete have trouble coping with the cooperative network learning dynamics that are at the core of innovation in complex technologies. Learning in complex networks is often very risky and can encompass prolonged tacit knowledge acquisition and application sequences. Complex learning involves substantial coordination, because investments must be made in different activities, often in different sectors, and increasingly in different countries. Collective learning tends to induce a self-reinforcing dynamic; it becomes more valuable the more it is used. Failure to recognize and adapt to these characteristics of complex network learning is a major source of market failure.
When technological innovation is incremental, which is the normal pattern of the evolution of most complex organizations and technologies, learning-generated market failures are less common. Well-defined market segments and well-developed network relationships with a wide array of users and suppliers are usually built on extensive incremental learning and adaptation. Over time, incremental innovations enhance the stability of markets, because a consensus develops concerning technological expectations. These expectations, in conjunction with demonstrated capabilities, provide a framework within which market signals can be effectively read and evaluated.
Alternatively, when innovation is not incremental, learning-based market failures proliferate. During periods of major change, stability and predictability erode, and markets provide unclear signals. When the innovation process is highly exploratory, the networks that are being modified are less responsive to economic signals, because the market has little knowledge of the new learning that is taking place and the new capabilities being developed. In such situations, linkages to other organizations (such as relationships with other networks or government) or the status of institutions (such as regulatory regimes) matter more than market processes, because they provide some stability and limited foundations on which decisions can be based. Even when well-defined markets do emerge, achieving stability can take a long time, often more than a decade for the most radical complex innovations.
Because innovation in complex technologies tends to foster many market failures, significant benefits arise from having connections to at least three kinds of institutions: (1) state-of-the-art infrastructure, including communication, transportation, and secondary educational systems; (2) appropriate standards-setting arrangements ranging from environmental regulations to network-specific product or process standards; and (3) closer linkages between firms and other national science and technology organizations, including national laboratories and universities. Establishing these connections can be facilitated by public policy.
Complex innovation takes place through market (competitive) and nonmarket (cooperative) transactions, and the latter involve not only businesses but also other institutions and organizations. Networks seeking to enhance the success of their innovations frequently find themselves involved in what Jacqueline Senker of Edinburgh University, in a study of the industrial biotechnology sector, refers to as “institutional engineering,” the process of “negotiating with, convincing or placating regulatory authorities, the legal system, and the medical profession.” In the United States, the federal government is most capable of affecting these market and nonmarket relationships in a systematic way.
Policymaking aimed at complex technologies is fraught with uncertainty. There is no way to be assured of successful policy in advance of trying it; the formulation of successful policy is unknowable in a detailed sense. Thus, policy prescriptions developed in the absence of the specific context of innovation are as dangerous as they are tempting. With this uncertainty in mind, the following policy guideposts seem useful.
Complex networks offer, through their capacity to carry out synthetic innovation, a broad capability for innovating, although it is not possible for individuals to understand the process in detail. Policy, too, must be made without any capacity for understanding in a detailed sense what will work. It will always be an approximation–never final, but always subject to modification. Flexibility is key. Small, diverse experiments will tend to be more productive in learning terms than one big push along what appears to be the most likely path at the outset.
Many existing U.S. technology projects and programs are relatively small and have had significant learning effects, but few were designed as learning experiments. For instance, the Advanced Technology Program (ATP) currently operates at a relatively modest level and has been credited with encouraging collaboration among industry, government laboratories, and universities. According to an analysis by Henry Etzkowitz of the State University of New York at Purchase and Magnus Gulbrandsen of the Norwegian Institute for Studies in Higher Education, ATP’s stimulation of industrial collaboration may be more significant than the work it supports. “ATP conferences have become a marketplace of ideas for research projects and a recruiting ground for partners in joint ventures,” they write. Such generation of learning by interaction was largely unanticipated. In the world of complex technologies, the unanticipated has become the norm.
Although policy must be adaptable and made without detailed understanding, it does not follow that knowledge and information are of little or no value. To the contrary, the most successful policymaking will usually be that which is best informed. Being informed in the era of complex technologies requires exploiting as much expertise as possible. Designing and administering complex policies requires, at a minimum, technological, commercial, and financial knowledge and skills. If these cannot be developed inside government, outside expertise must be accessed. Only policy informed by state-of-the-art knowledge of the repeated nonlinear changes taking place in the various technology sectors can be appropriately adaptive. Only those who are intimately involved in innovation in complex technologies can provide knowledge of what is happening. As a start in the right direction, the White House should take the initiative in reforming conflict of interest laws and regulations that are barriers to public-private sector interaction.
For example, to protect against collusion, current regulations preclude ex-government employees from closely interacting with their former colleagues for prescribed periods of time following their departure. But because technology can change rapidly, the knowledge of the former employee can quickly become obsolete. In today’s world of accelerating technological innovation, the costs of knowledge obsolescence probably outweigh the costs of collusion.
Another possibility involves policies that encourage those at project and program levels in government to make frequent use of advisory groups composed of people from industry, nonprofit organizations, universities, governmental organizations, and other countries. Beginning in the 1970s, advisory panels fell into disrepute and their creation at times required prior approval from the Office of Management and Budget. They were frequently seen as being not only costly but also as vehicles for inappropriate influence. What they offer in the era of complex technology is a valuable vehicle for knowledge exchange and learning. Such groups can facilitate the kind of trust that is especially valuable in dealing with tacit knowledge.
The objective of broader private-sector involvement is to enhance policy learning. Negotiations between private and government policymakers most likely will lead to consensus in some areas, but even if the immediate outcome is the recognition of conflicting interests, there is learning taking place if in the process new network practices, routines, or behaviors are identified, and new data sets are cataloged. Particularly important would be new insights from the private sector regarding the effects of previous public policies.
Traditional boundaries are of less use to those making policy. Complex networks and their technologies blur boundaries across the spectrum. For example, the proliferation of complex networks has made it difficult to define the boundaries of an organization as a policy target or objective. When a label such as “virtual corporation” is used to describe interfirm networks or “business ecosystem” is applied to networks that include not only companies but also universities, government agencies, and other actors, one can appreciate how amorphous the object of policy has become. This complexity is even greater when the focus is network learning, which typically involves a messy set of interactions among a variety of organizational carriers of both tacit and explicit core capabilities and complementary assets. In such situations, running even small learning experiments informed by private-sector expertise puts a premium on incorporating evaluation procedures to determine which organizations are being affected, and how. These policy evaluations must provide for reviews, amendments, and/or cancellation.
But policy evaluation must be more systemic than the traditional U.S. emphasis on cost efficiency for particular actors and projects. Network learning often confers benefits that are broader than immediate economic payoffs. For instance, networks may interact in ways that generate a form of social capital, a “stock” of collective learning that can only be created when a group of organizations develops the ability to work together for mutual gain. Here too, ATP is credited with producing positive, if largely unintended, social outcomes as a consequence of learning by interaction. More effort needs to be made to build such social factors into assessments of program success or failure. A promising option is to make more use of systems-oriented evaluation “benchmarking” or assessments of system-wide “best evaluation practices,” as compiled by international bodies such as the Organization for Economic Cooperation and Development.
Continuous coevolution between complex organizations and technologies is the norm. The dominant pattern will be the continuous emergence of small, incremental organizational and technological adaptations, but this pattern will be punctuated by highly discontinuous and disruptive change. The need for policy is greatest when change is discontinuous–when coevolving networks and their technologies have to adapt to major transitions or transformations. Policy that is sensitive to this process of adaptation must be informed by strategic scanning and intelligence. Government participation in the generation of industrial technology roadmaps is a particularly valuable way to gather intelligence regarding impending changes in innovation patterns. Roadmaps such as those produced by the semiconductor industry generally represent a collective vision of the technological future that serves as a template for ways to integrate core capabilities, complementary assets, and learning in the context of rapid change. Roadmaps also facilitate open debates about alternative technological strategies and public policies. Cross-sectoral and international road mapping exercises would be particularly valuable, because many of the sources of discontinuous change in complex technological innovation originate in different sectors and economies.
The great challenge for policymakers is to find an accommodation between the set of industrial system ideas and concepts that are the currency of contemporary policy debate and formulation and the reality of continuous technological innovation that has moved beyond that currency and is incompatible with it. We need a new policy language based on new policy metaphors. Metaphors are in many ways the currency of complex systems. By way of metaphors, groups of people can put together what they know, both tacitly and explicitly, in new ways, and begin to communicate knowledge. This is as true of the making of public policy as it is of technological innovation. The terminology we have used in this article allows one to address large portions of the technological landscape (such as the role of core capabilities in network self-organization) that are completely ignored when traditional labels and terms are used.
Policy guidelines that stress the shared public-private governance of continuous small experiments, chosen and legitimized in a new language, backed by strategic intelligence, and subject to careful evaluation may not sound like much.
But the study of complexity in organizations and technologies communicates no message more clearly than that even small events have unanticipated consequences, and one of the most dramatic messages is that sometimes small events have dramatic unanticipated consequences. Our policy guideposts are not a prescription for pessimism. Indeed, a major implication of innovation in complex technologies is that even modest, well-crafted, adaptive policy can have positive consequences that are enormous.