July 1997 Workshop
10-12 July 1997, Boulder, Colorado, USA


Project Overview

Major financial and intellectual resources in the earth sciences are focused on trying to predict natural and human-induced environmental phenomena. Such efforts reflect a demand by policy makers for predictive information that can help guide political decision-making on controversial environmental issues that include negotiation of international environmental treaties, environmental impacts of wastes, and control of development in areas prone to natural disasters. While policy makers and scientists view predictions from different perspectives, neither policy makers nor scientists possess explicit, systematic guidance for understanding if, how, and when the results of research focusing on prediction can be productively applied to the policy-making process.

Whereas timely, policy-relevant predictions may help policy makers respond to some environmental problems, the misapplication of prediction research to policy problems can undermine policy goals, waste scarce resources, and undermine the credibility of the scientific enterprise. This project on the use and misuse of prediction in the earth sciences involves two workshops (the first of which is reported herein) that bring together scientists, policy makers, and policy analysts, to develop, present, and integrate case histories in predictive earth science research (past and ongoing). The workshops will develop, debate, and propose usable principles and criteria that can help policy makers judge the potential value of scientific prediction for different types of political and social problems related to the environment, and thus contribute to the design of science and environmental policies that are fiscally responsible, scientifically efficient, and socially constructive. Towards this end, a significant component of this project will be the dissemination of workshop results to the relevant scientific and policy-making communities through publications, presentations, etc.

Project Goals

As environmental problems become more pervasive and severe while budgets for research become tighter, the importance of effective prioritization and allocation of scarce resources will increase, as will the need for timely and effective decision-making. The Geological Society of America (GSA) and the Environmental and Societal Impacts Group at the National Center for Atmospheric Research (NCAR) have initiated a project whose principal goal is to develop guidelines for the beneficial use of scientific prediction in the policy process, especially in the area of environmental policy problems. As a point of departure for further discussions and practical testing, we propose to develop a set of criteria and an analytical framework to help policy makers assess the potential value of scientific prediction for different policy problems. In developing such criteria, a complementary goal is to engage both scientists and policy makers in constructive dialogue about the role of predictive science in the policy process, and to foster more realistic expectations among both groups as to the potential contribution of prediction to policy making. Usable assessment criteria and realistic expectations can contribute to the design of science policies that are fiscally responsible, scientifically efficient, and socially constructive.

A Taxonomy of Scientific Prediction in the Earth Sciences

We use the word "prediction" to mean a scientific effort to determine the timing and characteristics of future events. Such efforts may be deterministic ("the sun will rise tomorrow at 6:32 a.m.") or probabilistic ("there is a 20 percent chance of rain tomorrow"). Predictions can be based on physical understandings ("a rock will drop because of gravity") or statistical understandings ("every time I let go of the rock, it falls"). In this project we are interested in the capacity of scientific predictions to improve policy outcomes and thus contribute to the welfare of society.

How scientists justify their claims on public funds in support of predictive research is directly linked to the expectations held by policy makers and the public about the role predictions can play in the resolution of societal problems. Typically, scientists and policymakers value predictions for different reasons:

A. Prediction for Science

Prediction is a test of scientific understanding. When the expectations of researchers coincide with observable events in the laboratory or in nature, it lends support to the power of scientific understanding to explain how the world works.

Prediction is central to the process of science -- it is fundamental to the scientific method. Scientists test their ideas by comparing predictions based on theory to actual events in nature or the laboratory. In such efforts the experimental situation must be carefully controlled or the observational situation carefully characterized to ensure that the theory is applicable. This methodology permits the elucidation of invariant--and thus predictive -- principles of nature, such as Newton's laws or Maxwell's equations.

In recent years, however, scientists have begun to pursue a different type of prediction. In addition to seeking to deduce fundamental laws of nature, sophisticated numerical models and/or suites of observational data are integrated to predict behavior of complex natural phenomena. The intended outcome of this type of prediction is to foretell the future behavior or evolution of the natural phenomena as a test of understanding. Such predictions are made possible by rapid advances in computer and data acquisition technologies. The scientific goals of such activities in the earth sciences are to better understand fundamental earth processes (such as heat exchange between the atmosphere and oceans). In many cases, the scope and complexity of the natural phenomena predicted make adequate characterization or control difficult.

B. Prediction for Policy

As decision makers debate alternative courses of action, such as the need for a new law or the design of a new program, they are actually debating the expected outcome of this law or program, and its future impact on society. Thus, from a policy perspective, there is an expectation that predictions will directly stimulate and enable beneficial political or societal action. Such expectations have a long history, predating modern science. For instance, armed with reliable knowledge of the coming flood, Noah was willing to build the ark and thus able to avoid the catastrophic end that befell those without such foresight. Similarly, with a reliable forecast more than 750,000 people were able to successfully evacuate from the path of hurricane Andrew in 1992.

C. Prediction for Science and Policy

In the past, many predictive activities central to science (i.e., hypothesis testing) were largely distinct from the predictive needs implicit in policy making (i.e., that a given policy decision will lead to a desired outcome). The rise of science aimed at predicting complex natural phenomena, has contributed to increasing overlap between the desires of scientists and the needs of policy makers: scientists desire to understand the complexities of nature and these natural phenomena are often near the center of important policy issues. The justifications presented by scientists and policy makers for public support of science reenforce this overlap. Today, predictive activities are frequently justified as a means to serve goals of science and policy.

Of course, when predictive research is justified in terms of its contribution to societal goals, no matter how indirectly, it has moved from "prediction for science" into the area of overlap. In today's environment of science policy it is likely that most predictive research is justified in terms of science and policy goals. This can lead to a tension between the desires of science and the needs of policy. This tension and its effects on the policy process and on policy outcomes forms the underlying motivation for this project.

different justifications used to secure public support for predictive research

The figure on this page illustrates conceptually the different justifications used to secure public support for predictive research. Prediction can be justified as exclusively a scientific activity (e.g., hypothesis testing to advance knowledge), exclusively a policy activity (e.g., hurricane landfall predictions to support evacuation decisions), or a combination of the two (e.g., better models of atmospheric circulation will lead to a better understanding of climate and support the needs of policy makers responding to climate change). Our focus in this project is on predictions developed for and justified by application to policy, and not on purely scientific applications, i.e., not on predictions internal to science and justified in terms of knowledge for knowledge sake. Thus, our area of interest is represented by the shaded area in the above diagram.

The Policy Problem

This situation sets the stage for a policy problem. Ironically, we lack the knowledge needed to anticipate -- to predict, if you will -- the circumstances in which predictive research can contribute to effective decision making. As a consequence, some of our environmental policies may rely inappropriately on predictions and thus run the risk of failing to achieve their intended effects. A problem exists in that we lack the evaluative capabilities to systematically assess which environmental issues might or might not be amenable to solution aided by predictions, and how best to assimilate predictions in the policy process once they have been made.

The use of prediction to test and advance our scientific understanding of nature has been vindicated throughout the history of science, and is a foundation for the authoritative status of scientific knowledge in modern society. The more recent use of prediction in science -- to contribute to societal goals through the foretelling of the future -- has not been subjected to systematic testing and evaluation with respect to policies that depend upon them, but is legitimated by the success of traditional predictive science.

Toward a Better Understanding of Prediction and Policy

The predictive capacity of science holds great intiutive appeal for policy makers who are grappling with complex and controversial environmental issues, by promising to enhance their ability to ascertain the need for and outcomes of particular policy actions. From a scientist's perspective, presenting policy makers the possibility of better understanding the future consequences of alternative courses of action provides a complelling justification for the support of scientific research. From a policy-maker's perspective, predictive science can improve policy outcomes by guiding rational policy choices, and in the process reduce the need for divisive debate and contentious decision-making based on subjective values and interests. In other words, in offering to improve policy outcomes, scientific predictions also offer to reduce political risk. For policy makers worried about public support and reelection, avoiding political risk is very appealing indeed. This appeal can be carried one step further, as when the very process of scientific research aimed at prediction is portrayed as a positive step toward solving a policy problem. Politicians may therefore see the support of research programs that promise to deliver a predictive capability in the future as an alternative to taking politically risky action in the present.

Supply and demand for federally funded research on prediction of environmental phenomena are tightly coupled. As environmental problems become more complex (technically, politically, scientifically, etc.) -- in a context where potential solutions are controversial and costly -- decision makers look toward scientists to help reduce uncertainties and dictate "rational" policy paths. Simultaneously, the growing analytical and computational sophistication of the earth sciences leads to an increased confidence in the capacity of these disciplines to predict the behavior of the environment. Furthermore, in a period of constrained federal research funding, decision makers and scientists naturally converge on areas of research that are expected to be mutually beneficial (i.e., the area of overlap in the figure on the previous page).

For both scientists and politicians the short-term benefits are clear: scientists receive federal funding to develop predictions; politicians can point to predictive research as "action" with respect to societal problems, while deferring difficult decisions as they await the results of research.

Over the long term, how well will this arrangement lead to improved policy making, and beneficial outcomes with respect to environmental problems? Prospects for success will certainly vary depending on the phenomenon being predicted and the policy problem being addressed. An analytical framework that allows policy makers and scientists to rigorously evaluate such prospects would be extremely useful. In particular, the capacity of predictive research to contribute to positive policy outcomes must be evaluated in light of the following concerns:

  • Phenomena or processes of direct concern to policy makers may not be easily predictable on useful time or spatial scales;

  • Accurate prediction of phenomena may not be necessary to respond effectively to political or socioeconomic problems created by the phenomena;

  • Predictive research may reflect discipline-specific scientific perspectives that can rarely provide "answers" to policy problems that comprise complex mixtures of facts and values, and which are typically perceived differently by different policy makers;

  • Necessary and/or feasible political action may be deferred in anticipation of predictive information that may not be forthcoming in a useful time frame; similarly, such action may be delayed when scientific uncertainties associated with predictions become politically charged;

  • Predictive information may be subject to manipulation and misuse, because the limitations and uncertainties associated with predictive models are often not readily apparent, and because the models are often applied in a climate of political controversy and/or high economic stakes. This may be particularly problematic when predictions are used to justify government regulatory decisions;

  • Emphasis on predictive sciences moves both financial and intellectual resources away from other types of actions, including other research, that might better help to guide decision making.

These and related concerns suggest that the usefulness of scientific prediction for policy making and the resolution of societal problems depends on relationships among numerous factors, such as the time scales under consideration, the ability of the policy process to define and formulate a "problem" that is capable of "solution," the scientific complexity of the phenomenon being predicted, the political and economic context of "the problem," and the availability of alternative scientific and political approaches to "the problem."

From a policy perspective, the challenge is to develop a useable framework that allows decision makers to understand and address these three questions for a wide range of environmental phenomena and problems. In other words, the value of predictive research as a policy tool will be determined within a complex scientific and a policy context. Defining and understanding this context is a crucial prerequisite for the design of environmental science policies that are fiscally responsible, scientifically efficient, and socially beneficial.

Back to Workshop Home Page
Back to Predictions Home Page