Special Roundtable Section976 BioScience • November 2005 / Vol. 55 No. 11No human endeavor completely escapes subjec-tivity, but science tries hard to achieve this ideal in order to substantiate claims that it has arrived at universallyvalid knowledge. It does this first of all by using methods andprocedures intended to free the investigative process fromhuman influences and frailties. These methods and proce-dures include mathematics, logic, nonhuman instrumen-tation, standardized and quantitative measurement,operational definitions, and replication. As valuable as thesetools are, scientists’ methods and their claims to knowl-edge nevertheless rest on an irreducible element of judgmentand subjectivity. This is why mathematical proofs must bechecked for error, instruments must be calibrated by humanobservers, measurements must be rechecked, definitionsmust be evaluated for their appropriateness, new studiesmust be judged as to whether they truly replicate or con-tradict old ones, and so forth. Science, despite its famous em-phasis on achieving objectivity by eliminating human error,can make its claims of objectivity only because it relies on the subjective judgments of fallible human beings and social institutions to detect and correct errors made byother fallible humans and institutions.I comment on this point not only because it is an interestingparadox for epistemology and philosophy of science but alsobecause of its importance for informing practical decisions.I focus here on understanding human–environmental systems.The effort to understand these systems reveals quite clearly theways that claims of scientific knowledge rest fundamentallyon subjectivity, judgments, and values. They reveal this withevery dispute over risk assessment, cost-benefit analysis, andother techniques used to inform decisions. I do not con-clude that there is no hope of basing claims of environmen-tal knowledge on well-founded information. There is a strategythat can be used to organize subjectivities to provide a solidenough base of understanding to support well-informed andsocially legitimate decisions. The strategy relies on adaptingtechniques for managing subjectivity that have been developedthrough long practice in two disparate human endeavors: sci-ence and democracy. These techniques promote and governprocesses of deliberation that communities use to seek truthand, in particular, to understand complex systems so that theycan make informed choices while confronting multiple andsometimes conflicting social objectives.When scientific understanding requires deliberationLet me begin by identifying a continuum of states of theworld, arrayed according to how strongly claims of knowledgeabout them can be supported. At one end of the continuumis an ideal type of situation about which absolute knowl-edge is achievable and confirmable by methods that escapesubjectivity. This ideal type does not exist in the real world,Paul C. Stern (e-mail: [email protected]) works for the National ResearchCouncil, Washington, DC 20418. © 2005 American Institute of Biological Sciences.Deliberative Methods forUnderstanding EnvironmentalSystemsPAUL C. STERNEnvironmental problem solving needs science but also inevitably requires subjective judgment. Science can help in dealing with subjectivity, becausescientists have long experience developing institutions and practices to address the subjective and value-laden choices that are essential to scientificprogress. Democracy has also developed approaches to the problem. The underlying principles can be applied to environmental policymaking. Thisarticle explores these issues in the context of decisions about environmental risks, drawing on the work of the National Research Council and othersources. It suggests some guidelines for risk deliberation—including broad-based participation, commitment to scientific quality, explicit attentionto values, transparency of deliberative processes, and rules for closure and reconsideration—and recommends that an experimental approach be employed to learn how best to use deliberative methods.Keywords: deliberation, decisionmaking, democracyNovember 2005 / Vol. 55 No. 11 • BioScience 977because there is a core of subjectivity in all observations anda core of judgment in all inferences from experience. How-ever, in the fields that most closely approach the ideal, the areas of subjectivity are relatively small and noncontentious.For example, experiments give varying estimates of the speedof light, and scientists’ judgments about the true value of thisconstant have often fallen outside the error bounds establishedby later observations (Henrion and Fischhoff 1986). Thediscrepancies in measurement and judgment, however, aresmall and decreasing over time, and there is no fundamen-tal disagreement about what is being measured or aboutwhich methods are appropriate for measuring it.At the other end of the continuum is a situation in whichthe only knowledge is subjective. Individuals’ fantasies maybe phenomena of this type, because although we agree thatother people have fantasies, we have no objective way toknow them. Science normally works to transform phenom-ena from being more like the second ideal type to beingmore like the first. Thus, scientists may try to measure fan-tasies by recording and analyzing verbal descriptions of themor by observing the brain activity of people while they fan-tasize.Many important practical decisions, including those ofenvironmental policy, require understanding of phenomenathat lie far from the first ideal type and that stubbornly resistthe efforts of science to bring subjectivity within acceptablebounds. These decisions need scientific input because humanchoices set in motion processes that conform to natural laws:decisionmakers need to understand the implications wellenough to choose actions that are likely to achieve the resultsthey desire. But science consistently falls far short of deliver-ing certainties for decisionmakers. Consider a few tough pol-icy choices: regulating environmental carcinogens, deployingmilitary force to stabilize critical regions of the world, and set-ting interest rates to promote economic growth while con-trolling inflation. For such choices, policymakers often wantmore certainty than science can provide. Science falls shortnot only because the phenomena are complex but also becausereasonable people disagree about which information is mostneeded to understand the choices