No model for policymaking
Useless Arithmetic: Why Environmental Scientists Can’t Predict the Future by Orrin H. Pilkey and Linda Pilkey-Jarvis. New York: Columbia University Press, 2007, 230 pp.
Stephen R. Carpenter
Technical solutions have been found for many complicated problems of environmental science. For example, enormous challenges have been overcome to deploy and maintain networks of sophisticated Earth-observing satellites that provide rich global perspectives on environmental change. Complicated problems can be broken down into sequences of smaller, tractable problems that involve predictable and controllable components. A solution that works for one complicated problem will work, with small adjustments, for a similar complicated problem; once one satellite is in place, many of the same approaches can be used for the next one. Through sequential solutions of complicated problems, technology has greatly improved human well-being in the past and will continue to do so in the future.
Many environmental problems, though, are beyond complicated: They are complex. Examples include global climate change; the sustainable mitigation of poverty; and managing tradeoffs among interacting ecosystem services such as food, fresh water, and wild living resources. Such problems self-organize from the interactions of trillions of organisms and decisions by millions of people in a changing world of turbulent atmosphere, ocean, and earth-surface dynamics. Successful solution in one place or time does not guarantee success elsewhere or in the future. Apparently successful solutions seem to sow the seeds of future failures. Prediction and control, the keys to solving complicated problems, fail in complex settings for several reasons, including lack of essential information, nonlinear dynamics, and human volition.
Complex problems must be faced with great humility because control is limited and predictions are unreliable. Yet predictions, however fraudulent, can have enormous economic and political influence if they are taken seriously by society. In this concise, powerful, and readable book, Orrin Pilkey and Linda Pilkey-Jarvis expose abuses of prediction in environmental decision-making. Their specific target is abstruse computer models used by private organizations or government agencies aiming to create spurious certainty, suppress alternative approaches, and influence public policy to reward narrow interest groups. Thus abuses that would be mere hubris if committed by an individual become sociopathic. The models are not designed to shed light on a problem but to create a politically advantageous shortcut to a self-interested outcome.
The book’s sharp critique might turn away some engineers and scientists who appreciate the value of computer modeling. That would be unfortunate, because the argument merits careful consideration. In their last chapter, the authors sharpen the focus of their attack to political misuse of elaborate computer models applied to complex problems. They acknowledge the many successful uses of such models to solve complicated problems of engineering, and the many valid uses of models in basic scientific research. Their beef is with the assertion that the models can resolve complex problems. The boundary between complicated and complex problems is neither precise nor static; it changes continually as society, science, and technology evolve. Thus, the authors argue that the proper uses of models in public policy need to be carefully considered by all environmental scientists.
The strongest chapters of the book address coastal engineering. Beach dynamics are richly dynamic on multiple scales. Local change depends on input-output balances of sand, which have deceptively complex relationships to underwater topography, wind, and currents. Measurements are easily made during calm weather, but these have little relevance to events during storms, when the massive, important changes occur. The authors show how oversimplified models have been applied repeatedly in schemes for beach nourishment or shoreline stabilization that turn coastal developments into accidents waiting to happen.
Modeling concrete and steel for bridges, dams, and elevated water towers is relatively easy. There are few surprises and the designs incorporate large safety factors, so failures are few. Modeling beaches, on the other hand, is very different. They are complex systems that operate under the control of many variables, which are often poorly understood. The various parameters involved in creating beach change work simultaneously and in unpredictable order, timing, and magnitude. There are many surprises. No one knows when the next storm will happen by, and this fact alone wreaks havoc on the neat and orderly world of mathematics at the shore.
Presumably the failed coastal developments enrich a few people while offloading the costs on others, although this social context is not developed in detail in the book. Case studies of open-pit mining reveal similar histories of failure in a situation where the economics are simpler.
Yet in other cases models are used more successfully. Climate change scenarios use enormous computer models, but here the social setting is different in important ways. The models, though complicated, are openly discussed and continually improved and reevaluated by a global community of scientists. The models are only a part of the body of evidence for climate change, which also involves direct measurements of climate and the atmosphere, trends in sea ice and lake ice, natural archives in ice or sediment, and many other sources. For the public and decisionmakers, the models may be almost invisible: One picture of a polar bear on a melting ice floe is worth a thousand computer runs. Pilkey and Pilkey-Jarvis also praise the transparent, open use of models in other areas of public policy such as the management of invasive species.
Though Useless Arithmetic is a compelling assessment of model abuse in environmental science, it is less successful in pointing to solutions. The authors draw a distinction between quantitative and qualitative models, and they link abuses to the quantitative ones. The distinction seems irrelevant and vague. The difficulty is not the models but the context in which they are used.
Uncertainty is not an intrinsic property of nature; it emerges from the problems that society faces and the institutions and intellectual tools (including models) used to address them. The abuses deplored by Pilkey and Pilkey-Jarvis all involve the false narrowing of uncertainty. This occurs when political processes use opaque models to close off alternatives and limit public debate. However, other political processes do the opposite. Global assessments such as the Intergovernmental Panel on Climate Change and the Millennium Ecosystem Assessment, as well as regional programs of adaptive ecosystem management, have used scenarios to embrace diverse models and perspectives. The hope is that fair decisions can emerge if all perspectives and biases are represented and the data and models are transparent and widely shared. By addressing a wide range of viewpoints and models, these processes evaluate information and uncertainty in ways directly pertinent to the social issue. An inclusive political process shapes the scientific assessment.
The harmonization of politics and science is an infinite game, always evolving and never evolved. Any game can involve good and bad moves. Useless Arithmetic is about a type of bad move, in which models are used in politics to overstate certainty and thereby achieve the goals of a narrow interest group. Scientists and engineers need to understand this challenge and help to avoid it. This engaging introduction from Pilkey and Pilkey-Jarvis is a good place to start. The text is clear, direct, and the right length for an airplane trip. Useless Arithmetic should be read widely, but readers will need to look to other sources to learn how institutions and politics can use science more appropriately to improve the general welfare.
Stephen R. Carpenter (email@example.com) leads the North Temperate Lakes Long-Term Ecological Research Program at the University of Wisconsin–Madison and served as co-chair of the Scenarios Working Group for the Millennium Ecosystem Assessment.