Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
1 The Idea of Risk Characterization During the past several decades many areas of government policy associated with hazards to health, safety, and the environment have become increasingly contentious. De- spite much new legislation and extensive efforts by the agencies charged with implementing the legislation, dissatisfaction and controversy con- tinue. A continuing debate on regulatory reform has not yet reached consensus on how governmental institutions and procedures should be structured to make decisions better and more broadly acceptable. Many believe that increased use of risk analysis1 is appropriate. The expecta- tion that clear and concise characterizations of existing information about risks, costs, and benefits will lead to informed and acceptable regulatory decisions is attractive; it may, however, be naive. One reason lies in inadequacies of the techniques available for risk analysis. A second is the fundamental and continuing uncertainty in information about risks. An- other, less well appreciated reason lies in a basic misconception of risk characterization and its relation to the overall process of comprehending and dealing with risk. Risk characterization involves complex, value- laden judgments and a need for effective dialogue between technical ex- perts and interested and affected citizens who may lack technical exper- tise, yet have essential information and often hold strong views and substantial power in our democratic society. See Glossary for the terms used in this volume. 11
12 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOCRATIC SOCIETY We believe the iterative analytical-deliberative process described in this volume holds much promise for improving risk characterization, in- forming decisions, and making those decisions more acceptable to inter- ested and affected parties. The technical and analytical aspects of risk analysis must be balanced with a concern for appropriate involvement by interested and affected parties in all steps of the decision-making process, including those leading to risk characterization. Analysis and citizen involvement are not separate steps to be carried out in sequence, but must be combined into an effective synthesis. Our approach involves a substantial change from the formulation of risk analysis that many fed- eral agencies and other organizations have been using for more than a decade. Many groups before us have studied aspects of risk decision-making processes in order to improve decisions. We undertook this study to evaluate and make recommendations about risk characterization, de- scribed in the committee's initial task statement as the part of the decision process at which "the information in a risk assessment is translated into a form usable by a risk manager, individual decision maker, or the public." Stating the committee's charge in this way highlights a central dilemma of risk decision making in a democracy: detailed scientific and technical information is essential for understanding risks and making wise deci- sions about them, yet the people responsible for making the decisions and the people affected by the decisions and who may therefore also take part in them are not themselves expert in the relevant science and technology. This dilemma has spawned numerous technical and policy attempts to resolve it. It is important to recognize, however, that in many areas of hazard management, the dilemma has not posed much of a problem. For example, airline and automotive safety have improved fairly continu- ously over long periods, and the public has trusted the responsible insti- tutions to monitor and maintain safety, even though few citizens under- stand the technologies. The same has been generally true for the safety of foods and drugs, earthquake engineering, and the issuance of some rou- tine environmental permits. In other areas, however, the dilemma has been stark. In the management of radioactive waste, toxic chemicals, and hazardous industrial facilities, for instance, technical experts and the re- sponsible agencies are mistrusted by many of those who participate in risk decisions. Our concern is especially focused on decisions in which, as in the latter set of cases, participants are likely to come into conflict about the adequacy of scientific knowledge; about issues of fairness, access, and consent in the decision process; or about basic goals and values. Such decisions are relatively few in number, but usually great in importance. Moreover, decisions that have not aroused this sort of contention in the past could do so in the future.
THE IDEA OF RISK CHARACTERIZATION 13 We have focused narrowly on one aspect of risk decision making- the problem of characterizing risk so that better informed decisions can be made but in other respects, we have taken a very broad view. First, we have not restricted ourselves to a particular kind of hazard or risk. Thus, the issues we discuss are relevant whether the goal is to understand the risks of cancer to humans, noncancer health risks, or risks to ecological, social, or political systems. We believe the principles we develop are applicable over a very broad range of hazards and risks.2 Second, we do not restrict our discussion to particular kinds of orga- nizations that may engage in characterizing risks. For specificity, we sometimes write as if the responsible organization is a federal, state, or local government agency, and in some places, the language may seem to be addressed even more specifically to federal regulatory agencies. We certainly intend that what we say will meet the needs of such organiza- tions. We believe that it also has wider applicability to public health organizations, industrial organizations, nonprofit organizations, and oth- ers who prepare descriptions of risks to health, safety, or the environ- ment. It may be, however, that adversarial settings such as courts do not place the same demands on participants because those settings provide their own mechanisms for including different perspectives. Third, we find that in order to improve risk characterization, one must consider other parts of the risk decision process, particularly the various analytic activities that provide the information used in character- izing risks. The purpose of risk characterization is to improve under- standing of risk, and everything that goes into such understanding is necessary for effective risk characterization. We develop this theme in detail in this chapter. Finally, we offer a strategic approach and a set of principles for a better understanding of risk, rather than a set of guidelines or procedures that can be applied in a routinized way for all risk situations. Our recom- mended strategy implies the need for very extensive effort and expense for characterizing some risks, but relatively little in most instances: the level and kind of effort required are highly situation-specific. Thus, it is imperative for organizations engaged in characterizing risk to consider the situation carefully at the outset and to build flexibility into their char 2Similarly, the issues we discuss are relevant regardless of whether decisions are nor- mally based on formal risk estimates (as is typically the case with cancer risks to humans); surrogate measuress, such as a dose observed to have no significant effect in laboratory animal studies, modified by a margin of safety (as is commonly done with noncancer health risks to humans); or made without reference to standard techniques of risk estimation or standard decision-making rules (as with many risks to ecological, social, and political sys- tems).
14 UNDERSTANDING RISK: INFORMING DECISIONS INA DEMOCRATIC SOCIETY acterization procedures to allow for unexpected situations. It is no doubt possible to devise more detailed guidelines, consistent with our prin- ciples, for risk characterization in particular kinds of situations. We be- lieve that the most useful contribution we can make in this volume is to propose a broadly applicable strategy and conception that allows organi- zations to deal with some of the most prevalent and serious challenges of risk characterization. BEYOND TRANSLATION The committee's initial charge presumed that the key to resolving the dilemma is to "translate" scientific knowledge into a form usable by deci- sion makers. We agree that it is essential to make science useful to deci- sion makers, but we have concluded that doing this involves much more than translation: it requires a different view of the purpose and form of risk characterization. Risk characterization has typically been seen as a summarization of scientific information. This understanding is succinctly stated in Risk Assessment in the Federal Government: Managing the Process (National Re- search Council, 1983:20), widely referred to as the Red Book: Risk characterization is the process of estimating the incidence of a health effect under the various conditions of human exposure described in exposure assessment. It is performed by combining the exposure and dose-response assessments. The summary effects of the uncertainties in the preceding steps are described In this step. Risk characterization is seen here as the final step in the process of risk assessment. It combines the results of a completed hazard identification, exposure assessment, and dose-response assessment into a concise esti- mate of adverse effect in a given population. Figure 1-1, taken from the Red Book, schematically represents the traditional view of risk character- ization and its relations to the other elements of risk decision making. In this schema (National Research Council, 1983:28), risk characterization involves "no additional scientific knowledge or concepts" and only a minimal amount of judgment. This view remains prevalent in federal agencies. For example, a 1992 memorandum by a deputy administrator of the U.S. Environmental Protection Agency (EPA) clearly defines risk characterization as a summarization or translation process coming at the completion of a scientific analysis (Habicht, 1992~. It was reaffirmed by the EPA in a 1995 policy statement using almost identical language (Browner, 1995~. It has also been applied, with some modification, to the characterization of ecological risks (U.S. Environmental Protection Agency, 1992a). And it is by no means confined to a single agency of the
15 z 5 LU z ~: y ~n 5 ~n ~n ~n ~n y ~n ~n ~ :) - - ~ o .~ z - - w ~_ w o C: o ._ ~n . _ o ._ a) =5 1 1 ~ ..~.. . - U) C~ U) V) .' , CD o ~ C%. i~ ~ C' ~ ~o ., a) ~ a a - s a) ~:5 ~ > cn =5 ~ a) ~ I ~ s _ ~ a) cn > a' ~ ·_ (t ~o cn ~ o o ~ 0 a) Q cn Q o ~ Q ~ s a) U) ~ o <: ~n a) Q s ~n ~ a) ~ CE a cn s o ~ ' au y q) C: . _ a ~n o C~ CO ~5 a' ~ ~s a) Q .= <1) o ~ ° ~ °~ CO o ~n ~ ~ ~ a) Q S O ~ 0) X ~ ~ ~ ~_ U ._ U) a' ~4 U) ·_4 sO O E a '_ w E ~ ~ C .= Q O a) cn a' O ~ O ~ t~ Q q) a) x o) s a) ~ _ tl:) ~ O o 3 a) N ~ 25 '~ a) a) a =5 ~ ~ .o ~n s ~ a~ ~ ~n o ._ - Q o Q ~_ 00 C~ . . 00 - ._ U ,- O ~ ~U r~ cn `_ ~
16 UNDERSTANDING RISK: INFORMING DECISIONS INA DEMO CK TIC SOCIETY government (see, e.g., the special Jurre 1994 issue of Risk Analysis on "The Risk Assessment Paradigm After Ten Years". We have concluded that the view of risk characterization as a sum- mary is seriously deficient, and we propose a more robust construction. Risk characterization must be seen as an integral part of the entire process of risk decision making: what is needed for successful characterization of risk must be considered at the very beginning of the process and must to a great extent drive risk analysis. If a risk characterization is to fulfill its purpose, it must (1) be decision driven, (2) recognize all significant con- cerns, (3) reflect both analysis and deliberation, with appropriate input from the interested and affected parties, and (4) be appropriate to the decision.3 The rest of this section describes and illustrates these four facets of risk characterization and the risk decision process. A Decision-Dri`ren Activity The purpose of risk characterization is to enhance practical under- standing and to illuminate practical choices. A carefully prepared sum- mary of scientific information will not give the participants in a risk deci- sion the understanding they need if that information is not relevant to the decision to be made. It is not sufficient to get the science right; an in- formed decision also requires getting the right science, that is, directing the scientific effort to the issues most pertinent to the decision. In 1994 the EPA completed a $6-million scientific reassessment of the health risks of dioxin (U.S. Environmental Protection Agency, 1994a), undertaken to resolve a major controversy about the dose-response rela- tionship between dioxin exposure and possible human health effects, par- ticularly cancer. Under the working assumptions of a linear dose- response relationship for cancer, existing data imply that one form of dioxin, TCDD, is one of the most highly carcinogenic of all chemicals. But analyzing these same data under an alternative hypothesis, dioxin could be much less dangerous. EPA's 96-page risk characterization attempts to synthesize an estimated $1 billion worth of scientific research and a 2,000- page reassessment document. Yet the characterization has not resolved the scientific issues (e.g., Clapp et al., 1995; Environ Dioxin Risk Charac 3This idea of risk characterization is the crystallization of much thinking about risk over at least two decades. Much of it is implicit in the common belief among practitioners of risk analysis that it is imperative to engage in repeated interaction with the client; we simply add that risk characterizations have many clients. Intellectual antecedents for many of our ideas can be found in earlier work (e.g., Morse and Kimball, 1951; Howard, 1966, 1968; Morgan, 19&1; National Research Council, 1983, 1989, 1994a; Edwards and von Winterfeldt, 1987; Keeney, von Winterfeldt, and Eppel, 1990; von Winterfeldt, 1992).
THE IDEA OF RISK CHARACTERIZATION 17 terization Expert Panel, 1995), and its authors and critics agree that a substantial additional research effort will be necessary if this is to be done. To what extent would an improved dioxin risk characterization be useful for making decisions, even if scientists could agree on it? Would it be the right science? Among the most likely present uses for such a risk characterization are to inform decisions concerning operating permits for municipal and industrial facilities, siting waste incinerators, making policy decisions about Vietnam veterans exposed to dioxin, and reme- diating Superfund sites. Yet in all of these contexts, dioxin is only one of many hazardous chemicals involved and cancer is only one of many out- comes of concern, so dioxin-induced cancer is at best only part of the problem and a dioxin risk characterization, though relevant, can only hope to provide some of the information needed for the decision. More- over, one effect of exclusive or intense attention to quantifying the dioxin- cancer link and its uncertainty may be to draw attention away from other risk-related concerns, which may be more important to participants in the decisions and may require different kinds of analysis. These concerns might include questions about the fairness of exposing a community that may have an abundance of toxic chemical sites to yet another site, about whether the local population has characteristics that make it unusually susceptible to damage from an additional body burden of dioxin, or about effects of the contemplated action on local property values. Some of these (or other) issues may be the most important ones for a particular decision. A risk characterization focused solely on scientific questions about the dose-response relationship of dioxin to cancer may be highly unsatisfac- tory to some people because it is only marginally relevant to their most serious concerns. Since the 1950s, national policy on the disposal of high-level radioac- tive waste from the civilian nuclear power industry has been to store the wastes permanently in deep, underground repositories. Billions of dol- lars have been spent on studies to characterize the risks of leakage of radioactive materials from proposed repository sites into the environ- ment. Much recent effort has been devoted to what is currently the only proposed high-level repository site, at Yucca Mountain, Nevada. The analyses have repeatedly been criticized on technical grounds, and new technical objections have led to new studies. Although the studies have convinced many members of the technical community that the site is potentially acceptable, most Nevada citizens and many others remain unconvinced. Some of the objections probably result from disagreement about what decision needs to be made. A major gap in the argument to use the Yucca Mountain site is the lack of a convincing case that a permanent repository is needed now for
18 UNDERSTANDING RISK: INFORMING DECISIONS INA DEMOCRATIC SOCIETY environmental, safety, or health reasons. Surface storage has been judged by the Nuclear Regulatory Commission to be acceptably safe well into the next century, and "no comprehensive appraisal is now available of the probable costs and risks of continuing the present temporary waste dis- posal practices" for decades more in comparison with the risks and costs of putting the waste in a permanent repository (National Research Coun- cil, 1995:13~. In the Yucca Mountain case, the government's risk charac- terizations seem to have relied too much on only one subset of scientific information, presuming that it was obvious which question needed to be answered. Opponents were concerned with a different set of issues, which were not addressed in the risk analyses, such as the fairness of placing nuclear waste in a region that does not have any nuclear power plants and is already host to the nation's nuclear testing facility (several such concerns are listed in National Research Council, 1995:21-23~. Conse- quently, to many people, characterizations of the Yucca Mountain site are at best irrelevant, and at worst dangerously misleading because they fo- cus attention on the wrong question. A contrast to these two examples comes from the Man and Biosphere (MAB) Program organized by the U.S. Department of State as part of a larger international effort. In an MAB activity over several years, more than 100 natural and social scientists from various federal and state agen- cies and from universities have considered policy options for managing surface water so as to maintain a sustainable ecosystem in and around Florida's Everglades (Harwell et al., in press). Changes in the ecosystem and possible responses to them entail risks to endangered species, to drinking water quality in nearby metropolitan areas, and to the liveli- hoods of sugar growers. The scientists considered all these risks care- fully, but from a perspective different from that typical in risk assess- ments. They defined the problem not as one of estimating and reducing risks, but as one of developing a shared vision of desired conditions of the ecosystem. They then identified development strategies consistent with such a vision and proposed governance structures that could adaptively manage the social-ecological system as it changed and as new knowledge developed. They considered several scenarios for change in human man- agement of the ecosystem and analyzed them in terms of their compatibil- ity with goals of sustainable economic and social development and with a widely shared vision of ecosystem use. The MAB effort is noteworthy for its problem-driven approach, particularly its extensive and explicit efforts to understand the decisions to be made, rather than presuming that deci- sion makers would gain the understanding they needed from estimates of the ecological, health, and economic costs and benefits of previously de- fined choices. In fact, the process generated policy options that had not
THE IDEA OF RISK CHARACTERIZATION 19 previously been considered and that might be more acceptable, both socially and ecologically, than any that might otherwise have been con- sidered. (A more detailed description of the MAB activity is in Appendix A.) Recognizing All Significant Concerns The people who participate in risk decisions public officials, experts in risk analysis, and interested and affected parties may be concerned with a variety of possible harms or losses. Sometimes, risks to social, ethical, or ecological values are at least as important as risks to health and safety. The analysis that will be the basis for a risk characterization must pay explicit attention to the breadth of the significant issues. This is often best done by involving the spectrum of decision participants explicitly in formulating the problem to be analyzed. In recent years a number of states have organized "comparative risk" projects to develop strategies for setting priorities for environmental pro- tection efforts, based on a ranking of risks. Building on the experiences of other states, California began its project in 1992 by providing for input from a wide variety of citizens, which led to considerable elaboration of the issues the project addressed. For example, at the request of partici- pants in the process, an environmental justice committee was created, and it raised some fundamental questions about risk ranking as a strat- egy. Arguing that risk-based ranking gives insufficient emphasis to com- munity participation, pollution prevention, and the disproportionate risk burdens borne by some communities, the committee proposed giving these three concerns a more important place. The comparative risk project responded by paying increased attention to the distribution of risks to human health and welfare in its analysis. This process to the satisfac- tion of some and the consternation of others initiated broader statewide debate about the goals of environmental policy (California Environmen- tal Protection Agency, 1994; Stone, 1994; also see Appendix A). Another example of the need to recognize all significant concerns comes from East Liverpool, Ohio, where a hazardous waste incineration project has been controversial for over a decade. Risk assessment studies conducted before, during, and after the construction of the facility have not convinced a number of local constituencies, including public health officials, to accept the incinerator. A series of recent test burns, designed to reduce uncertainty about the risk, only increased the controversy. An underlying cause of the opposition was that people were concerned with issues not addressed in the risk assessments. Some of these, including risks during start-up and shut-down and waste transportation, were even- tually incorporated in risk analyses, but several other concerns never
20 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY were. One of these involved the broader policy issue that approving a waste incinerator might encourage increased production of hazardous waste. Another was that delaying definitive risk analysis until test burns were performed might permit so much investment in construction and testing that government could not then refuse the incinerator, regardless of the results of the risk assessments. Because of this concern, some of the neighbors distrusted the entire risk assessment process. A third concern was with the physical effects of pollutants on the local population, which was exposed to various other industrial emissions. A fourth was with the adequacy of risk assessment methodology for making public health deci- sions: the local public health department was not satisfied with the "theo- retical assumptions" in the risk assessments and wanted decisions to be based on ongoing monitoring of air, soil, and crops. (The East Liverpool incinerator case is described in more detail in Appendix A.) An Analytic-Deliberative Process Improving risk characterization requires attention to two discrete but linked processes: analysis and deliberation. Analysis uses rigorous, replicable methods developed by experts to arrive at answers to factual questions. Deliberation uses processes such as discussion, reflection, and persuasion to communicate, raise and collectively consider issues, increase understanding, and arrive at substantive decisions. Deliberation frames analysis and analysis informs deliberation. Thus, risk characterization is the output of a recursive process, not a linear one. Analysis brings new information into the process; deliberation brings new insights, questions, and problem formulations; and the two build on each other. The analytic- deliberative process needs input from the spectrum of interested and affected parties. Four recent cases provide examples of this process. In 1992 and 1993 the Environmental Protection Agency sponsored a negotiation to set maximum concentration levels for the by-products of chlorine and other disinfectants used to eliminate microbial contamina- tion of drinking water. The agency had signed a consent order setting a 1994 deadline for proposing a rule, but it subsequently concluded that any rule it could propose by that date would be vulnerable to legal chal- lenge because of lack of data. In an attempt to avoid long delays in litigation, EPA invited the interested and affected parties to join it in negotiating a rule. The negotiating group relied on a technical advisory committee to analyze the risks and on processes of dialogue and persua- sion to try to reach agreement on a rule. The negotiators soon ran up against a key information gap: they had little information about the quality of untreated water in different re- gions, so they could not adequately characterize the risks from microbial
THE IDEA OF RISK CHARACTERIZATION 21 pathogens in water left untreated or the level of risk likely to remain alter disinfection. Without reliable estimates of these and other factors, they could not be confident about making a rule. The composition of un- treated water was important to some of the interested and affected parties because they wanted to consider rules allowing non-chlorine disinfectant technologies. To consider the feasibility and wisdom of using such tech- nologies would require additional analysis, using new information. The negotiators dealt with the approaching regulatory deadline by proposing two rules: a provisional rule limiting disinfectant by-product concentra- tions and a rule requiring large public water supply systems to collect information on pathogens in source water, disinfectant by-products and their chemical precursors, and other matters. They agreed on a plan for reconsidering and possibly revising the limits in a few years on the basis of the new information collected, which they believed might well justify a revision of the rule. (A description of the regulatory negotiation for disin- fectant by-product concentrations appears in Appendix A.) Decades of experience with new drugs that came close to approval, only to be found to have unacceptable side effects, had made the Food and Drug Administration (FDA) cautious about approving new clinical trials of compounds that had not received extensive safety and efficacy testing in animals. The epidemic of AIDS (acquired immune deficiency syndrome) and the virus that causes it created strong demand for a faster approval process for new drugs. Although the FDA resisted at first, involvement of AIDS activists and that part of the medical community involved in clinical trials resulted in new, expedited protocols for certain situations. Reexamination of the risk situation led to recognition that some assumptions made for other drug approval decisions did not neces- sarily apply to drugs for AIDS. In 1989 the Florida Power Corporation began to search for a site for a new 2000-megawatt coal/gas-fired power generation station in its service area. Drawing on the knowledge and judgments of the corporation's technical and managerial staff and the knowledge and judgments of an external environmental advisory group, a consulting team constructed weighted lists of criteria for excluding potential sites. Most of the analysis was done by a consulting firm; most of the deliberation was done by the corporation's staff and the advisory group. Results of each of five rounds of deliberation instructed the consultants on which factors and weights to use in analyzing the potential sites. This iteration between analysis and deliberation was repeated, with the selection criteria becoming more de- fined each time, and more possible sites eliminated from the list. During the fifth and final phase, the six remaining sites were ranked, and one preferred and two alternate sites were selected. As of November 1995 the,
22 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY preferred site was in the licensing stage. (More detail on this example is in Appendix A.) A rabies outbreak among raccoons in New Tersey and the risk of its spread to humans prompted the creation of an Interagency Rabies Task Force that included not only representatives from several departments of state government and county government, but also a range of outside experts. The task force considered conducting a field trial of air-dropped oral rabies vaccine in the one unaffected area of the state to prevent the spread of disease to raccoons there, but it did not want to proceed without assurances of public acceptance because the test would expose the public to baits containing the vaccine and because property owners would have to be asked to include their properties in the test area. With a survey and public meetings, the task force elicited public concerns and found wide- spread support for the program. During the field trial, mailings and press releases kept the public informed of progress. Content analysis of news- paper articles and letters and studies of telephone inquiries and com- plaints also were used to gather data about public response to the drop. After the trial was completed, the Task Force conducted mail and tele- phone surveys and found broad public support not only for the rabies experiment, but also for the way in which the potential public concerns were handled (Pflugh, no date). Matching the Process to the Decision Different kinds of risk decisions require different kinds and levels of analysis and deliberation in support of risk characterization. For a series of similar risk situations, one might establish routines for risk analysis, characterization, and decision making that embody clear and consistent expectations about how the problem is defined, which options are to be considered, what kinds of evidence are to be considered, who is to partici- pate in the process, and so forth. For novel, complex, or highly controver- sial risk situations which often involve questions about major potential impacts and the equity of the distribution of risks and benefits routines are likely not to be satisfactory. It is likely to be necessary to develop unique procedures for characterizing risk in these situations. Some examples of procedures involving repetitive risk decisions are those for reapproving existing permits for discharge of pollutants from industrial plants, for testing new drugs prior to approval decisions, for issuing Remanufacturing approval for the industrial production of new chemicals, and for deciding whether to exclude an individual from re- ceiving a vaccine or giving blood. We do not mean to imply that all the current procedures for these and similar decisions are appropriate; only that it is often appropriate to develop standard procedures. In fact, situ
THE IDEA OF RISK CHARACTERIZATION 23 ations and knowledge change even for routine decisions, and standard procedures for risk analysis and characterization should be reevaluated from time to time. Risk situations need to be accurately diagnosed to determine whether existing standard procedures should be applied, whether new procedures need to be devised, whether additional information is needed to decide which approach to follow, and what extent and type of analytic and delib- erative effort may be needed to come to such decisions. In medicine, experienced clinicians use a combination of knowledge, experience, and judgment to make diagnoses. The situation is closely comparable for those who must diagnose a risk situation and prescribe the appropriate kinds and level of analysis and deliberation needed and the appropriate breadth of participation. PARTICIPATION AND KNOWLEDGE IN RISK DECISIONS In the framework we have outlined, risk characterization cannot suc- ceed as an activity added at the end of a risk analysis, but must result from a recursive process that includes problem formulation, analysis, and deliberation. Two essential aspects of that process are appropriately broad participation by the interested and affected parties and appropriate incor- poration of science. Rationales for Participation There are three compelling rationales for broad participation in risk decisions. They have been classified as normative, substantive, and in- strumental (Fiorino, 1990~. The normative rationale derives from the prin- ciple that government should obtain the consent of the governed. Related to this principle is the idea that citizens have rights to participate mean- ingfully in public decision making and to be informed about the bases for government decisions. These ideas are embodied in laws, such as the Administrative Procedure Act and the Freedom of Information Act, al- though these laws and their associated procedures have not always been implemented in ways that involved meaningful participation (e.g., Houghton, 1988; Kathlene and Martin, 1991; Lynn and Busenberg, 1995~. The substantive rationale is that relevant wisdom is not limited to scientific specialists and public officials and that participation by diverse groups and individuals will provide essential information and insights about a risk situation. As we show in detail in Chapter 2, nonspecialists may contribute substantively to risk characterization for example, by identifying aspects of hazards needing analysis, by raising important questions of fact that scientists have not addressed, and by offering knowl
24 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY edge about specific conditions that can contribute more realistic assump- tions for risk analyses. Nonspecialists may also help design decision processes that allow for explicit examination, consideration, and weigh- ing of social, ethical, and political values that cannot be addressed solely by analytic techniques, but also require broadly participatory delibera- tion. The instrumental rationale for broad public participation is that it may decrease conflict and increase acceptance of or trust in decisions by government agencies. Mistrust is often at the root of the conflicts that arise over risk analysis in the United States (see, e.g., Bella, 1987; English, 1992; Flynn and Slovic, 1993; Kasperson, Golding, and Tuler, 1992; Laird, 1989; Pijawka and Mushkatel, 1992; Renn and Levine, 1991; Slovic, Flynn, and Layman, 1991; U.S. Department of Energy, 1992~. A combination of psychological tendencies to notice, believe, and give more weight to trust- destroying than to trust-building information, and social factors, such as the tendency of mass media to favor bad news and of some special inter- est groups to encourage distrust to influence policy debates, make trust very fragile (Slovic, 1993a). Some observers have suggested that improv- ing risk analysis and characterization may have little practical effect on public policy without efforts to rebuild trust by improving participation (Kunreuther, Fitzgerald, and Aarts, 1993; Leroy and Nadler, 1993; Slovic, 1993a). Simply providing people an opportunity to learn about the prob- lem, the decision-making process, and the expected benefits of a decision may improve the likelihood that they will support the decision (Peelle, 1979; Peelle et al., 1983; Peelle and Ellis, 1987~. Even if participation does not increase support for a decision, it may clear up misunderstandings about the nature of a controversy and the views of various participants. And it may contribute generally to building trust in the process, with benefits for dealing with similar issues in the future. Role of Science Reliable technical and scientific input is essential to making sound decisions about risk. Scientific and technical experts bring indispensable substantive knowledge, methodological skills, experience, and judgment to the task of understanding risk. A few less obvious points are worth emphasizing about the role of scientific analysis in risk decisions. First, such analysis requires contribu- tions from many, diverse disciplines. In particular, risk analysis often needs the substantive and methodological expertise of the economic, so- cial, and behavioral sciences: for instance, effects on property values, tourism, scenic value, human population migrations, fairness, and public trust in government may be important outcomes of risk decisions, and
THE IDEA OF RISK CHARACTERIZATION 25 they are in some cases amenable to rigorous scientific analysis. Even health risks cannot be estimated accurately without a good understand- ing of the behavior of the individuals and organizations that control or are affected by hazardous substances or processes. Second, in addition to their specialized disciplinary knowledge, sci- entists bring a capacity to build systematic and reliable ways of analyzing and interpreting information about new situations. As already noted and as elaborated further in Chapter 2, nonspecialists sometimes have infor- mation and knowledge to contribute to the risk decision process. It is important to incorporate such knowledge in a valid scientific framework. Third, scientific analysis may not always be neutral and objective as a decision-making tool, even when it meets all the tests of scientific peer review. Good scientific analysis is neutral in the sense that it does not seek to support or refute the claims of any party in a dispute, and it is objective in the sense that any scientist who knows the rules of observa- tion of the particular field of study can in principle obtain the same re- sults. But science is not necessarily neutral and objective in its ways of framing problems. For example, analyses of the risks of drunk driving that highlight drivers' behavior as a cause of traffic fatalities draw atten- tion away from the equally significant factors of automobile and highway design; analyses of the cancer risks of industrial chemicals divert atten- tion from the possibly comparable risks from naturally occurring chemi- cals in foods (National Research Council, 1996~; and analyses of the risks of indoor air pollution draw attention away from the problems of ambient air pollution and vice versa. Similarly, analyses of the costs of environ- mental regulation often serve the policy arguments of the opponents of regulation, while analyses of the risks of unregulated activities bolster the arguments of the proponents of regulation. Each kind of analysis is ap- propriate by itself, but if the overall scientific effort is tilted too far toward only one of the legitimate formulations of a problem, it tends to yield biased understanding. Science is not necessarily neutral either, in its choices of assumptions. Analysis is compromised for decision-making purposes when it is based on assumptions about the conditions of hazard exposure that are known to be unreasonable by decision participants who were not consulted when the assumptions were selected. Even standard statistical assumptions can raise questions of bias. The assumption of the null hypothesis as used in risk analysis contains an implicit bias because it places a greater burden of proof on those who would restrict than those who would pursue a haz- ardous activity, presuming these activities are safe until proven other- wise. Evidence that science has been censored or distorted to favor par- ticular interested parties has long been a source of conflict over risk characterizations (e.g., Rosner and Markowitz, 1985; Lilienfeld, 1991~.
26 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY Chapter 2 discusses many ways in which judgments made in the course of risk analysis can undermine the quality of risk characterization, even when the analysis meets stringent scientific tests. Fourth, scientists may be in a specially powerful position to influence decisions because many hazardous substances or activities have non- obvious and delayed effects that can be uncovered and quantified only with highly technical methods. Without specialized skills, nonscientists may be at a disadvantage in trying to confirm or challenge scientists' claims or judgments. Fifth, science alone can never be an adequate basis for a risk decision. This point deserves special emphasis in the light of recent proposals for "risk-based" decision rules that could tie public risk decisions to stan- dardized technical procedures of risk analysis. Risk decisions are, ulti- mately, public policy choices. In principle, analysis of a set of alternative decisions could show which would produce the fewest deaths, the fewest new cancers, the fewest workdays lost to illness, or the least cost to a manufacturer under given circumstances, but it cannot tell how these different effects should be weighed in the context of the decision. No amount of analysis can determine whether cancer-incidence rates should be more important to society than the number of workdays lost, or whether preventing cancer should be more important than preventing reproductive disorders, or whether reducing the prevalence of environ- mental illness in a broad population should be more important than en- suring an equitable distribution of the risk across subpopulations or a reduction of risk to particular subpopulations (e.g., children, the elderly). No amount of analysis can tell whether a 30 percent lifetime cancer risk to one individual is better or worse than a 15 percent risk to two individuals. No amount of analysis can tell whether the loss of one more wetland is more important than the loss of ten jobs, or ten thousand jobs. Analysis can gather useful information about which tradeoffs citizens as individu- als would prefer, but scientists cannot and should not be expected to make decisions that involve societal values. A specialist's role is to bring as much relevant knowledge as possible to participants in a decision, whose job is to make the value-laden choices. These characteristics of science and scientific analysis all show the importance of appropriately broad-based deliberation as well as analysis: to determine what kind of analysis a decision requires; to incorporate information from disparate sources; to determine when analysis is appro- priately balanced; and to determine how to synthesize the results of analy- sis to make them useful to decision participants. Good science is a neces- sary in fact, an indispensable-but not sufficient basis for good risk characterization.
THE IDEA OF RISK CHARACTERIZATION AN EXPANDED FRAMEWORK 27 The aim of risk characterization, and therefore also of the analytic- deliberative process on which it is based, is to describe a potentially haz- ardous situation in as accurate, thorough, and decision-relevant a manner as possible, addressing the significant concerns of the interested and af- fected parties, and to make this information understandable and acces- sible to the parties and to public officials. If the underlying process is unsatisfactory to some or all of the interested and affected parties, the risk characterization will be unsatisfactory as well. A risk characterization can be only as good as the analytic-deliberative process that produces it. Figure 1-2 presents our conception of the risk decision process. The rest of this chapter specifies the role of risk characterization in this pro- cess; the rest of the book elaborates on the key elements in the figure and on ways to improve risk characterization. A New Definition and Its Implications We begin with a definition of risk characterization and then elaborate key parts of it: Risk characterization is a synthesis and summary of informa- tion about a potentially hazardous situation that addresses the needs and interests of decision makers and of interested and affected parties. Risk characterization is a prelude to decision making and depends on an iterative, analytic-deliberative pro cess. A risk characterization has many users. Risk characterization docu- ments are often prepared as if the only significant users will be legally designated decision makers, such as government officials. For example, the 1995 EPA policy statement on risk characterization (Browner, 1995) distinguishes risk characterization from "risk communication" on the grounds of who uses it. Risk communication, according to the statement, "emphasizes the process of exchanging information and opinion with the public," while risk characterization "addresses the interface of risk assess- ment and risk management." In this usage, a summary of knowledge about risk is a risk characterization if it is addressed to agency officials (it is presumed to be "available to the public" as well), but not if it is in- tended primarily for others. We do not make this distinction because summaries prepared primarily for internal use may ultimately be used by a range of parties and because such summaries can often benefit from exchanges with "the public" as well as within the agency. Agencies should recognize from the start that a risk characterization should be
28 / = ...~... it.. ..~., no. ........... En. m.. ..... ... o ._ In ._ :~ ,,,,.., ...: ~ U) V) a; u o o ._ CD ., - a' U) .~ o o .~ V) a; ._. a; U) ¢
THE IDEA OF RISK CHARACTERIZATION useful to multiple parties with different interests, concerns, and informa- tion needs. 29 A risk characterization may need to consider a wide variety of out- comes or consequences. In principle, the full range of potential harms and losses from a hazard is appropriate for treatment in a risk character- ization. In addition to the biological and physical outcomes that are typi- cally covered, decision makers and interested and affected parties often need to know about the significant economic costs and benefits of alterna- tives, the secondary effects of hazard events, or the efficacy of alternative regulatory mechanisms. For some decisions, consequences to human health and environmental quality are only part of what is of concern; concerns may also include such matters as geographical, racial, or eco- nomic equity, intergenerational tradeoffs, and informed consent by those who will be affected by a decision. Many of these issues can be analyzed systematically, and the summary of such analyses should be included in a risk characterization. Problem formulation is a paramount consideration. Because a risk characterization is geared to a risk decision, the knowledge to be devel- oped also has to be geared to the decision. To get the right science, it is necessary to ask the right questions. But identifying the questions to be addressed in a scientific analysis is not a straightforward task. For ex- ample, when assessing risk associated with a Superfund site containing hundreds of substances, or the ecological effects of introducing a new aeneticallv engineered organism into the natural environment, there tarp ~ , a many possible harms that could be analyzed, and it is not at all obvious which ones should be chosen. Consequently, the problems selected for analysis hazardous substances or processes, undesirable effects, and options for action need to be determined in consultation with the deci- sion makers and the interested and affected parties. A risk characteriza- tion will fail to be useful if the underlying analysis addresses questions and issues that are different from those of concern to the decision makers or the interested and affected parties. The key role of problem formula- tion in understanding risk has been strongly stated with respect to eco- logical risks (U.S. Environmental Protection Agency, 1992a), but its im- portance extends much more broadly. Effective risk characterization depends on an iterative process with feedback. The risk decision process is a goal-directed activity, and it should be iterative. It involves a series of tasks and feedbacks that allow for learning by all participants. Risk characterization, that is, the task of synthesis, depends on preceding tasks in the process, such as problem
30 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC - TIC SOCIETY formulation and information gathering. It also depends on users' reac- tions to previous risk characterizations and decisions, which may lead to redefinition of a problem and change the needs for analysis and synthesis. Risk characterization depends on an analytic-deliberative process. Understanding a risk depends on the interplay of two processes. Analysis includes various ways of reasoning and drawing conclusions by system- atically applying theories and methods from natural science, social sci- ence, engineering, decision science, logic, mathematics, and law. Delibera- lion includes the methods by which people build understanding or reach consensus through discussion, reflection, persuasion, and other forms of communication-processes that allow for interaction across different groups of experts and between experts and others. Both analysis and deliberation are essential, and they interact within each of the tasks lead- ing up to risk characterization: deliberation frames analysis, and analysis informs deliberation. Effective risk characterization depends on the appropriate repre- sentation, involvement, or participation of the interested and affected parties. Successful risk characterization depends on input from three kinds of actors: public officials or other designated decision makers; analytic experts, such as natural and social scientists; and the interested and affected parties to the decision. Scientists, of course, are often em- ployed by government agencies or interested parties. Public officials and some scientists are usually included in the process that leads to risk char- acterization, but the interested and affected parties are sometimes over- looked. Their inclusion is critical to ensure that all relevant information is included, that it is synthesized in a way that addresses the parties' con- cerns, and that those who may be affected by a risk decision are suffi- ciently well informed and involved to participate meaningfully in the decision. The interested and affected parties have a right to influence which questions should be the subject of analysis and can contribute both to developing information and to the deliberative parts of the process. The most appropriate types of inclusion will depend on the particular risk decision and may vary at different points in the process leading to a risk characterization. Sometimes, the situation may require direct participa- tion and involvement by interested and affected individuals or groups; at other times, representation by surrogates may be most appropriate. Iden- tifying the interested and affected parties and obtaining their appropriate involvement are important at all steps of the process that informs risk . . . Decisions.
THE IDEA OF RISK CHARACTERIZATION 31 The appropriate level of effort for a risk characterization is situa- tion specific. The above discussion may seem to suggest that risk charac- terization should always be based on extensive public participation at every step, lengthy deliberation, detailed analysis of a great variety of possible adverse outcomes, and the like. Such is not our intention. A1- though we believe analysis, deliberation, and participation have too often been inappropriately restricted in processes leading to risk characteriza- tion, extensive and expensive efforts in these directions are only occasion- ally warranted. judgment is critical in determining the amount, content, and timing of analysis, deliberation, and participation that are appropri- ate for supporting a particular risk characterization. Good judgment re- sults from two things: careful diagnosis of the decision situation to arrive at preliminary judgments on these matters and openness to reconsidering those judgments during the process. The procedures that govern risk characterization should leave enough flexibility for the process to be ex- panded or simplified to suit the needs of the decision. Our conception of risk characterization may perhaps be best seen by reference to Figures 1-1 and 1-2. In both conceptions, risk characteriza- tion involves synthesizing or summarizing information; the difference lies in who is involved in producing the characterization, what is synthe- sized, and how. In Figure 1-1, risk characterization is an activity con- ducted by experts in risk analysis that synthesizes or summarizes the results of analytical work by the same or similar experts. As this view of risk characterization is often implemented, expert judgment and statisti- cal techniques are the most important methods used to synthesize infor- mation. Risk characterization is conceived as the last step in a process of information gathering and interpretation. It makes sense of available information, but does not affect what information has become available. In Figure 1-2 risk characterization is conducted by a more diverse group of participants that may include, depending on the needs of the situation, not only analytic experts (labeled "natural and social scien- tists"), but also public officials and interested and affected parties. (The categories overlap: scientists, for example, may work for government, interested parties, or independently.) The information that goes into a risk characterization is determined by a similarly diverse group and is developed to meet their concerns, as well as the needs of the decision. Thus, risk characterization is not only the end of an analytic process, but also an important shaper of that process. The needs of risk characteriza- tion help formulate the problem for scientific analysis and influence the ways information is generated and interpreted. These relationships are represented by some of the various feedback loops in Figure 1-2. Expert judgment and statistical analysis are used to synthesize information, but
32 UNDERSTANDING RISK: INFORMING DECISIONS INA DEMOC~TIC SOCIETY this step also relies on deliberative methods that allow information to be considered from multiple perspectives. For risk characterization to meet the needs of a decision, it is important for each step of the process prior to synthesis to integrate analysis and deliberation and to involve, as appro- priate, scientists, public officials, and interested and affected parties. Figure 1-2 represents the interplay of analysis and deliberation and of these various participants throughout the process leading to a risk charac- terization and, beyond that, in decision making and implementation. Ar- rows indicate both the presence of feedbacks and the major direction of the process, which moves toward decision and action. The figure shows public officials, scientists, and interested and affected parties as partici- pating throughout the process, which represents the default presump- tion. The figure presents more detail about the processes preceding a decision than those following it because it is the former processes that are our main concern. What happens during and after a decision is more complex than shown, but those details are not as directly important to the success of risk characterization. Our definition of risk characterization will seem overly broad to some readers who think of risk characterization as simply the summary of avail- able scientific information about risk. This narrower definition is widely used and familiar from previous National Research Council (1983) work. We found this definition wanting because it suggests that a risk character- ization (and by inference the understanding of the people who use and comprehend it) is acceptable if it adequately reflects and represents exist- ing scientific information. - We have written a broader definition that highlights the fact that risk characterizations that meet this test can and do still fail when, for instance, the underlying analysis fails to address the questions that the users of the characterization see as relevant, when the characterization fails to reflect important perspectives and concerns, or when the process inappropriately restricts participation. Our definition retains the sense of risk characterization as a synthesis or summary, but it offers a broad conception of what it should synthesize, for whom, and how it should be developed. The definition makes clear that even though risk characterization does not include all the activities represented in the arrow at the left of Figure 1-2, its success depends on the quality of all of these prior activities. The need for a broader concept of risk characterization derives, we believe, from a shift in the roles of risk analysis and characterization in public policy over the past two decades. Before the 1970s, much of the deliberation that drove federal risk decisions occurred in Congress; by the time agency officials entered the process, legislation had already formu- lated the problems, defined the decision processes, and identified the adverse outcomes that would trigger regulatory action. Regulatory agen
THE IDEA OF RISK CHARACTERIZATION 33 cies such as the Food and Drug Administration and the Environmental Protection Agency had the task of determining facts, such as whether a chemical was a carcinogen, and their substantive decisions were sup- posed to follow more or less automatically from findings of fact (e.g., Interagency Regulatory Liaison Group, 1979; Rodricks, 1988; Paustenbach, 1989; Albert, 1994~. When the concept of risk characterization was devel- oped in the late 1970s and early 1980s, almost all regulatory experience with risk had been with decisions of this type. Over the past two decades, however, agencies have increasingly been called on to conduct risk analy- ses with less well-specified purposes-for example, to call to legislative attention new problems that may require regulation; to assess loosely formulated risk problems, such as those involving risks to ecosystems; and to address local issues, such as those of hazardous wastes, where legislation has not specified how agencies should arrive at decisions or which outcomes they should consider. A narrow concept of risk charac- terization and a linear view of the risk decision process may have been adequate when an organization was dealing with only a small part of the process; now, when public agencies are routinely responsible for much more of the process, a broader view is necessary. The Risk Assessment-Risk Management Distinction The traditional view of the risk decision process makes a sharp dis- tinction between two functions, risk assessment (understanding) and risk management (action); see Figure 1-1. Risk assessment is usually defined as the scientific analysis and characterization of adverse effects of envi- ronmental hazards. It may include both quantitative and qualitative de- scriptors, but it often excludes the analysis of perceived risk, risk com- parisons, and analysis of the social and economic effects of regulatory decisions (e.g., National Research Council, 1983:18~. Risk assessment is often presumed to be free of value judgments, with some important ex- ceptions, such as choices about whether and to what extent to include worst-case assumptions in risk assessments, a choice that may be made differently depending on whether the assessment is being conducted to determine regulatory priorities or priorities for testing (National Research Council, 1983:40~. Risk management refers to the activities of identifying and evaluating alternative regulatory options and selecting among them. Risk managers are supposed to deal with broad social, economic, ethical, and political issues in choosing from among a set of decision options by using the results of the risk assessment and their understanding of the other issues. Making tradeoffs, which may be called risk-benefit, cost- benefit, or risk-risk evaluations, is part of risk management. The conceptual distinction between risk assessment (understanding)
34 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOCRATIC SOCIETY and risk management (action) remains useful for various important pur- poses, such as insulating scientific activity from political pressure and maintaining the analytic distinction between the magnitude of a risk and the cost of coping with it. For the purposes of improving decision-rel- evant understanding of risk and making that understanding more widely accepted, however, a rigid distinction of this sort does not provide the most helpful conceptual framework.4 The reason, in brief, is that the analytical activities generally considered to constitute risk assessment are not sufficient by themselves to provide the needed understanding. Much of this volume elaborates on how those activities must be shaped and complemented by deliberation in order to yield useful risk characteriza- tions. Our judgment partly reflects several developments since 1983. Risk characterizations have been needed for a much wider range of policy questions, including many in which the nature of the problem and the identity of the available choices is not at all obvious; articulate and scien- tifically informed public opposition to risk decisions has revealed gaps in many risk analyses; experiences with risk communication have demon- strated that official summaries of risk are often incomprehensible, confus- ing, or irrelevant to many of the affected parties; and public trust in many of the organizations that conduct risk assessments has declined. These developments underline the limitations of an approach to informing risk decisions that presumes that it is sufficient to get the science right: that the sound way to build understanding of risks is to apply methods from epidemiology, toxicology, statistics, and a small number of other scien- tific specialties. We believe that acceptance of too strict a separation between risk assessment and risk management has contributed to an unworkably narrow view of risk characterization. Careful studies of the risk decision process have increasingly ac- knowledged the limitations of a strict separation. They were recognized first in the Red Book itself (National Research Council, 1983), which pointed to the need to iterate between risk assessment and risk manage- ment so that assessment could incorporate analytical assumptions that may need to be different for functions such as initial screening and the evaluation of regulatory options: it was noted that "a single risk assess- ment method may not be sufficient" (p. 40) and that the choice of appro- priate assumptions required interaction between the assessment and man- agement functions. As this study further noted (p. 142~: 4We therefore generally avoid the terms risk assessment and risk management in this book although we do use the former term in discussing government agency functions or products that are normally identified by that name.
THE IDEA OF RISK CHARACTERIZATION Separation of the risk assessment function from an agency's regulatory activities is likely to inhibit the interaction between assessors and regu- lators that is necessary for the proper interpretation of risk estimates and the evaluation of risk management options. Separation can lead to disjunction between assessment and regulatory agendas and cause de- lays in regulatory proceedings. 35 The Red Book recognized that "interpretation of risk estimates" involves an important element of judgment because of gaps in data and theoretical understandings, and this theme was elaborated much further in the more recent National Research Council (1994a) report, Science and judgment in Risk Assessment. Methods of risk analysis can make only a limited contri- bution to improving such judgments. Progress can be made, however, by strengthening the processes, only some of which are analytical in nature, that are used for informing risk decisions. Our framework emphasizes a series of tasks that support risk charac- terization and help inform risk decisions and the two processes by which these tasks are performed: analysis and deliberation. It emphasizes that there is a role for both scientific method and for appropriately broad- based deliberation in each of the tasks. It makes explicit that although good analysis is essential, it is not the only way to increase understanding among participants in risk decisions; indeed, relying on analysis alone is detrimental to the enterprise. This framework implies that those responsible for risk decisions should look at each task differently. Before determining what informa- tion to gather, and well before considering how to summarize it for par- ticipants in a decision, they should ask several diagnostic questions, such as: What are the decisions that can or need to be made? Which outcomes for individuals, society, or the environment are of concern? Who are the potential participants public officials, scientific and technical experts, and interested and affected parties in the decisions? What information would be needed to address the questions or problems as identified by the participants and to satisfy them that their concerns are being given adequate consideration? Who should be involved in answering these questions so that the answers are acceptable to the participants? The answers to such questions will determine how to structure the process for informing the particular decision to be made. Structure of the Book The rest of this book substantiates our framework for risk character- ization and sets out some of its implications for practice in government agencies and other organizations. Chapter 2 summarizes evidence that provides much of the rationale for our framework, particularly evidence
36 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY of the many important judgments embedded in each step leading to risk characterization. Chapters 3, 4, and 5 focus on the analytic-deliberative process. Chapter 3 defines deliberation, explains the need for appropri- ately broad-based deliberation for risk characterization, and discusses some principles for organizing effective deliberation. Chapter 4 discusses the general principles and purposes of analysis in the context of risk characterization, a subject that has been elaborated in greater detail else- where, and focuses particularly on two analytical issues: simplifying the understanding of risk by combining many of its dimensions into one and characterizing uncertainty. Chapter 5 addresses the challenge of finding an appropriate balance of analysis and deliberation within each of the major steps of the process leading to a risk characterization. Chapters 6 and 7 deal with implementing risk characterization. Chap- ter 6 discusses the implementation of our framework. It addresses the issue of practicality and the problem of matching the analytic-deliberative process to the decision. It especially emphasizes the diagnostic effort that is required at the beginning of the process. It also addresses the problem of building the capability to implement the framework. Chapter 7 pre- sents a set of principles for implementing the process and approach to risk characterization that we advocate. Appendix A details some of the risk decision cases that are referred to briefly throughout the text, and Appendix B briefly discusses some common approaches to deliberation and public participation, noting the research literatures on them.