Click for next page ( 38


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 37
2 Judgment in the Risk Decision Process A risk characterization is part of a process that begins with the formulation of a problem the likelihood of a harm-and ends with a decision. A risk characterization cannot make up for deficiencies in other parts of the process; inversely, if the other parts of the process are done well, it is far more likely that the risk characterization will be both clear and useful. Some of the analytical difficulties affecting risk characterization are well known, such as the difficulty of determining an appropriate math- ematical model for extrapolating from animal toxicological data to assess the health consequences of human exposures and of comparing best esti- mates of different risks when their uncertainty distributions differ in shape or in variability (see, e.g., Finkel, 1990; National Research Council, 1994a). In this chapter we focus on other difficulties, often overlooked in the extensive literature on risk analysis, that are equally important for understanding and coping with risk. Many of these difficulties result from judgments made at each step of the process that can undermine the quality of risk characterization and, if they are unacceptable to some of the interested and affected parties, become lightning rods for conflict. Such difficulties tend to arise when the knowledge and perspectives of these parties were not adequately incorporated into the process that led to the judgments. Many of the difficulties can be prevented or reduced if the process is recognized from the start to require both analysis and delibera- tion and if it is organized to ensure that the judgments are informed by appropriate deliberations. 37

OCR for page 37
38 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY We consider in detail the steps of problem formulation, selecting op- tions and outcomes to consider, and information gathering, as well as synthesis, usually the major focus of risk characterization. We discuss process design at the end of the chapter. We document the variety of judgments made during each of these steps and some of the ways these judgments can undermine understanding of risks and contribute to mis- trust and public conflict about risk decisions. The chapter concludes with a strategy for avoiding these outcomes by designing the analytic-delib- erative Process so as to inform the kev judgments with the knowledge 1 J J (I and perspectives of the range of decision participants. Certainly, many risk characterizations and risk decision processes have been appropriate for the decision at hand. However, as we note in Chapter 1, some high-profile, controversial risk characterizations have suffered from deficiencies, and sometimes the damage to decision mak- ing has been significant. The deficiencies also threaten some lower-pro- file risk characterizations. PROBLEM FORMULATION Perhaps the most basic difficulty with risk characterization is that the people who will or should participate in the risk decision process fre- quently have divergent perspectives on the decision at hand. Differences of perspective cause problems because efforts to inform decisions neces- sarily proceed from some implicit formulation of the problem: a risk characterization that deals selectively with only one perspective on a prob- lem will be inadequate for those with significantly different perspectives. The Concept of Risk ~. ~. ~ Judgments pervade any understanding of risk (National Research Council, 1994a). Some writers even question the idea that risk should be conceptualized as a quantifiable physical reality (e.g., Douglas and Wildavsky, 1982; Funtowicz and Ravetz, 1992; Krimsky and Golding, 1992; Otway, 1992; Pidgeon et al., 1992; Slovic, 1992; Watson, 1981; Wynne, 1992~. They argue that the concept of risk helps people interpret and cope with the dangers and uncertainties of life, including but not limited to the prospect of physical harm, and that the concept is shaped by human minds and cultures. That is, there are many different kinds and qualities of dangers and many potentially useful ways of making sense of them, and even though many of these are measurable in prin- ciple, it is judgments and values that determine which ones are defined in terms of risk and actually subjected to measurement. The multidimensionality of risk and the many ways it can be viewed

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS 39 help explain why risk characterizations sometimes lack authority for some of the interested and affected parties to a decision, even when the charac- terizations are supported by high-quality analysis. Individuals and groups that do not share the judgments and assumptions about the prob- lem formulation that underlie a risk characterization may well see the information it provides as invalid, illegitimate, or not pertinent. They may see the characterization as flawed because the underlying risk analy- sis is based on controversial assumptions (often implicit) about which perspectives are legitimate, which solutions are reasonable, and which types of information are useful or relevant (Vaughan and Seifert, 1992~. The history of risk analysis is filled with instances in which analysis, at least to some of the parties, seemed to beg the question. One such case is the risk analysis for the Yucca Mountain nuclear waste repository site, mentioned in Chapter 1. Billions of dollars were spent on assessing the quantitative and calculable risks associated with permanent disposal at one site, when many people believe it would have been more productive to assess the risks of temporary storage while engaging in a more thor- ough debate on the merits of a permanent solution. Another example was the comparison of coal and nuclear power generation in the 1970s that did not consider slowing the growth of energy demand as one approach to finding sufficient electric generating capacity to meet national needs. Even the apparently straightforward act of defining the hazardous pollutant to be characterized can embed important assumptions about the nature of the problem. Should one consider narrow classes of compounds such as dioxins, or broad classes such as the thousands of organochlorines and assess chlorine as the relevant environmental risk? Such choices, though they should be well informed by toxicology and other relevant science, involve important acts of judgment that shape risk characteriza- tion and even decision making (Fischhoff et al., 1981; O'Brien, 1995~. In the chlorine example, the definition of the hazard is highly consequential for the chemical industry and for proponents of pollution prevention; a risk characterization based solely on either formulation might be unsatis- factory to one of the interested and affected parties. Missing Considerations Three considerations that are often missing from the formulation of risk problems have led to disputes about the subsequent risk character- izations: fairness, prevention (of pollution or risk), and rights. Acknowl- edging these concerns may lead to different (usually broader) problem formulations than those that emerge from the ordinary routines of gov- ernment agencies. The way such concerns are or are not addressed can

OCR for page 37
40 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMO C~TIC SOCIE~ directly affect choices about the options and outcomes to consider in char- acterizing risk. F. alrness For some interested and affected parties in risk decisions, managing environmental risks has become a question of fairness, moral responsibil- ity, and distributional equity (Beach, 1990; Bullard and Wright, 1992; Law- less, 1977; Nelkin, 1989; Sandman, Weinstein, and Klotz, 1987; Vaughan and Seifert, 1992~. An example from Chester, Pennsylvania, shows how fairness issues can arise in risk characterization. Chester is an industrial city with a declining population (now about 40,000) consisting largely of low-income African Americans. It has become the site of numerous haz- ardous facilities, including two oil refineries, a trash incinerator, some Superfund sites, and an autoclave facility for infectious materials. When a proposal arose to site a soil decontamination plant in Chester, the Penn- sylvania Environmental Protection Agency proposed to do a risk assess- ment of the plant project, examining its likely emissions and projecting the incremental health risks to the local population on the basis of models of exposures and dose-response relationships. But community represen- tatives raised several other issues, one of which was the claim that adding a new hazardous facility was unfair in a city where residents were al- ready bearing more than their share of toxic exposures. The city's questions led the regional office of the U.S. Environmental Protection Agency (EPA) to agree to conduct an analysis based on a dif- ferent problem formulation: a cumulative risk assessment that would characterize pollution in Chester generally and identify the areas of high- est risks. Such a risk analysis would focus serious scientific attention on matters that are not considered in an incremental-risk analysis, including the interactive effects of the hazards present in the city, the health effects of the new hazard on a population whose health status may be compro- mised by other exposures, and the comparison of overall pollution risks in the exposed population with those in more affluent communities nearby. These issues would not have been addressed in the original state- proposed analysis. The city's position was that any risk characterization that ignored these issues would be incomplete and inadequate in terms of providing the information needed to make a major decision about public health (personal communication, Gregory Schirm, 1994~. Some fairness concerns, described in terms of "environmental jus- tice" for minority and low-income populations, were given prominence by Presidential Executive Order 12898, issued in February 1994. The Executive Order recognized that federal agencies' risk analyses had not previously made equity issues a routine part of the problem definition

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS 41 and directed them to do so. Effective implementation of the order would make the analysis of some aspects of fairness and equity an essential input into risk characterization. Prevention Current debates about preventing pollution and risk also show clearly how problem formulation shapes risk characterization and the entire risk decision process. One proponent of pollution prevention has criticized standard practice in risk analysis for asking the question, "Which envi- ronmental problems can we ignore?" (because their risks are negligible), rather than the question, "How can we avoid exposures to hazards?" (O'Brien, 1995~. While both formulations omit the key question of cost, the latter question invites consideration of a much wider range of possible policy options (especially, pollution prevention). For many interested and affected parties, a risk characterization that does not address preven- tion could never provide the information they need to accept a risk deci- sion, no matter how well the narrower problem is analyzed or character- ized. The controversy over control of the Mediterranean fruit fly, a major pest to the $2 billion fruit and vegetable industry in California, illustrates, among other things, how judgments about which aspects of risk to char- acterize can obscure or highlight debates over pollution prevention. Much of the Medfly debate and the associated risk analysis and characterization in the early 1990s focused on estimates of risks to the health of residents who might be exposed to the malathion spray. The debate centered on dose-response questions and on the appropriateness of assumptions about the behavior and therefore the exposure of residents who are warned to stay indoors during an imminent spraying but who might not do so. It is based on the formulation that the Medfly problem in California comes from flies that have recently entered the state on infested fruit and that spraying where flies have been seen will eradicate the problem (Califor- nia Department of Food and Agriculture, 1994~. Some critics argue that this formulation is incorrect because the flies are now established as breeding populations in California and cannot be eradicated by a spraying program (Carey, 1991; 1994~. They argue further that the state has a vested interest in the isolated infestation assumption because under it spraying might easily and cheaply convince potential importers of California produce that it is fly free. If, however, the Medfly is an established pest, malathion will not perform as claimed, and biologi- cal pest control would become an alternative option worth considering. Biological control would also avoid the projected human exposures to malathion spray. The formulation of the problem as one of keeping an

OCR for page 37
42 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMO CRITIC SOCIETY established pest population under control changes the questions for risk analysis: the focus shifts to the risks of biological control approaches and the comparative effectiveness of chemical and biological Medfly control; it makes pollution prevention, in the form of nonchemical means of pest control, a key policy option. If the Medfly problem has been incorrectly formulated, much of the risk analysis and characterization has focused on an incomplete set of options, and the needs of decision participants for understanding have been poorly served. Rights Risk characterizations have also become controversial when they pay little attention to issues of the rights of individuals or groups to control their own lives. An example is the continuing debate about the fluorida- tion of public water supplies. Advocates declare that this technique is a safe and cost-effective method of improving people's dental health. Op- ponents in addition to questioning the certainty of the scientific esti- mates and raising the issues of increased risk of bone fractures in the elderly and fluoridosis in those with poor kidney function speak about individual rights and the undesirability of a public health policy that eliminates individual choice in regard to exposure to a chemical agent that has both benefits and risks (Lawless, 1977; Martin, 1989~. Another example is the contentious debate about environmental "un- funded mandates,', federal decisions that require states and localities to spend their money on certain environmental and health projects, which consequently reduces funds available for other projects. Many interested and affected parties claim that they have rights to clean air and water and that the federal government should protect these rights. Other parties, chiefly state and local officials, complain that federal decisions that com- pel their action, even if based on sound risk analyses, abrogate the right of localities to use their funds to reduce risks in the most cost-effective ways. While federal officials may have sponsored risk analyses to focus on re- ducing the risks of environmental chemicals, a state public health depart- ment or a mayor faces a different problem: a choice between reducing risks to citizens from chemical residues in water or from birth defects, traffic accidents, or violent crime. A risk characterization for water pollu- tion alone may be quite beside the point for a local official confronting such a choice. SELECTION OF OPTIONS AND OUTCOMES Problem formulation has practical implications for other steps in the risk decision process. Among the most important is the way it shapes

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS 43 choices about which options to consider and which possible adverse out- comes to analyze- choices that are critical to the success of risk character- ization.1 For a risk characterization to meet the needs of participants in a decision, it must consider the range of plausible decision options. The parties to a decision may not agree on which options are worth consider- ing, but a risk characterization that does not consider an option that one of the participants views as promising is likely to be seen as biased and inadequate. The controversies over the Medfly and unfunded mandates are examples. When problem definitions truncate the list of options too severely, risk characterizations can be doomed to controversy long before they are undertaken. Organizations responsible for risk characterizations should make ef- forts to identify the range of decision options that experts and the spec- trum of interested and affected parties consider viable. Generating an adequate list of options may be difficult. It demands familiarity with the context of the decision, knowledge about the scientific and technical as- pects of the possible risks, and, sometimes, creativity and imagination. Also, it often demands that organizations listen to the interested and affected parties. Although identifying the range of options is challenging, it is a key to successful risk characterization. Considering a sufficiently broad range of possible harms or losses is equally important to risk characterization. Typically, analysis focuses on only a few adverse outcomes such as cancer or birth defects in human beings, loss of a species, or elimination of a habitat that are judged to be the most serious of the possible harms. The analysis is sometimes further restricted to the effects of exposure to a particular agent (a substance or process), through a particular medium (e.g., air, water, or food intake). Agencies may narrow the list of outcomes because they have narrow responsibilities under law, because funds for analysis are limited, because of political pressures, or for other reasons. Consideration of only a few possible outcomes is usually justified on the assumption that if a decision protects adequately against the selected outcomes, it will also protect against the others, because environmental and health hazards are strongly correlated. This assumption becomes increasingly suspect as the range of outcomes of concern expands from overt human health risks to effects on the immune system and related systems (e.g., allergenicity); behavioral effects; psychological effects, such 1Selection of outcomes has sometimes been treated as a stage within problem formula- tion: for example, the U.S. Environmental Protection Agency (1992a:12) refers to outcomes as "ecologically based endpoints". We prefer to make a sharper conceptual distinction between the two tasks.

OCR for page 37
44 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY as anxiety and depression; ecological effects; and social, economic, and ethical impacts. Some participants may doubt that protection against one of these adverse outcomes is tantamount to protection against all. Some of the nonmeasured outcomes may be more salient to some parties, and they may argue that these other adverse outcomes should be subject to analysis and characterization. Demands for such expansion of the out- comes to be included can be especially meritorious when data are sparse on certain outcomes, but the risks have nevertheless been dismissed on the assumption that they are small enough to ignore. An adequate risk characterization must address all the outcomes or consequences of a hazardous situation that are reasonably important to the relevant public officials and to the interested and affected parties to the decision. Agencies should tailor their analyses to the decision to be made, addressing the potential adverse outcomes most significant for that decision. Ecological Effects One important class of nonhealth outcomes is harm to nonhuman organisms and ecosystems. The EPA has taken the lead in developing a conceptual framework for conducting ecological risk assessment (U.S. Environmental Protection Agency, 1992a, 1992b, 1992c, 1992f, 1992g, 1993h) and is preparing guidelines for this activity. Analysis is difficult because the effects may fall on individual animals or plants, on local populations of a certain species, on ecosystems (thus affecting many spe- cies), or on the survival of endangered species. At larger scales, effects on the distribution of ecological communities across the landscape are cen- tral to regional-scale ecosystem management (Grumbine, 1994; Harwell et al., in press). There may be important ecological outcomes to consider and characterize at each of these hierarchical levels of ecological systems (Harwell et al., 1990~. Ecological risk analysis requires an understanding of how the af- fected ecosystem functions. There are numerous interrelationships among taxa, across responses, and across organizational levels. In addition, some of the most important effects may be indirect, operating through several interrelationships. Many of these effects are inadequately understood, difficult to measure, or laden with uncertainty (National Research Coun- cil, 1993a). Some ecologists even dispute whether the concept of ecologi- cal risk (or its inverse, ecological health) is useful for policy analysis (e.g., Lackey, 1994, 1995~. None of these scientific difficulties of estimation, however, negate the importance for policy decisions of considering eco- logical outcomes. Interested and affected parties may want to take ac- count of ecological effects even if the level of scientific understanding of

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS 45 them is poor. Qualitative assessments of relative ecological risks can provide useful insights for environmental decision making (Harwell et al., 19921. A critical need is to develop appropriate tools for assessing the value of ecological systems, including both economic and noneconomic (e.g., intrinsic) values. Economic and Social Effects Economic consequences are sometimes inextricable from the other aspects of a risk. We are not referring here to the well-recognized eco- nomic costs of regulating risks, but to the economic costs of the hazards themselves. Many risk characterizations do not consider the full range of adverse economic outcomes, even though they are important for decision purposes and amenable to scientific analysis. A 1987 example from Brazil is illustrative. When two men seeking scrap metal pried open a metal capsule containing 100 grams of cesium 137, subsequent exposures to neighbors and family resulted in 4 deaths and 50 cases of radioactive contamination that required medical treatment. Described in this sense, the accident appears to have been of local significance, without major national or international impact. But that depiction fails to capture the $20 million dollars in cleanup costs and the subsequent 50 percent drop in the wholesale value of agricultural products from the Brazilian state in which the accident occurred. Sales of manufactured goods were also affected, despite the lack of plausible contamination (Freudenburg,1988~. The incident illustrates that losses in terms of human health may not be the only adverse outcomes of a hazardous situation that are worth charac- terizing. One important class of economic effects are those a given hazard has on nearby property values (Greenberg, 1995; Gregory, Flynn, and Slovic, 1995~. A relationship between changes in property values and proximity to hazardous waste sites has been repeatedly demonstrated (Greenberg and Hughes, 1992, 1993; Schulze et al., 1994; Skaburskis, 1989~. For fami- lies living near hazardous facilities, property value losses are sometimes a significant factor that they want estimated and considered in a decision process. Such information can provide guidance for making decisions about the costs of risk remediation plans (McClelland et al., 1990~. An- other kind of economic effect is the cost of insurance premiums and emer- gency preparedness that flow directly from the possibility of an adverse event (Freudenburg, 1988~. These costs are borne by the potentially af- fected population regardless of whether they actually suffer from the adverse event. Risk characterizations typically do not address social effects, perhaps because they are considered outside the purview of formal risk analysis.

OCR for page 37
46 UNDERSTANDING RISK: INFORMING DECISIONS INA DEMOC~TIC SOCIETY Yet they are legitimate objects for risk characterization because partici- pants in decisions need to understand them to make informed choices, and many social effects are amenable to systematic analysis. Social effects that may need to be considered in a risk decision include neighborhood disruption and issues of social equity and stigma (Gregory, Flynn, and Slovic, 1995~. Some risk decisions can significantly alter a community's character. Neighbors often express fears that a hazardous facility will be destructive to the community (Zeiss and Atwater, 1991~. It is not uncom- mon to see these kinds of concerns taken seriously in negotiations about citings, but they are not usually treated in conventional risk characteriza- tions. In part, what is at stake is community control (Elliot, 1984; Zeiss and Atwater, 1991~. People are more willing to tolerate a risk if they feel they have some control over the exposure (Slovic, 1987~. By the same token, removing control from a community has social costs, even if the commu- nity accepts a risk. For example, cleaning up a contaminated site near a residential community is a disruptive process. Several studies have docu- mented the tension between residents who want total removal of con- tamination (usually younger families with children) and those who want minimal disturbance and cost (usually older couples living on fixed in- comes who have owned their homes for many years and have no children living at home). The latter group sometimes opposes clean-ups that would disrupt their lives (e.g., remove their gardens, demand temporary evacuation, or involve high costs), while the former group finds the status quo more disruptive (Claus, 1995; Fessenden-Raden et al., 1987; Levine, 1982~. Risk analyses could, but rarely do, explicitly consider the effects of such kinds of neighborhood disruption. If the analyses implicitly set this potential loss equal to zero, affected parties may find the risk character- ization unsatisfactory. Many affected parties in risk decisions expect the government to en- deavor to achieve some fair balance between the risks a community or an individual bears and the benefits received. Recently, this expectation has been voiced as a concern for environmental justice for minority communi- ties (Bullard and Wright, 1992; Greenberg, 1993~. There is evidence of unequal distribution of noxious facilities between communities as a func- tion of economic and racial or ethnic differences (Bullard, 1990; Commis- sion for Racial Justice, 1987; U.S. Environmental Protection Agency, 1992h) and of differential harm associated with these risk sources (Greenberg, 1995~. However, equity concerns had rarely been considered in conven- tional risk assessments until Executive Order 12898, in 1994. To the extent that this directive is implemented, agency risk characterizations will be

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS 47 gin to provide information that will inform public deliberation on equity issues.4 People sometimes hold negative associations for things, places, orga- nizations, or people they connect to risks (Slovic, 1993b). Such stigma can have a tangible economic impact: in 1989, concerns about the use of Alar on apples led to nationwide decline in apple sales of over $100 million (about 10%) after risk assessments were publicized that linked the sub- stance to cancer in children (Rosen 1990~. Researchers have also identi- fied the potential economic effects from stigma associated with the pro- posal to construct a high-level radioactive waste facility in Yucca Mountain, Nevada (Slovic, Layman, et al., 1991~. For many siting deci- sions, the effects of stigma cannot be reduced by engineering and design alone, but it may be possible to address them through compensation or insurance (Fort, Rosenman, and Budd, 1993~. Although these effects or potential losses from stigma are difficult to quantify or compensate (Gregory, Flynn, and Slovic, 1995), they are nevertheless important to consider. Effects on Future Generations Many risk decisions may impose risks on future generations that re- quire a different kind of consideration from risks to people living today. The high-level nuclear waste disposal facility planned for Yucca Moun- tain, Nevada, is a striking example: releases of radioactive material from this facility could cause harm thousands of years in the future. Such situations present two questions for risk analysis: How can one be certain that the risks to future generations are known? How can one represent the interests of future generations in a current risk decision process? The difficulty of the first question is illustrated well by the Yucca Mountain controversy. As described in Chapter 1, a fundamental as- sumption of U.S. and international policy on radioactive waste disposal has been that safe, permanent disposal was the strategy most likely to reduce the risks to future generations. But a 1993 technical review com- mittee set up by the state of Nevada one of the interested and affected parties questioned even that most basic assumption. Arguing that no one today can predict what human beings might be able, or motivated, to do at the Yucca Mountain site over the next 10,000 years, the Technical Review Committee (1993:14) concluded that rather than protecting future 2Various formal analytical techniques exist for informing discussions about distributional equity: all involve controversial techniques for valuing human lives (e.g., Zeckhauser, 1975; Anderson, 1988; Leigh, 1989; Ellis, 1993).

OCR for page 37
62 CO o n Z~zO 0 0 0 ~' ~CO ~ ~o ~, ~' ~2 - E ~ u AL o a) <: o co Q ~O Z an 2 2 ~o ~ | ~ ~E . , ~ ~ ~ ~I E (D o , ~In: ~ ~ nil ~ ~ o ii - ~ O O a - ~ 0 ~ An ~ O g ~. Dig ~ . 3 a) 0 = 0 o ,~ ~o .. -A a) 6 o- by ~i- ~ o ~ ~ ~ O cn en a). - ~ ~D ~ ~6 ' ' ousts E 3 0 ~ ~ ~ 0 a) Ui

OCR for page 37
63 N -O o 0m , O O Q) o ' 'out Rm2~ 0 a) CO O Q a) ~ O ~ 3 ZY! g O ~ Q a, a) ~ In 0 ~ o. a) O ~ ~ ~ In o Q ~a) to oZ ~ ~' a) _ 0 ~ O ~ ~ 0 y C) Z Z C) ~ 3 ~ ~ g O _ u . _d o Cal u: cn ' ~ ~ 50 V .= ~ U) 50 o N a,, C ~ O ~ ~ ._ 00 ~ ~ ~ ._ O ~ o o CO $~ ~ - E~ .tn oo .= 50 o . .- ~ - c~ oo . . oo - ~ 50 ~ o~ rx x ~ o ~ . ~ .Q o uD . u) 5- o ~ cn ~ ~ .m - ~ ) t = ~ o ~ ~ ~o ~ - ~ o ~ ~ ~ .- . _ ~ _1 .- a; cJ) ~ ~ o ~ =.= ~ > .Q ~ a, ~ ~ ,= LI = LO ~ ~ ~ ~ u L~ - cn o a;

OCR for page 37
64 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY with the ways that risks tend to be characterized in quantitative risk assessments (Slovic, Fischhoff, and Lichtenstein, 1979; The Royal Society Study Group, 1992; Lindell and Malmfors, 1994~. Conflicts over "risk" may reflect differences between specialists in risk analysis and others on their definitions of the concept. In this light, it is not surprising that citations of statistics about "actual risks" often do little to change most people's attitudes and perceptions. Nonspecialists factor complex, qualitative considerations into their estimates of risk, in- cluding judgments about uncertainty, dread, catastrophic potential, con- trollability, equity, and risk to future generations. The legitimate, value-laden issues that underlie these multiple di- mensions of risk need to be considered in risk policy decisions (Fischhoff, Watson, and Hope, 1984~. For example: Is risk from cancer (a dread disease) worse than risk from auto accidents (not so dreaded)? Is a risk imposed on a child more serious than a risk accepted voluntarily by an adult? Are the deaths of 50 passengers in separate automobile accidents equivalent to the deaths of 50 passengers in one airplane crash? Is the risk from an industrial emission worse if the facility is located in a neighbor- hood that has a number of other hazardous facilities nearby? The difficult questions multiply when outcomes other than human health and safety are also considered. As noted in an earlier study (National Research Council, 1989:51~: Technological choices sometimes involve weighing the value of a river vista, a small town style of living, a holy place, or the survival of an endangered species, in addition to human health, against probable ben- efits. Such matters are ultimately matters of values. The fact that hazards differ dramatically In their attributes or charac- teristics helps explain why certain technologies or activities, such as nuclear power, evoke much more intense public opposition than others, such as motorcycle riding, that cause many more fatalities. The implica- tions of "risk perception" for synthesis have been well described in a previous study (National Research Council, 1989:52~: Those quantitative risk analyses that convert all types of human health hazard to a single metric carry an implicit value-based assumption that all deaths or shortenings of life are equivalent in terms of the impor- tance of avoiding them. The risk perception research shows not only that the equating of risks with different attributes is value laden, but also that the values adopted by this practice differ from those held by most people. For most people, deaths and injuries are not equal some kinds or circumstances of harm are more to be avoided than others. One need not conclude that quantitative risk analysis should weight the risks to conform to majority values. But the research does suggest that it is presumptuous for technical experts to act as if they know, without

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS careful thought and analysis, the proper weights to use to equate one type of hazard with another. When lay and expert values differ, reducing different kinds of hazard to a common metric (such as number offatalities per year) and presenting comparisons only on that metric have great potential to produce misunderstanding and conflict and to engender mistrust of expertise. 65 This analysis is still pertinent with reference both to techniques for char- acterizing relative risks in terms of expected deaths and to techniques that compare hazards on a common monetary metric, such as contingent valu- ation methods or various forms of cost-benefit analysis. A number of risk analysts have sought technical solutions to the prob- lem of taking qualitative aspects of risk into account. Generally, they have proposed broadening risk analysis to incorporate one or more of the various characteristics identified in studies of perceived risk: for example, distinguishing between voluntary and involuntary activities in assessing risk-benefit balances (Starr, 1969~; giving proportionally more weight to large accidents than to numerous small accidents that cause the same amount of damage or number of deaths (Wilson, 1975; Griesemeyer and Okrent,1981~; and adjusting risk estimates to take into account the impor- tance of various risk-perception characteristics (Rowe, 1977; Litai, Lanning, and Rasmussen, 1983~. None of these proposals has yet been developed to the point of application in actual risk assessments. A related approach has successfully integrated several of the dimen- sions of risk in a formal way and done so in real applications. This approach has been developed by two Swiss analysts to aid decisions about the safety of ammunition storage depots and transportation systems, in- cluding the design of a high-speed railway in Germany (Bohnenblust and Schneider, 1984~. The method characterizes the risk reduction in terms of cost-effectiveness, and it attributes more value to reducing risks that are involuntary, poorly understood, potentially catastrophic, and hard to con- trol. Although this Swiss model has been applied to a number of impor- tant decision problems, there is a need to align the model more closely with recent research that has been done on social, cultural, and psycho- logical factors (see, e.g., Krimsky and Golding, 1992~. This might be ac- complished, at least in principle, by refining quantitative approaches to risk analysis (see Chapter 4~; it might also be achieved qualitatively through deliberative processes (see Chapters 3 and 5~. Another way to incorporate some of these dimensions into risk char- acterizations is to apply multiattribute utility analysis (e.g., von Winter- feldt and Edwards, 1986; Fischhoff, Watson, and Hope, 1984; Gregory, Lichtenstein, and Slovic, 1993; Brody and Rosen, 1994; see also Appendix A). In this technique, individuals identify value dimensions, attributes, or outcomes that are important to them, assign relative weights to them, and evaluate the outcomes identified by risk analysis (or a set of policy

OCR for page 37
66 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOCRATIC SOCIETY options) on each attribute. Multiattribute utility analysis allows individu- als to compare options that yield different packages of risks and benefits. It allows for evaluations to be explicitly subjective: individuals can, for example, assign a numerical value to a quality such as "incidental en- counters with neighbors." But it does not, by itself, solve the problem of providing risk estimates for populations because there is no acceptable formula for aggregating individuals' evaluations. Should each indivi- dual's evaluation have equal weight, or should those who might bear the risk have a greater weight? This problem of estimating risks for whole groups when the risks are of various kinds pushes the limits of analysis. (In Chapter 4 we discuss the strategy of using analytical techniques to reduce the dimensionality of risk. In Chapter 5 we discuss the alternative strategy of combining deliberative processes with analysis to help the participants in decisions develop working understandings of multi- attribute risks.) The Meaning of Risk Estimates Separable from the technical questions about how best to estimate the risk of a particular agent with respect to a particular outcome is the ques- tion of what the risk estimate means, or should mean, to participants in risk decisions. Risk characterizations often fail because they attribute meaning to scientific estimates in ways that mislead participants in the risk decision process or that are incomprehensible to them. This section discusses several such sources of failure: in the treatment of uncertainty, in inferences about which populations will be affected, and in inferences about how a risk estimate should be interpreted in light of other risks to which a population is exposed. Uncertainty Risk characterizations often give misleading information about un- certainty in several ways. They may give the impression of more scien- tific certainty or unanimity than exists (or of more uncertainty or dissen- sion). They may suggest that uncertainty is a matter of measurement when in fact it is a matter of disagreement about whether a particular theory applies or differences in judgment about how to infer something that is unknown from something that is known. And they may give the impression that certain risks do not exist when in fact they have not been analyzed. Civil engineers, public health professionals' and others often take account of uncertainty by a strategy of "conservatism." This means that they recommend decisions or actions that leave a margin for error that is

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS 67 intended to protect the public if the actual risk turns out to be greater than they predict. It is often argued that risk analysts should instead present their best available estimate to decision makers, along with an explicit characterization of its uncertainty, and allow the decision makers to de- cide explicitly how much margin of safety to allow.4 Either approach embodies a value choice about the best way to characterize risk and pro- tect public health and safety, and there is no scientific technique for deter- mining which approach is preferable (National Research Council, 1994a). There is strong agreement that risk analysts should explicitly summa- rize uncertainty, and there are methods for doing so (e.g., Morgan and Henrion, 1990~. But despite the admonitions of officials in some govern- ment agencies (Habicht, 1992; Browner, 1995) and the recommendations of outside panels (e.g., National Research Council, 1994a), many risk char- acterizations still present point estimates of risk, representing these as upper-bound estimates and providing little or no analysis of the extent of overestimation. In spite of the obvious shortcomings of point estimates and the efforts to develop alternative ways of describing uncertainty- such as with probability distributions or scenario approaches no alter- native has gained widespread recognition as acceptable and practical within EPA and other regulatory agencies. It is difficult to characterize what is known about uncertainty without making the risk appear either larger or smaller than analysts believe it to be (see, e.g., Johnson and Slovic, 1995~. And the difficulties are not yet yielding to analysis: the more that characterizing uncertainty is debated as an analytical problem, the more complex it appears to be (see, e.g., Finkel, 1990; National Re- search Council, 1994a:Chapter 9~. Characterizing uncertainty analytically puts risk analysts on the horns of a dilemma: simple characterizations are likely to give an erroneous impression of the extent of uncertainty, but more careful and elaborate characterizations may be incomprehensible to nonspecialists and so un- usable by decision makers and some other participants. Like the problem of finding a single, unbiased summary of accepted scientific knowledge, the problem of summarizing uncertainty may have no technical solution. We believe, however, that a solution might be found in the processes that lead to a risk decision, processes that combine iterative deliberation and analysis and provide participants with enough understanding of uncer- tainty to appreciate where scientists agree and where they disagree. (The last section of this chapter outlines key issues in process design; Chapter 4 presents a more detailed discussion of understanding uncertainty.) 4For an illuminating exchange of views on the "conservatism" issue, see Finkel (1994) and McClellan and North (1994).

OCR for page 37
68 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOCRATIC SOCIETY Specific Populations The question of who is at risk is important both for decision makers and the persons whose health and safety is of concern (Vaughan and Seifert, 1992; see also Konheim, 1988; Nelkin, 1989~. In presenting aggre- gate risk estimates to community residents who are concerned about a hazard, officials often fail to answer the important question for many interested and affected parties: "What does this mean for me or my family?" (e.g., Sharlin, 1986~. When confronted with statistical risk esti- mates, people often seek to reframe the question in terms of personal risk (Plough and Krimsky, 1987; Siegel and Gibson, 1988~- an issue not ad- dressed by the aggregate numbers. If a risk characterization does not address such questions, both it and the analysis behind it may be discred- ited by participants in the decision process. If it does address them, different ways of framing the same information can create different un- derstandings. Such failings often arise when a risk characterization assumes that estimates of risks to one population are sufficient to answer risk questions about what may be a different population. For example, the 1989 contro- versy over the use of Alar on apples centered on the risks to children, who drink more apple juice than adults, but the standard was set on the as- sumption that an adult male weighing 70 kilograms was an adequate surrogate for everyone (lasanoff,1987~. The risk reduction expected from vaccination programs is usually presented for the entire population al- though rural children, who live far from treatment centers, may not re- ceive as much benefit as other children because vaccination programs may not reach them. Migrant farm workers and their families may be inadequately protected against workplace risks because standards are based on exposure under very different working conditions (Vaughan, 1993a). To adequately summarize risks to some populations may require be- havioral analyses as well as the traditional analyses of exposure and sen- sitivity. For example, different groups of farm workers who are exposed to pesticides vary in their ability to understand warning materials and in their propensity to take self-protective action when given the opportunity (Vaughan, 1993a). Knowledge about reading ability and the psychologi- cal factors underlying self-protective behaviors are not usually incorpo- rated in risk characterizations although they can obviously affect both the risks to exposed individuals and the effectiveness of options to reduce those risks. Thus, a risk characterization that fails to carefully consider which populations the estimate is for may be inadequate to inform deci . . . slon magma.

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS Multiple Exposures 69 Risk analyses for multiple exposures are often based on the assump- tion that risk from genotoxic carcinogens are additive and that non-cancer risks are not, unless the agents operate by similar mechanisms. The de- gree to which this assumption holds is subject to debate because of the limited data available to address it. It is even less clear how well the assumption may apply for combinations of biological, chemical, and phy- sical risks to an ecosystem. EPA has considered multiple chemical eco- logical risks, but only at certain sites and within narrow bands. Although the agency also recognizes physical hazards and those that arise from management practices (U.S. Environmental Protection Agency, 1992a), it has not considered their possible interactions. For humansr evidence of serious drug interactions suggests that even with chemical hazards there are some instances in which the assumption of additivity may be ques- tionable. Increasing concerns about synergistic effects warrant careful consideration of how to address them in risk characterizations. A related issue is the treatment of past exposures or the past health conditions of some of the population at risk. This issue may arise as one of health (e.g., the possibility that past exposures have synergistic effects with present ones) or of equity. As the Chester, Pennsylvania, case sug- gests, residents of an industrial community who believe that they have already had more than their share of exposure to chemical risks may demand on equity grounds that past exposures be considered as part of the risk characterization. Communication The success of a risk characterization depends on its effective delivery to the participants in a risk decision. Typically, not all participants will understand a risk message in the same manner. Analogies are often used to make risk summaries more understandable, but analogies are usually very specific and sometimes depend on culture, status, age, gender, and other characteristics for their interpretation. If the manner in which the risk message will be interpreted by different groups or participants is not considered, uneven risk protection across groups could result (Vaughan, 1993a, 1995~. Another even more fundamental problem is that of compre- hension. Non-English-speaking people obviously get no benefit from a risk characterization in English. Messages prepared in written form will be ignored by people who cannot read or who are used to receiving information in other forms. The history of interaction between an organization that is presenting a risk characterization and the interested and affected parties can be an

OCR for page 37
70 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOC~TIC SOCIETY other source of communication problems (e.g., Krimsky and Plough, 1988~. A party that has had unsatisfactory experiences with that organi- zation or that issue may simply be unreceptive to new information from that source. For example, in a decision-making process to permit experi- mental land application of sewage sludge on farmland, an elaborate pub- lic involvement process collapsed partly because the same community had been involved in a landfill siting controversy just one year earlier. A widespread belief that the community was targeted as a "dumping ground" overpowered any positive reaction to the public involvement plans. People in the community organized against the new proposal partly because of ill feelings toward the state regulatory agency (Renn et al., 1991~. CONCLUSION: THE IMPORTANCE OF PROCESS DESIGN This chapter surveys the variety of judgments made sometimes, without careful consideration in the course of analyzing and character- izing risk that can become lightning rods for controversy. They become problematic when they conflict with the judgments of some of the inter- ested and affected parties to a decision, so that the resulting risk charac- terization does not address these parties' needs. The best way to prevent such problems, we believe, is not to call all such judgments into question in every decision process. Doing this would make risk analysis and char- acterization inordinately complex and resource intensive. We believe the best preventive is to devise analytic-deliberative processes that will pay appropriate attention to the judgments involved in problem formulation and the other tasks, inform these judgments with the best available knowl- edge and the perspectives of the spectrum of decision participants, and thus guide risk characterizations toward addressing the needs of the deci slon. When understandings of risk depend on potentially controversial judgments, it seems prudent to involve those who are likely to be affected by the decisions that rely on those judgments. If, for instance, there are many scientifically defensible ways of counting deaths and if the choice has serious implications for the concerns of some of the interested and affected parties, it makes sense to involve those parties in selecting the measures of death that will be used to characterize risk. Organizations responsible for characterizing risk should anticipate the value-based judg- ments that are likely to become contentious in the context of a particular risk characterization and consider putting them on the agenda for the analytic-deliberative process. We emphasize strongly that improved risk characterization based on

OCR for page 37
JUDGMENT IN THE RISK DECISION PROCESS 71 a better designed process will not eliminate conflict about risk. The best it can hope to do is to eliminate or reduce those conflicts that are based on misunderstandings, mistrust, miscommunication, inadvertent neglect of a point of view, and the like. It might be said that although good practice does not predictably lessen conflict, bad practice predictably increases it. Designing an analytic-deliberative process involves many choices. Who should be involved in the tasks that support risk characterization, beginning with problem formulation? In what ways and through what procedures should they be involved? At what points in the process should they be involved? Under what conditions should past assumptions, con- clusions, or decisions be reconsidered? These choices can affect the ulti- mate content of a risk characterization, the ways participants in a decision understand the risks, and acceptance of the process. Federal agency officials with a legislative mandate to protect the pub- lic against dangerous exposures to a toxic substance commonly respond to preliminary evidence of a possible hazard by directing toxicologists, epidemiologists, and other technical experts on the hazard to estimate the health risks associated with the substance. The process involves these experts, agency officials and policy makers, attendees at any required public hearings (whose ideas may or may not be given serious consider- ation), and any legislators and interest groups that know about the pend- ing decision and are able to gain access to the process. This standard process often leads to objections from interested and affected parties that they have been disenfranchised, that their ideas have been ignored, that their concerns have not been taken seriously, that the risk analysis was incomplete or irrelevant, that the analyses are so complex and arcane that they cannot participate meaningfully, and so forth in short, serious dis- affection with the process and the resulting risk characterization. Such outcomes have led many observers to recommend increased public involvement in risk decision making, better two-way communica- tion between agencies and interested and affected parties, involvement of these parties early in the decision process, and other changes that would make risk decision making processes more broadly participatory (e.g., Kunreuther, Fitzgerald, and Aarts, 1993; Leroy and Nadler, 1993; Slovic, 1993a; National Research Council, 1994b). We agree that more complete involvement of interested and affected parties in risk characterization is often essential for improving the process. It can also be essential for arriving at sound analyses. We note here some key principles of increas- ing meaningful participation in risk characterization that are developed in more detail in the next several chapters: give explicit attention to the design of the process that informs risk . . . clec~s~ons;

OCR for page 37
72 UNDERSTANDING RISK: INFORMING DECISIONS IN A DEMOCRATIC SOCIETY solicit and seriously consider input from the interested and af- fected parties as appropriate at various points in the process leading to risk characterizations; and plan for iteration in the decision process, that is, for reconsidering past assumptions, conclusions, and process-related decisions on the basis of new data and changes in the decision situation. We reiterate that risk characterization is more than a synthesis of information developed by analytical techniques. Analysis has inherent limitations in the face of the multidimensional and value-laden nature of many risk decisions. The success of risk characterization depends not only on doing and describing analysis well, but also on choosing analyses that address the needs of decision participants and on making the choice through a process that those parties trust. Organizations responsible for characterizing risks should plan to blend analysis with deliberative pro- cesses that clarify the concerns of interested and affected parties, help prevent avoidable errors, offer a balanced and nuanced understanding of the state of knowledge, and ensure adequately broad participation for a given risk decision.