Aversion to Risks:Public Paranoia or Technological Oppression?

Industry spokespersons attribute the blocking of oil refineries, nuclear reactors, and toxic waste dumps to public ignorance and mass paranoia. They charge that misguided and irrational citizens have successfully delayed so many technological facilities, driving up their costs, that wise investors now avoid them.8

Pete Seeger, however, has another story. He and the members of the Clamshell Alliance, as well as many other environmental and consumer activists, would claim that, just as the people created the moral victories won by the civil rights movements and the Vietnam protests, so also the people have successfully challenged potential technological oppressors. In their view, just as the people rejected a war fought without their free, informed consent, they also are rejecting public environmental risks likewise imposed on them without their free, informed consent. For them, to delay or stop construction of risky industrial facilities is a great moral triumph for populist democracy.

Industry sympathizers do not agree. They claim that laypersons' aversion to societal risks stems not so much from any real or apparent danger, such as toxic waste contamination, but from group attitudes that are anti-industry, antigovernment, and antiscience. They charge that the paranoid, neo-Luddite baby boomers who now dominate the environmental movement cut their political teeth during the Vietnam-era protests and then went on to become Yuppie lawyers, professors, and social workers. Holding their earlier political beliefs, they have merely transferred their activism from military to environmental issues. Thus, Pete Seeger now sings about "nukes," not "Nam." And Seeger's hair has turned gray, while the baby boomers long ago cut theirs, probably for an important job interview.9

Who is right? Is public aversion to societal risks caused by mass

paranoia and ignorance of science? Or by yet another form of oppression inflicted by "big industry," "big technology," and "big government"? Not surprisingly, I shall argue that the correct answer lies between these two extremes. Despite a regrettable and widespread ignorance of science, nevertheless environmentalism is not merely the product of an irrational "construct." Despite rampant technological illiteracy, irrationality is not the sole explanation of typical public aversion to involuntarily imposed societal risks. Likewise, it cannot account for widespread distrust of technologies having the potential to cause catastrophic accidents and increased cancers.

The main purpose of this volume is to sketch a middle path between the industrial charges of scientific illiteracy and the populist charges of technological oppression. In so doing, I shall argue for an alternative approach to contemporary, societally imposed risks. My focus is not on personally chosen risks, like diet drinks or oral contraceptives, since each of us is able to avoid such hazards. If my analysis is correct, then we need a new "paradigm," a new account of when the acceptance of public hazards is rational. We also need to recognize that laypersons are often more rational, in their evaluation of societal risks, than either experts or governments appear to have recognized.

The Rise of Risk Assessment and Evaluation

As Chapter Four will explain in greater detail, government and industry experts perform most risk or hazard assessments. Their analyses include three main stages: (1) identification of some public or societal hazard; (2) estimation of the level and extent of potential harm associated with it; and (3) evaluation of the acceptability of the danger, relative to other hazards.10 (Most of the discussion in this volume will focus on the third stage,risk evaluation. ) Once assessors have completed these three assessment tasks, policymakers then determine the best way to accomplish risk management of a particular public threat—for example, through regulation, prohibition, or taxation.

As a specific tool for societal decisionmaking, risk or hazard analysis is relatively new. Although Mesopotamian priests, before the time of Christ, regularly evaluated the impacts of proposed technological projects, risk assessment as a "developing science" did not arise until the late 1960s and the early 1970s.11 Public concern about the human and environmental risks of thousands of technologies arose in part because of tragedies like Love Canal and because of works like Rachel Carson's Silent Spring. 12 Another important milestone in raising environmental consciousness was the Club of Rome's famous 1972 report, Limits to

Growth. It predicted global human, economic, and environmental catastrophe in the twenty-first century unless we were able to stop exponential increases in pollution, resource depletion, population, and production.13

Widespread worries about impending environmental catastrophe and a rapidly increasing cancer rate were evident as early as 1969, as evidenced by the passage of the U.S. National Environmental Policy Act (NEPA), "the Magna Carta of environmental protection."14 NEPA required, among other things, that all federal agencies prepare an environmental impact statement (EIS) every time they considered a proposal for federal actions significantly affecting the quality of the environment.

In addition to the passage of NEPA, much risk-analysis effort also arose as a direct consequence of the creation of new federal agencies, such as the Occupational Safety and Health Administration (OSHA). Pressured by growing public concern about environmental risks, and faced with approximately 100,000 occupation-induced fatalities per year,15 the United States created OSHA in 1970. Many of the first hazard assessments—for example, regarding asbestos—were done under the direction of OSHA or other federal regulatory agencies, such as the Food and Drug Administration (FDA) and the Nuclear Regulatory Commission (NRC).

One of the main difficulties with risk assessments done in the 1970s and 1980s, however, was that there were inadequate standards for the practice of this new set of techniques. As a consequence, some hazards, such as carcinogens, were being monitored and regulated very stringently, whereas others, equally dangerous, were evaluated more leniently. To help address these methodological inconsistencies and regulatory difficulties, in 1982 the U.S. Congress passed the Risk Analysis Research and Demonstration Act (RARADA). This bill established a program, under the coordination of the Office of Science and Technology Policy, to help perfect the use of hazard assessment by federal agencies concerned with regulatory decisions related to the protection of human life, health, and the environment.16 Numerous risk assessors, prior to the RARADA, bemoaned the fact that government safety regulations for the automobile industry, for example, presupposed an expenditure of $30,000 for the life of each automobile passenger saved, whereas analogous government regulations for the steel industry pre-supposed an expenditure of $5 million for the life of each steelworker saved.17

Despite the passage of the RARADA, however, quantitative risk assessment is still practiced in somewhat divergent ways; for example,

the monetized "value of life" presupposed by government regulations (and used at the third, or evaluation, stage of assessment) varies dramatically from one federal agency to another.18 This value, in turn, has a great effect on the acceptability judgments associated with various risks, particularly if the evaluation is accomplished in a benefit-cost framework. Even for the same hazard, risk analyses often do not agree, in part because there are many ways to evaluate harms at the third stage of assessment. There are many ways to answer the question "How much risk (in a given area) is socially, politically, economically, and ethically acceptable?" Hazard evaluations often contradict one another, not only because scientists frequently dispute the relevant facts but also because policymakers and the public disagree about what responses to risk are rational. Some persons claim that only technical experts are capable of making rational judgments about risk acceptability, whereas others assert that only potential victims, usually laypeople, are in a position to be truly rational about evaluation of possible hazards.

"Rationality" in Risk Evaluation and Philosophy of Science

'Rational', however, is a highly normative term. Controversies about the "rationality" of various evaluations of risk are no easier to settle than analogous debates in science. Conflicts among philosophers of science (about what methodological rules, if any, guarantee the rationality of science) generate alternative accounts of scientific explanation, as well as disputes over which scientific theory is correct. Likewise, conflicts among risk assessors (about what methodological rules, if any, guarantee the rationality of responses to hazards) generate both alternative accounts of acceptable harm and disputes over whose risk-evaluation theory is correct.

In the debate over the rationality of science, philosophers and scientists are arrayed on a spectrum extending from pluralist or relativist views to logical-empiricist positions. At the left end of the spectrum, the pluralist end, are epistemological anarchist Paul Feyerabend and others who believe that there is no scientific method, that "anything goes," and that "no system of [scientific] rules and standards is ever safe."19 At the other end of the spectrum are logical empiricists, such as Israel Scheffler and Rudolf Carnap, who believe that there are at least some universal and fixed criteria for theory choice and that these criteria guarantee the rationality of science.20 Somewhere in the middle, between the relativists and the logical empiricists, are the so-called naturalists, such as Dudley Shapere, Larry Laudan, and Ronald Giere. They

maintain that theory evaluation can be rational even though there are no absolute rules for science, applicable in every situation.21

The challenge, for any philosopher of science who holds some sort of middle position (between the relativists and the logical empiricists), is to show precisely how theory choice or theory evaluation can be rational, even though there are no universal, absolute rules of scientific method that apply to every situation. Perhaps the dominant issue in contemporary philosophy of science is whether, and if so how, one can successfully develop and defend some sort of naturalistic middle position, as Larry Laudan, Ronald Giere, and Thomas Kuhn, for example, have tried to do.22

An analogous problem faces the hazard evaluator trying to articulate a middle position. In the debate over what methodological norms, if any, guarantee the rationality of risk evaluation, analysts are arrayed on a spectrum extending from the relativists to the naive positivists. At the left end of the spectrum are the cultural relativists,23 such as anthropologist Mary Douglas and political scientist Aaron Wildavsky. They believe that "risks are social constructs," that "any form of life can be justified. . . . no one is to say that any one is better or worse,"24 that there is "no correct description of the right behavior [regarding risks],"25 and therefore that the third stage of risk assessment, risk evaluation, is wholly relative.26 At the other, naive-positivist, end of the spectrum are engineers such as Chauncey Starr and Christopher Whipple. They maintain that risk evaluation is objective in the sense that different risks may be evaluated according to the same rule—for example, a rule stipulating that risks below a certain level of probability are insignificant.27 They also claim that risk assessment, at least at the stage of calculating probabilities associated with harms and estimating their effects, is completely objective, neutral, and value free.28 In their view, the objectivity of risk identification and estimation guarantees the rationality of specific evaluations of various hazards.

The challenge, for any risk evaluator who holds some sort of middle position (between the cultural relativists and the naive positivists), is to show how risk evaluation (the third stage of assessment) can be rational and objective, even though there are no completely value-free rules applicable to every risk-evaluation situation. My purpose in this volume is (1) to articulate why and how both the cultural relativists and the naive positivists err in their general accounts of risk evaluation; (2) to explain the misconceptions in a number of specific risk-evaluation strategies allegedly deemed "rational"; and (3) to argue for a "middle position" on the methodological spectrum of views about how to guarantee the rationality of risk evaluation. I call this middle position "sci-

entific proceduralism," and I defend it by means of arguments drawn from analogous debates over naturalism in contemporary philosophy of science.

Outline of the Chapters: Risk Evaluation Is Both Scientific and Democratic

In Chapter Two, "Science against the People," 1 introduce the problem of conflict over rational evaluations of risk. Specifically, I show how the cultural relativists and the naive positivists have wrongly dismissed lay evaluations of risk as irrational. The bulk of this chapter focuses on faulty epistemological assumptions underlying relativist and naive-positivist arguments about risk evaluation.

After defusing these arguments against "the people," in Chapter Three ("Rejecting Reductionist Risk Evaluation") I analyze in greater detail the two most basic risk frameworks out of which such antipopulist arguments arise. I show that both of these frameworks, naive positivism and cultural relativism, err in being reductionistic. The cultural relativists attempt to reduce risk to a sociological construct, underestimating or dismissing its scientific components. The naive positivists attempt to reduce risk to a purely scientific reality, underestimating or dismissing its ethical components. I argue that the sociological reductionists err in overemphasizing the role of values in risk evaluation, whereas the scientific reductionists err in underemphasizing the role of ethical values and democratic procedure in risk evaluation.

Because locating the flaws in accounts of rational risk evaluation comes down to clarifying the appropriate role of values at the third stage of hazard assessment, Chapters Four and Five attempt to provide a general overview of the various evaluative assumptions that are integral to all three stages of risk analysis. Chapter Four ("Objectivity and Values in Risk Evaluation") shows that—despite the presence of cognitive or methodological value judgments, even in pure science—science itself is objective in several important senses. After outlining the value judgments that arise in the three stages of risk assessment (risk identification, risk estimation, and risk evaluation), the chapter presents a case study from the field of energy studies. The case study shows how alternative value judgments at the first two stages of assessment can lead to radically different policy conclusions (third stage) regarding hazard acceptability.

Chapter Five ("Five Dilemmas of Risk Evaluation") shows how epistemic value judgments arise in the more scientific stages of risk assessment (viz., risk identification and estimation). It sketches some

analogous difficulties arising at the third, or risk-evaluation, stage. It argues not only that methodological value judgments are unavoidable in risk evaluation, but also that the judgments often pose both methodological and ethical dilemmas, problems for which there are no zero-cost solutions. Chapters in the last section of the volume (Part Three) argue that these five dilemmas raise troubling ethical questions and thus provide a basis for improving hazard evaluation.

Whereas the first part of the book (Chapters One through Five) provides an overview of risk analysis and evaluation and a discussion of the flaws in the two most general accounts of hazard evaluation, Part Two of the book (Chapters Six through Ten) addresses more specific difficulties in risk evaluation. Each chapter in Part Two evaluates a questionable methodological strategy common in various methods of risk evaluation.

Chapter Six, "Perceived Risk and the Expert-Judgment Strategy," argues that risk assessors' tendencies to distinguish "perceived risk" from "actual risk" are partially misguided. Typically, they claim that only "actual risk" (usually defined by experts as an average annual probability of fatality) is objective, whereas "perceived risk" (based merely on the feelings and opinions of laypersons) is subjective. I argue that both "perceived risk" and "actual risk" are partially subjective, since both involve value judgments. Further, I suggest that the evaluation stage of risk assessment will be more successful if analysts do not overemphasize the distinction between perceived and actual risk. Instead, they should focus on mediating ethical conflicts between experts and laypeople over risk evaluation.

Continuing in a similar vein, Chapter Seven also identifies a problematic strategy associated with the attempt to define risk in a largely quantitative way. This chapter, "Democracy and the Probabilistic Strategy," attacks two common methodological assumptions about risk evaluation. One is that risk abatement ought to be directed at the hazards to which persons are most averse. The other is that risk aversion ought to be evaluated as directly proportional to the probability of fatality associated with a particular hazard. After arguing against this probabilistic strategy, I show that other, nonprobabilistic criteria for risk evaluation (equity of risk distribution, for example) are equally plausible, in part because accurate knowledge of probabilities is sometimes difficult to obtain. If one employs these other criteria, I argue, one can conclude that technical experts should not be the only persons chosen to evaluate risks and therefore dictate which societal hazards are acceptable. Control of risk evaluation needs to become more democratic.

Chapter Eight, "Uncertainty and the Utilitarian Strategy," argues that, in many risk evaluations, it is more reasonable to pursue a "maximin" strategy, as most laypersons request, rather than the utilitarian (or Bayesian) strategy used by most experts. The main argument of this chapter is that, in situations of uncertainty, Bayesian accounts of risk evaluation are often unable to provide for considerations of equity and democratic process.

Chapter Nine, "Uncertainty and the Producer Strategy," addresses another problematic method of risk evaluation, one closely related to Bayesianism. The chapter asks whether, in a situation of uncertainty, one ought to implement a technology that is environmentally unsafe but not recognized as such (thereby running a "consumer risk") or fail to implement a technology that is environmentally safe but not recognized to be so (a "producer risk"). In cases of doubt, on whose side ought one to err? Chapter Nine argues that there are scientific, ethical, and legal grounds for minimizing consumer risk and maximizing producer risk, especially in cases of uncertainty.

Just as experts tend to overemphasize producer risk and underemphasize consumer risk, they also tend to discount hazards that are spatially or temporally distant. Because of this "discounting" tendency, risk assessors in developed countries often ignore the hazards their nation imposes on those in underdeveloped areas. I call this tendency the "isolationist strategy." Chapter Ten, "Third-World Risks and the Isolationist Strategy," argues that this risk-evaluation strategy is unethical.

Discussion of the isolationist strategy in Chapter Ten marks the end of the second part of the volume. Although this second section criticizes several of the problematic risk-evaluation methods (such as the probabilistic strategy, the Bayesian strategy, and the isolationist strategy) employed both by contemporary hazard assessors and by moral philosophers, it provides neither a technical nor an exhaustive account of all the questionable risk-evaluation methodologies.29 Instead, its purpose is both to provide an overview of representative risk-evaluation errors (the strategies criticized in Part Two) and to cast doubt on the thesis that expert assessment dictates the only risk evaluations that are 'rational'. Instead, rational risk evaluation and behavior may be more widely defined than has been supposed. And if so, there are grounds for doubting experts' claims that lay responses to, and evaluations of, societal risks are irrational.

Together, the chapters in the first and second sections of the volume provide an overview of much of what is wrong with contemporary hazard assessment and with allegedly rational risk evaluation. The pur-

pose of the third section is to sketch some solutions to the problems outlined in the two earlier parts of the book. Chapter Eleven, "Risk Evaluation: Methodological Reforms," makes a number of specific suggestions in this regard. It begins by offering an alternative risk-evaluation paradigm, "scientific proceduralism." According to this paradigm, risk evaluation is procedural in that it ought to be guided by democratic processes and ethical principles. It is scientific or "objective" in at least three senses: (1) It can be the subject of rational debate and criticism. (2) It is partially dependent on probabilities that can be affected by empirical events. (3) It can be criticized in terms of how well it serves the scientific end or goal of explaining and predicting hazardous events and persons' responses to them.

After arguing that risk evaluation is largely objective, because it is based in part on probabilities, and because it is assessed on the basis of its explanatory and predictive power, I also argue that risk evaluation ought to be defined in terms of social and ethical values. Explaining how risk evaluation can be both objective and evaluative, Chapter Eleven outlines a number of specific suggestions for methodological improvements in hazard evaluation—for example, the use of ethically weighted risk-cost-benefit analysis (RCBA) and the ranking of experts' risk opinions on the basis of their past successful predictions. The chapter also substantiates the claim that hazard assessment—although burdened both with reductionist definitions of risk (Chapter Three) and with a number of biased methodological strategies (Chapters Six through Ten)—is objective in important ways. Hence, it makes sense to continue to use quantified risk analysis (QRA). That is, although in practice problems of risk evaluation have led to poor policy, in principle they are capable of being solved by means of improved risk-evaluation methods and more participatory styles of hazard management.

Having considered the methodological solutions to some of the difficulties with QRA and risk evaluation, 1 conclude by addressing certain procedural and institutional reforms needed to make risk management more rational. Chapter Twelve, "Risk Management: Procedural Reforms," argues that, consistent with a more naturalized view of all knowledge, we must place less emphasis on whose hazard evaluations are correct or incorrect and instead focus on negotiating workable risk-management principles and practices. In addition, we ought to make use of several insights from medical ethics, such as requiring free, informed consent prior to imposing risks; guaranteeing legal rights to due process and compensation for all unavoidable risks and harms; and applying the theory of "market share liability," as in the celebrated DES case.

These chapters come nowhere close, of course, to providing a complete explanation of what makes a risk evaluation rational. This account does not pretend to be complete, in part because the problems of risk evaluation are too numerous to be treated in a single, nontechnical volume, and in part because I attempted (as far as possible) to avoid repeating analyses given in my earlier works.30 These chapters will have accomplished their modest aim if they enable us to be more critical of existing attempts to define "risk evaluation" in highly stipulative and question-begging ways. They will have taken us in the right direction if they teach us to be suspicious whenever someone gratuitously attributes motives and causes to those allegedly exhibiting "irrational" risk evaluations. They will have helped us if they encourage us to consider alternative models of rationality and to remember that chronic errors in risk-evaluation heuristics are not limited to laypeople alone.31 This being so, determining when a risk evaluation is rational is as much the prerogative of the people as of the experts. Science need not co-opt democracy.

Attacks on the Public's Aversion to Risk

At least three groups of persons maintain that citizens' worries about environmental risks, from carcinogenic pesticides to loss of global ozone, are typically biased or irrational: (1) industry spokespersons, (2) risk assessors, and (3) a small group of contemporary, antipopulist social scientists. All three have attacked the environmental fears of laypeople. Industry spokesperson Edith Efron, for example, maintains that both citizens and scientists have been corrupted by ideology. She says that the ideology takes the form of attempting to incriminate industry in the name of "cancer prevention." The politics of citizens, she says, derive from "warring attitudes toward the American industrial system." In Efron's view, most persons who fear environmental hazards (such as alleged carcinogens and nuclear power) are irrational because their concern is dictated not by the facts about risk but by their paranoid and primitive "fantasies" about "industrial mass murder."13

Risk assessors, often experts in the employ of industries responsible for the hazardous technologies they are paid to evaluate, constitute a second class of persons critical of alleged citizen "irrationality." Norman

Rasmussen and Christopher Whipple, for example, each authors of famous risk analyses, have accused the public of "inconsistency" in its attitudes toward hazards.14 In their view, people who travel by automobile and, at the same time, oppose commercial nuclear fission are inconsistent, because they accept a large risk but reject an allegedly smaller one. Other hazard assessors claim that, if laypeople understood the relevant mathematics involved in calculating risk probabilities, they would no longer have "pathologic fear" of contemporary technologies like atomic energy. In other words, risk assessors claim that there are few rational reasons for fearing existing environmental threats, and that public concern is primarily a matter either of irrationality or of ignorance.15

A minority group of contemporary, antipopulist social scientists constitutes the third main camp of persons who are critical of lay evaluations of technological and environmental risks. Perhaps the most famous members of this group are anthropologist Mary Douglas and political scientist Aaron Wildavsky. Although their individual views are somewhat different, they are coauthors of a best-selling book, Risk and Culture. 16 In his Searching for Safety and in the coauthored Risk and Culture, Wildavsky argues, just as Weinberg does, that Americans are biased, witch-hunting opponents of technology. He and Douglas claim that laypersons ("sectarians" not at the center of industrial or governmental power) are dominated by "superstitions" about environmental risks and by fundamentalist desires for unrealistic environmental "purity." Like Hoyle, Fremlin, Weinberg, Thompson, and other critics, they allege that these contemporary "superstitions" and "biases" of contemporary environmentalists are no different in kind from those of pre-scientific, primitive people.17

Admitting that they have a "bias toward the center,"18 Wildavsky, Douglas, and others (such as Efron and Thompson) claim that "America is a border country," a nation whose citizens, in general, are "sectarians" who reject the technological and environmental risks imposed by those at the economic and political "center" of the nation.19 They identify Americans, in general, as environmentalists and sectists. Hence, when they attack U.S. environmentalists and sectarians, they are attacking the U.S. lay public as a whole.

Numerous industry spokespersons, engineers, hazard assessors, and natural and social scientists tend to use at least five basic arguments to attack the societal risk evaluations of laypersons: (1) Laypeople, the "border" citizens, are anti-industry and antigovernment and are obsessed with environmental impurity. (2) Laypeople are removed from centers of influence and power, and therefore attack the risks chosen

by those who are at the "center." (3) Laypeople are unreasonably averse to risks because they fear things that are unlikely to occur, and they are unwilling to learn from their mistakes. (4) Laypeople are irrationally averse to risks because they do not realize that life is getting safer. (5) Laypeople have unrealistic expectations about safety and make excessive demands on the market and on hierarchies of power.20

Not all those with antipopulist risk views are consistently antisectarian.21 Douglas, for example, is not antisectarian in some of her other works.22 However, in Risk and Culture, Douglas and Wildavsky specifically admit that they are biased in favor of centrists (nonsectarians; those with market and hierarchical views); Wildavsky makes the same admission in his Searching for Safety. 23 Apart from whether Thompson, Wildavsky, or Douglas, for example, gives an antisectarian account of risk aversion, it is clear that many other risk writers do so (for example, Maxey, Cohen and Lee, Whipple, Weinberg, Rothschild, Hoyle, Efron, and Fremlin). Hence, the five arguments, apart from their specifics, represent paradigmatic attacks on lay evaluations of societal risk. Because they are representative, it is important to assess them.24

The Need for a New Theory of Rationality

If the preceding analysis of arguments (for the bias and irrationality of lay responses to societal risk) is correct, then many risk assessors, industry spokespersons, and natural and social scientists have built their view of "rational" risk evaluation on a questionable foundation. Without significant evidence, they have attributed motives to environmentalists who are averse to risk. They have confused causes of behavior with correlations; they have defined 'rational' and 'environmentalist' in highly stipulative, question-begging ways.

If such arguments against lay views of environmental risk are wrong, then we may need a new theory of rationality.70 We also may need new criteria for rational risk evaluation. After showing (in the next chapter) how risk experts subscribe to outmoded positivistic or relativistic notions of rationality, I shall sketch a new account of rational risk evaluation. Its main policy goal will be to enfranchise the very people who are most likely to be victimized by environmental hazards.

The Case for Redefining "Rationality"

Echoing the famous description in A Tale of Two Cities, many people would describe our own era as both the best of times and the worst of times. We are sending humans into space, creating new life through biotechnology, and bombarding ever smaller subatomic particles into their elementary components. We have been less adept at securing world peace, feeding the starving, and protecting the environment. Some persons might even say, despite our medical advances, that we have not been wholly successful even in saving lives. In 1982, the National Cancer Institute warned that one in every three Americans would die of cancer. Cancer already takes more American lives, each year, than were lost during all of World War II, the Korean War, and the Vietnam War combined; it is responsible for eight times as many annual deaths as automobiles.1

Apart from the magnitude of such statistics, at least three facts about cancer deaths are significant. For one thing, many epidemiologists claim that about one-third of the fatalities are correlated with cigarette smoking. Also, the deaths are occurring among the young, not merely among typical elderly victims; cancer is the leading cause of death for all U.S. children between the ages of one and ten and for all women between the ages of thirty and forty. It is second only to accidents as the leading cause of death for all Americans under the age of thirty-five. Cancer rates for all age groups have been increasing in the last three decades. Most important, the U.S. Office of Technology Assessment claims that up to 90 percent of all cancers are "environmentally induced" and hence "preventable."2

If hundreds of thousands of cancer deaths are preventable, why aren't they stopped? Part of the answer has to do with human ignorance

of the hazards we create. Most people are probably unaware, for example, that the United States annually generates one ton of hazardous wastes for every man, woman, and child; that 90 percent of the 80,000 known contaminated sites for toxic wastes threaten groundwater; and that cancer rates near such dumps are typically 50 percent above average. Similarly, most American mothers probably don't know that their breast milk is so contaminated with toxins that it could not be sold in supermarkets.3 Likewise, most people probably don't know that legal emissions of only four radioactive isotopes from the U.S. commercial nuclear program during the next hundred years will cause over 30,000 deaths4 even if there is no catastrophic reactor accident, which alone could cause 145,000 fatalities.5

If people are unaware of risks close to home, risks carried by their own bodies and in the groundwater they drink, they are likely even less cognizant of hazards farther from home: 25,000 people are killed daily, worldwide, by water contamination and shortage.6 In Africa alone (where 47 percent of the land is too dry for agriculture and where one million babies are born every three weeks), a million and a half persons have died of famine since 1950.7 Sixty percent of the world's humans are cutting fuelwood faster than it can be replenished, and in countries like India 88 percent of the forestland has been lost since 1950. As a consequence, 28 percent of the ice-free surface of the earth is faced with desertification, drought, and famine.8 In fact, the Sudan Desert has recently been extending southward at a rate of three miles per year.9 Coping with the risks of desertification and famine, however, has increased the pressure to maximize farm yields, in part through use of chemicals. But such chemicals bring their own hazards. The World Health Organization (WHO) estimates that pesticides cause 50,000 deaths annually, many resulting because one-third of all U.S. pesticide exports, although banned in this country as unsafe, are marketed and used legally throughout the third world.10

As this catalogue of risks indicates, famine, firewood shortages, and pesticides are all interrelated. They are interrelated because virtually every benefit, from food and firewood to railroads and electricity, is bought at some environmental and human cost, such as pesticide risks. Perhaps we tolerate many of these hazards, from pesticides to deforestation to cancer fatalities, because we subscribe to a particular philosophy about the rationality of risk acceptance and risk trade-off.

In this chapter, I argue that at least two of the major positions regarding societal risk evaluation and acceptance are highly questionable, those of the cultural relativists and the naive positivists. 11 Proponents of

both positions err, in part, because they are reductionistic and because they view the hazard evaluations of citizens who are risk averse as either biased or irrational. After showing why both views of risk evaluation are built on highly doubtful epistemological presuppositions, I argue in favor of a middle position, which I call "scientific proceduralism." This Popperian account is based on the notion that objectivity in hazard evaluation (stage three of assessment) is tied to at least three things: (1) the ability of risk evaluations to withstand criticism by scientists and laypeople who are affected by the hazards; (2) the ability of risk evaluations to change, as new facts about probabilities are discovered; and (3) the ability of risk evaluations to explain and predict both risks and human responses to them. Because of point 1, the position is sympathetic toward many populist attitudes about involuntary and public hazards. Because of points 2 and 3, the position is sympathetic to the scientific need for objectivity in risk evaluation.12 Although scientific proceduralism is not the only reasonable view of risk evaluation, I argue that it is at least a rational one. I also argue that rational behavior should not be defined purely in terms of the evaluations of either the cultural relativists or the naive positivists. Most important, risk experts should not "write off" the common person.

Since hazard evaluation is dominated by these two questionable positions (cultural relativism and naive positivism), it is reasonable to ask whether criticizing them threatens the value of quantified risk assessment (QRA). Indeed, many of those allegedly speaking "for the people" are opposed to objective scientific and analytic methods of assessing environmental dangers. I am not. In Chapter Eleven, I shall argue that, although QRA is subject to various shortcomings and biases, it can and ought to be improved and used in policymaking; although it is in practice flawed, it is in principle necessary to rational, objective, democratic risk assessment.13

Three Philosophies about Risk Evaluation: Cultural Relativism, Naive Positivism, and Scientific Proceduralism

Contemporary controversy over rational evaluation of risks is, at the core, a conflict over values. Cultural relativists such as Douglas and Wildavsky, authors of the best-selling Risk and Culture, overemphasize value judgments in hazard evaluation (stage three of assessment). They claim that "risk is a collective construct."14 I shall argue that, in so doing,

they erroneously reduce the third stage of hazard analysis (risk evaluation) to anthropology or sociology, and they ignore its objective, scientific component. Naive positivists, such as Starr, Whipple, Hafele, Okrent, Jones-Lee, Morgan, and Rothschild, make the opposite mistake. They attempt to avoid value judgments in evaluating societal risks. In assuming that there are often value-free, mathematical criteria that are sufficient for evaluating risks, they presuppose that all risks can be objectively measured, independent of even methodological value judgments.15 I show that, by virtue of their position, they attribute a misguided (that is, wholly value-free) notion of objectivity to risk assessment.16

The naive positivists are wrong, as this chapter will argue, because risk estimates are not completely objective (in the sense of being wholly value free)—as is evidenced by the U.S. National Academy of Sciences' statement that most risk assessments are performed in situations of probabilistic uncertainty .17 As witness to this uncertainty, the current technological landscape is littered with the bodies of victims of various hazards. From Chernobyl to Bhopal, there are victims of risks that experts allegedly measured objectively, catastrophes that were not supposed to happen. They include radwaste dumps whose contents were "objectively" and definitively said not to be able to migrate18 and toxic chemicals that were "objectively" claimed not to cause harm to humans.19

But even though risk evaluation is not wholly objective, neither is it merely evaluative nor only a construct.20 Constructs don't kill people; faulty reactors, improperly stored toxics, and poor risk evaluations do. At least some hazards are real, and many of these are measurable. Hence, the cultural relativists are wrong in overemphasizing value judgments in risk assessment and evaluation. Risk evaluations assess real hazards, not just constructs.

In my view of risk evaluation, then, it is false to say that hazard assessments can be wholly value free (as many naive positivists claim), and it is equally false to assert (as many cultural relativists do) that any evaluation of risk can be justified.21 That is, some risk evaluations are more warranted, more objective, than others, although none is wholly value free. Later in the chapter, I shall specify some of the criteria for democratically reaching more objective risk evaluations. For now, let's investigate the position of the cultural relativists, and then that of the naive positivists, to see where they go wrong. This will provide a basis for defining a middle path between them, a new account of risk evaluation.

Cultural Relativism

Cultural relativists begin from an astute (although not original) insight. They recognize that risk estimation is not wholly objective (value free), and they criticize assessors for their repeated error in assuming that lay estimates of risk are mere "perceptions" whereas expert analyses are "objective."22 Clearly, the cultural relativists are to be commended for discussing this error, since both experts and the public have risk "perceptions," and no one has privileged access to the truth about risk acceptability merely because she is an engineer rather than a housewife. Even though the preponderance of technical errors made in estimating hazards obviously lies with laypeople, not experts, the cultural relativists are correct in affirming that engineers and housewives both employ value judgments, especially in evaluating risk acceptability.

Cultural relativists err, however, in assuming that, because everybody can be wrong at some time, everybody is wrong all the time. They say that any judgment or risk evaluation is merely a social construct. Some of these relativists reserve their harshest criticisms for the U.S. public. They claim that "risks are social constructs,"23 but they single out U.S. environmentalist or sectarian laypersons (as opposed to technical experts) as having particularly biased constructs.24

Of course, as I argued in the previous chapter, if cultural relativists believe that "no one is to say that one [risk judgment] is better or worse" than another,25 they are obviously inconsistent. It is inconsistent to affirm both that all risk evaluations are equally plausible and that environmentalists' risk positions are more questionable than those of the "centrists" (persons whose value systems, in their scheme, derive from the market or from hierarchical organizations).26 The last chapter outlined allegations against the risk evaluations of U.S. laypersons. This chapter examines why cultural relativism, especially in the area of risk evaluation, is wrong. Among other things, cultural relativism contributes to a proindustry bias toward risks, a bias that disenfranchises the lay public and supports the status quo.

Risk relativists tend to employ five arguments (all variants of themes advanced by writers such as Sumner, Engels, Rothschild, and Herskovits27 ): (1) Increased knowledge and additional reasoning about risks do not make people more rational about hazards.28 (2) Risk assessments are like judgments in aesthetics.29 (3) "Any form of life," including risk behavior and attitudes, "can be justified," since all people—including experts who disagree about hazard analysis—are biased in their perceptions of danger.30 (4) Modern persons are no different from "primi-

tives" (Douglas and Wildavsky's term) in that social structures dictate their views on, and responses to, alleged hazards.31 (5) More specifically, environmentalists' views on risk are a result of their "sectarian problems."32 Let's examine each of these arguments.

Naive Positivism

Interestingly, in reducing all risk determinants to social structures, cultural relativists share a common error, reductionism, with the naive positivists.69 Just as the cultural relativists attempt to reduce risk evaluations to sociological constructs, ignoring their objective, scientific content, the naive positivists attempt to reduce them to scientific rules, minimizing their ethical content.70 The cultural relativists overemphasize values in risk evaluation and assessment, whereas the naive positivists underemphasize them. Other positivists, such as Hempel, maintain a defensible, objective, empiricist approach without espousing this reductionism.71 Let's see why the naive positivists' account of risk objectivity goes awry.

A Path between Naive Positivism and Cultural Relativism

If public policy about risk is to address the real sources of lay unrest and expert disagreement, that policy must be based on an accurate

account of the reasons for the controversy. At least part of this unrest and disagreement, I have argued, is attributable to a misunderstanding about the role of values in hazard assessment. I have argued that hazard assessment can be objective and testable,102 but not wholly value free. Cultural relativists, but not "soft relativists,"103 overemphasize values and reduce all risk evaluations to mere sociological constructs, devoid of objective science. Naive positivists underemphasize values and attempt to reduce all risk evaluations to algorithms of allegedly pure science.104 They ignore the ethical part of risk evaluation. A more plausible account of scientific rationality and objectivity is, I believe, one that falls midway between the views of the cultural relativists and those of the naive positivists. In this view, which I call "scientific proceduralism," the rationality appropriate to risk analysis is best defined by three propositions: (1) that, as Hempel asserts, there is at least one general, universal criterion for theory or paradigm choice in beth science and risk assessment: explanatory power as tested by prediction; (2) that most of the remaining criteria for theory choice, although evaluated and interpreted on the basis of how well they serve explanatory power as tested by prediction, are situation specific or determined largely by practice (i.e., naturalistic); and (3) that we can best guarantee "scientific objectivity," as I define it, by testing the predictive and explanatory power of our risk theories and by subjecting risk evaluations to intelligent debate, criticism, and amendment by the scientific community and laypeople likely to be affected by the risk.105

Since I won't have time to defend all three propositions, I'll rely in part on the arguments of McMullin and Hempel for claim 1, that there is at least one general, universal criterion for theory choice in science and risk assessment—viz., explanatory power as tested by prediction.106 To see how this general criterion might operate, one need only recall some of the arguments made earlier in the volume. For example, I claimed that the theory of risk behavior adopted by risk relativists such as Douglas and Wildavsky fails in explanatory power because it relies on social determinants of risk aversion and does not provide an account of hazard aversion arising from personal and psychological variables.107 Likewise, the predictive power of their theory is flawed, as was already argued, since the Sierra Club is politically quite representative of the entire population.108

To understand how explanatory power as tested by prediction can function as a universal criterion or goal for theory choice in science and in hazard analysis,109 however, we have to understand why scientific objectivity does not require value-free empirical confirmability. It is not

reasonable to require value-free, empirical confirmability of all risk evaluations because such confirmability is not the only test of objectivity, either in science or anywhere else, and because essential methodological value judgments (in science) cannot be confirmed. Rather, we often call a judgment "objective," even in science, if it is not obviously biased or subjective. Objectivity, in this sense, is not tied to value-free empirical confirmability but, rather, is linked to avoiding bias values, to giving evenhanded representation of the situation. What I call "scientific objectivity" (and I say "scientific" not because the objectivity is unique to science, but simply because it is characteristic of science and related activities, such as making risk evaluations) is closely related to this sense of objectivity as evenhandedness.

Presumably, one could be blamed for failure to be objective, in the sense of not being evenhanded, if one were biased in a particular methodological value judgment about risk—for example, if one drew a conclusion about risk acceptability after deliberately misrepresenting some of the data on probability of the hazard. Since we do often blame people for not being objective—that is, for not being fair or evenhanded—it is clear either that objectivity in this sense must be attainable or that one can be more or less objective, contrary to what the cultural relativists believe.

But how might one guarantee scientific objectivity in the sense of avoiding bias values and providing evenhanded representation of the situation?110 Perhaps the best way to contribute to rationality and objectivity is to check the predictive and explanatory power of our hazard evaluations and to subject them to review by scientists and by laypersons likely to be affected by them. The assessments then might be said to possess "scientific objectivity," in some minimal sense, after they have been subjected to tests of predictive and explanatory power (see note 66), and to criticism, debate, and amendment. In this account, scientific rationality is served by a risk assessor working individually, pursuing a goal of explanatory power tested in part by prediction.111 Scientific objectivity, however, can only be guaranteed both by predictive and explanatory power and by scientists and affected laypersons working together to criticize and amend proposed accounts of risk acceptability.112

It seems reasonable to define "scientific objectivity" in terms of criticism, as well as predictive and explanatory power, especially because such criticism need not be wholly subjective. When I make an epistemic value judgment about two accident probabilities, or about the acceptability of a particular risk, for example, I am not talking merely autobiographically and subjectively. I am talking about characteristics of possible, public, external events that are capable of being known and

understood by other people. Moreover, the skills associated with making these judgments are a function of training, education, experience, and intelligence. Therefore, objectivity does not necessarily require one to have an algorithm guaranteeing the correctness of risk judgments. An algorithm is not necessary, in part because empirical factors (such as observed accident consequences or actual accident frequencies) could change the likelihood that our risk judgments are correct; hence, such judgments cannot be purely subjective. (That is why their predictive power can be tested.113 )

Finally, to make empirical confirmability alone (instead of predictive and explanatory power, for example) a necessary condition for the objectivity of all risk judgments would be to ignore the way that reasonable people behave. Reasonable people accumulate observations and inferences until the probability of their judgments is so great that they do not doubt them. They make assumptions when their evidence supports them; they do not rely exclusively on value-free empirical confirmation. Only if one were engaged in a search for objectivity that came close to avoiding all error would it seem reasonable to complain about well-supported risk evaluations that met the procedural criteria (predictive power, explanatory power, surviving criticism) for scientific objectivity just outlined.114 Since even science has never claimed such an unrealistic notion of objectivity (that came close to infallibility), it does not seem reasonable to demand more than the previous three criteria as standards for "objective" risk judgments.115 My new sense of "scientific objectivity" also seems plausible because it relies, in part, on the social and critical character of science and science-related activities (like risk assessment). As Karl Popper,116 William James, Ludwig Wittgenstein, and John Wisdom realized, the naive positivists make too strict a demand on scientific objectivity by requiring value-free confirmability, alone, in all evaluative situations.

The cultural relativists also make too strict demands by presupposing that "objective" risk estimates must be wholly value free and therefore infallible and universal.117 They believe that, because every risk evaluation is value laden and because both experts and laypeople disagree about hazards, no risk evaluations can be said to be objective. Because of the lack of consensus on risk evaluation, they claim that none is certain, that any of them can be said to be justified, and that one is as good as another.118 Although they are correct in recognizing the incompleteness of the naive positivists' model of rationality, the cultural relativists go too far in leaping to the premature conclusion that, because the naive positivists' criterion of value-free empirical confirmability fails, therefore risk evaluations are purely relative.119

Conclusion

More generally, if relativistic arguments requiring that all objective risk evaluations be based on universal and stable rules were correct, and if these arguments were extended to other areas of epistemology and policy, then they would invalidate most of our knowledge claims.127 Rather than admit that most of our knowledge claims are invalid, it seems more reasonable to say that the naive positivists and the cultural relativists have conceived of "objective" risk evaluations in partially unrealistic ways. Moreover, both groups of reductionists have confused general principles of risk-evaluation methods with specific procedures. I have argued that a more realistic way to conceive of rational risk assessment is in terms of "scientific proceduralism."

In subsequent chapters of this volume, I shall show how the failure to define rational risk evaluation in terms of "scientific proceduralism" leads both to flawed policy and to disenfranchising members of the public likely to be victims of societal risk. Later, in Chapter Eleven, I shall show how my proposals for reforming risk evaluation and management meet the criteria for scientific rationality and objectivity outlined in this chapter; I shall also answer several prominent objections to scientific proceduralism. These objections fail, in large part, because they miss the basic insight of Israel Scheffler: scientific "objectivity requires simply the possibility of intelligible debate over the merits of rival paradigms."128

Why We Need a Procedural Account of Rationality

In 1988, the U.S. Department of Energy (DOE) applied for state permits to begin testing at Yucca Mountain, Nevada, the only candidate for the nation's first high-level nuclear waste dump. In winter 1990, Nevada's governor said that he would never issue the permits, even though the Justice Department has brought suit over the state's failure to comply. Nevada has argued that Yucca Mountain is full of fractured rock that could allow contaminated water to migrate. DOE says the site is safe.1

If previous experience at other U.S. hazardous waste sites is representative, the people of Nevada have cause for concern. In 1962, for example, geologists calculated the risks associated with a proposed site for shallow land burial of transuranics and low-level radioactive wastes. Scientists from industry and academia praised the (Maxey Flats) Kentucky location and calculated that, if plutonium were buried there, it would take 24,000 years to migrate one-half inch. They said that "the possibility of subsurface migration offsite is nonexistent."2 They were confident about their predictions because experts typically have argued that, "in the main, epidemiology studies have not shown that these [hazardous waste] sites produce adverse health effects in humans who live near them"3 Yet, only ten years after the facility was opened, plutonium and other radionuclides were discovered two miles off site.4 The geological predictions were wrong by six orders of magnitude.

Although the Kentucky facility has been closed, it remains a problem. The site contains more plutonium (in curies) than any other commercial repository in the world,5 and scientists have been unable to halt the radwaste migration. The half-life of plutonium 239 is 24,390 years, and less than a millionth of a gram is sufficient to cause lung cancer.6

Not surprisingly, journalists have called the Kentucky site the world's "worst nuclear dump," a "fountain of death."7

One of the many questions raised by this case is whether the faulty hydrogeological predictions at the Maxey Flats radwaste site were highly atypical. If they were, then perhaps Nevadans have little to fear. If they were not, then Yucca Mountain may also be a cause for concern. Are science, in general, and risk assessment, in particular, more reliable than these erroneous migration predictions suggest?

Probably not. Experts reveal that uncertainties of six orders of magnitude are "not unusual" in risk assessment.8 Likewise, a famous study on the siting of LNG (liquefied natural gas) terminals in four countries showed that widely varying risk estimates for the same event are pervasive; in Oxnard, California, for instance, reputable hazard assessments (of citizens' average annual chance of dying in a liquefied natural gas accident) differed by three orders of magnitude.9 Likewise, current U.S. government predictions for the likelihood of a Three Mile Island type of accident differ by two orders of magnitude.10 Other cases are much the same; they suggest that scientific predictions and risk estimates are laden with methodological value judgments and enormous uncertainties.11

To claim that science is laden with value judgments, however, is to challenge the whole edifice of epistemology in the West. Science is the paradigm of all knowledge.12 If it contains value judgments, then we must radically rework the ideal of scientific objectivity. As Rudner put it, "the slightly juvenile conception of the coldblooded, emotionless, impersonal, passive scientist mirroring the world perfectly in the highly polished lenses of his steel rimmed glasses—this stereotype—is no longer, if it ever was, adequate."13

As noted in Chapter Three, we need to build a new account of rationality and objectivity. This account (scientific proceduralism) falls midway between the positions of the naive positivists and the cultural relativists. Neither of them, as the last chapter showed, gives an adequate account of scientific rationality and objectivity. The naive positivists underemphasize the value judgments in science, whereas the cultural relativists overemphasize them. Both the naive positivists and the cultural relativists erroneously claim that citizen aversion to societal risks is biased. Calling people "biased" or "irrational" in their risk evaluations, however, presupposes that they are making illegitimate value judgments about acceptable risk. To determine the correctness of the "irrational" label, one must know something about which value judgments are legitimate and which ones (if any) are compatible with scientific rationality and objectivity. By further clarifying the nature of

value judgments used in science, hazard assessment, and risk evaluation, I shall be able to present a new account of which risks are rational to accept. I begin by outlining the "standard view" of risk assessment. Next I argue that this "standard view" is wrong; it ignores the characteristic, epistemic values central to doing even pure science. Finally, I outline some typical value judgments essential to methods used at each of the three stages of risk analysis, closing with a familiar example of a highly value-laden hazard assessment, Inhaber's energy study.

The "Standard Account" of Risk Assessment

Many practitioners of risk or hazard assessment or analysis, as was mentioned in Chapter One, divide it into three stages: risk identification, risk estimation, and risk evaluation. 14 Most of the discussion in this volume focuses on the third stage, risk evaluation. At the first stage, one identifies a particular threat to human health and safety—for instance, vinyl chloride. At the second stage, scientists such as epidemiologists or environmental engineers estimate the risk of death or injury associated with particular levels of exposure to that hazard. At the third stage, scientists such as economists and health physicists, as well as psychologists and policy makers, evaluate what level of exposure to the risk, if any, is acceptable to society. After these three tasks of hazard analysis are completed, regulators and lawmakers begin the process of risk management, ensuring that the actual risk exposure conforms to the standards of acceptability already determined by assessors.

Proponents of the "standard account" of risk assessment, naive positivists such as Starr, Whipple, Maxey, Cohen, Lee, and Okrent, maintain that the expert methods employed in the first two stages of assessment (risk identification and risk estimation) are completely objective, neutral, and value free. They say that the somewhat subjective methods employed in the third stage (risk evaluation) can be made more objective and neutral merely by improving the analytic techniques used by experts to evaluate risks.15 Risk assessment, in their view, is a largely scientific discipline, to be perfected along hypothetical-deductive lines.16 They are on the quest for algorithms that will give them power to predict and assess risks in a wholly neutral way, much as Einstein and David Bohm hoped for deterministic, predictive power for quantum mechanics.17

Critics of the standard account argue, however, that risk analysis— especially its third stage, risk evaluation—ought to be accomplished both through hypothetical-deductive testing and through democratic give-and-take involving citizens. It cannot depend solely on an expert-

controlled analytic or scientific methodology.18 If these critics are correct, then risk evaluation is not merely a scientific investigation but also a political procedure to be negotiated among experts and the public. Many assessors ignore the procedural aspects of hazard evaluation, calling merely for "scientifically verifiable models with testable hypotheses" and urging fellow assessors to improve risk assessment so that it earns "more respect in the scientific community."19

Risk Evaluation, Like Science, Is Evaluative

Although they reveal good intentions and a healthy empiricism, such pleas are in part misguided. Confirmation and acceptance by the technical community are clearly goals for wholly scientific enterprises. These goals, however, are never fully realized, especially in risk evaluation. Although the first two stages of hazard analysis are largely scientific, they nevertheless rely on a number of methodological or epistemic value judgments.20 Moreover, the third stage of hazard assessment requires comparing a given risk to a great many other ones, in order to determine whether it is tolerable to society. This means that the judgment of acceptability is a function not only of predictive and explanatory power but also of the kinds of comparisons used. Hazards may be compared on the basis of their probability, their consequences, their benefits, their equity of risk distribution, their voluntariness, the level of consent given to them, and so on. Obviously, these comparative notions are not subject to empirical confirmation, even though risk analysis is touted as a "developing science."21

This comparative side of risk evaluation, however, is one of its strengths. It provides opportunities for promoting efficiency by maximizing the benefits of government expenditures for health and safety. It also promotes equity and consistency in allocation of funds among safety programs. As one thinker put it, risk assessment enables us to ask important questions, such as "why OSHA intends, in a set of proposed regulations on coke-oven emissions, to protect the lives of steelworkers at $5 million each, while a national Pap-smear screening program that would save women's lives at less than $100,000 each has gone unfunded."22

Perhaps the greatest liability of comparative, quantified risk assessment, especially risk evaluation, however, is that it attempts to reduce qualitatively diverse risks to mere mathematical probabilities and numbers of fatalities. Critics of hazard analysis also charge that it threatens to remove health and safety from democratic social control and to place it solely in the hands of experts. Apart from whether opponents of

risk analysis are correct in their allegations about its political liabilities, the third, or evaluative, stage of hazard assessment is the most controversial precisely because it involves quantitative comparisons of diverse risks. Because of such comparisons, risk evaluation is both objective or scientific (in the three senses discussed in the previous chapter) and procedural in focusing on the normative goals of efficiency and equity.

Several reasons why risk analysis, and especially risk evaluation, are value laden have been pointed out in the previous chapter.23 Assessors must make value judgments about which data to collect; how to simplify myriad facts into a workable model; how to extrapolate because of unknowns; how to choose statistical tests to be used; how to select sample size; determine criteria for NOEL (no-observed-effect level);24 decide where the burden of proof goes, which power function to use, what size of test to run, and which exposure-response model to employ.25 these methodological value judgments in hazard identification and estimation suggest that, even though it is objective in the three senses already discussed, risk assessment is also value laden.

To claim that science and risk assessment are value laden, however, is to challenge naive-positivist orthodoxy, as the previous chapter argued.26 Although most philosophers of science realize that science may be described in terms of some sort of Kuhnian revolution, few seem to understand that this recognition of scientific revolution was caused by a new awareness of the role played by value judgments in scientific work. The reason why many philosophers continue to follow the naive-positivist, Carnapian view of the nature of science (as free from values) is likely that, although they correctly believe that emotive values have no place in science, they erroneously associate values only with emotion and feeling.27

Admittedly, science should not include emotive values. As the previous chapter argued, however, values can be cognitive and not just emotive. Consequently, one can argue that there are methodological value judgments in science,28 and that they substantively affect scientific conclusions.29 The case of physicist Wolfgang Pauli’s postulation of the neutrino illustrates this point. Pauli emphasized the value of internal consistency with conservation principles and therefore postulated the existence of the neutrino before it was observed. Other physicists of his time emphasized a different value— external consistency with principles demanding empirical proof—and therefore rejected the neutrino.30

In response to these claims about values, some positivistically oriented persons would likely argue that, although science does rest on nonverifiable "postulates," it does not involve values per se.31 The prob-

lem with this sort of response, however, is that the "postulates" (e.g., Keynes's postulate that economics can obtain legitimate conclusions by ignoring all human desires except the desire for wealth) involved in science are matters of controversy among experts. Therefore, someone could come up with different "postulates" or "values" and defend them on different grounds. This means that science is not more objective by virtue of employing "postulates" rather than "values." Both labels are threats to naive accounts of objectivity.32

Value Judgments in the Three Stages of Risk Assessment

One of the most important threats to naive accounts of objectivity in risk analysis is the postulate or value judgment that one ought to define risk as a compound measure of the probability and magnitude of adverse effect. This measure is often expressed as the "average annual probability of fatality." It presupposes that one is justified in interpreting "risk" in terms of fatalities alone and ignoring injuries; financial losses; carcinogenic, mutagenic, and teratogenic effects; and political and philosophical harms, such as threats to due-process rights. Various emissions standards for industrial pollutants, for example, threaten due-process rights because it is virtually impossible to tell which pollutant is responsible for an injury. It is therefore almost impossible to collect damages for particular harms caused by industrial air pollutants, even though air pollution in the eastern United States poses nearly as great a danger of death as smoking a pack of cigarettes each day. The standard account of risk, however, ignores such dangers because of its narrow, highly evaluative definition of risk.33 Indeed, each of the three stages of hazard assessment (risk identification, risk estimation, and risk evaluation) involves numerous evaluative judgments.34

It is difficult, first, to identify risks because each of the five commonly used methods of carcinogen identification, for example, has serious defects whose presence requires the assessor to make epistemic and pragmatic value judgments. For instance, the use of case clusters (looking for adverse effects to appear in a particular place) is helpful only when specific hazards cause unique diseases and only when the population at risk is known in some detail. Since often it is not known whether these conditions are satisfied, especially in cases involving new risks, assessors must evaluate the situation as best they can.

Comparison of compounds on the basis of structural toxicology, a second method of risk identification, likewise is problematic. It reveals only that a potential toxin, for example, has the same structure as a

carcinogen. Yet, in using this method, assessors typically assume that this similarity of structure is sufficient to determine that a substance is a positive carcinogen. Use of mutagenicity assays also relies on a normative judgment, that most chemical carcinogens are mutagens. But mutagenicity assays are rarely sufficient to support the conclusion that a particular mutagen is also carcinogenic. Long-term animal bioassays likewise depend on an epistemic value judgment, that particular results from animal experiments are applicable to humans.

Another class of risk-identification methods, biostatistical epidemiological studies, employs more sophisticated case clusters that aim to show an association between an agent and a disease. The obvious deficiencies with this method are that it is often difficult to accumulate the relevant evidence, particularly if exposure is at low dosage or if the effects are delayed, as in cancers with latency periods of up to forty years. In the absence of complete data and long years of testing, assessors are forced to interpret and to extrapolate from the information that they do have, and hence to make both pragmatic and epistemic judgments. Moreover, most substances are not even tested with epidemiological methods; apart from other sources of hazards, roughly 60,000 chemicals are used annually in various manufacturing processes, and at least 1,000 new ones are added each year. A new toxic chemical is introduced into U.S. industry every twenty minutes.35 Deciding which chemicals to test, when not all can be tested, is perhaps one of the greatest pragmatic value judgments in risk identification. Yet without biostatistical epidemiological testing, it is unlikely that hazardous chemicals will be removed from the market. Only a few pesticides commonly used in the United States, for instance, have received biological epidemiological testing, even though a recent U.S. National Academy of Sciences study reports that 60 percent of all herbicides and 90 percent of all fungicides used in the United States can cause cancer in animals.36

At the second stage of risk assessment, risk estimation, one is concerned primarily with determining a dose-response relationship for a particular hazard, the population at risk, and the dose it receives from the hazard. Determining a dose-response curve is important in order to set exposure limits and to trace causal pathways back from a harm (for example, cancer incidence in a population) to its origin in a hazard such as radiation. Dose-response methods are conceptually problematic because risk assessors do not have open to them the obvious standard techniques of justification. They cannot expose a series of populations to a dose of some carcinogen and then count the cancer fatalities in each population. Instead, assessors must make many value judgments

about the validity of extrapolation from high to low doses, from animals to humans, and from one group of humans to another.

The problems associated with such extrapolation are well known. For example, health physicists often must try to estimate a dose-re-sponse curve for low-level radiation exposure when they are given data points only for higher-level exposures. Environmentalists, industry representatives, and members of health physics associations have extrapolated, respectively, to a dose-response curve for radiation that is typically logarithmic or supralinear; quadratic, linear-quadratic, or linear with a large threshold; and linear with no threshold. On the basis of their differing value judgments about the low-dose end of the curve, they have concluded, respectively, that low-level radiation is very dangerous, not very dangerous, and moderately dangerous. To give some idea of the differences in the estimates produced when the several models for radiation dose-response curves are used, consider the excess incidence of cancers predicted, other than leukemia and multiple mye-loma. For a 10-rad exposure, the quadratic model predicts 560 excess cancers per million persons exposed, whereas the linear-quadratic predicts 4,732 excess cases, and the linear model predicts 10,630 excess cancers. Thus, there is a twentyfold difference between the quadratic and the linear models! More surprisingly, there is an eighty-six-fold difference between the predictions of the quadratic and the supralinear models. However, the National Academy of Sciences' (NAS) Committee on Biological Effects of Ionizing Radiation claims that there is no reason to prefer one model over the other at low doses. The committee therefore seems to suggest that one value judgment about which curve to use is not always more objective than another.37

Estimating the population at risk and the dose received is just as problematic as determining the dose-response curve, because estimators must assume that their theories provide a reasonable measure of exposure; that the exposure is related in some clear way to the administered dose; that the absorbed dose is a specific function of the administered dose; that phenotypic variation in populations does not inordinately affect the dose received; that the critical organ dose is related precisely to the absorbed dose; and that the effect is a particular function of the critical organ dose. Because actual measurements of particular doses of a chemical, for example, cannot be made in all situations, a mathematical model of assumed exposure must be used, and assessors must make an epistemic judgment as to its suitability. Sometimes these judgments differ radically; toxicologists, for example, typically view the notion of a safe threshold dose as controversial, whereas engineers, who equate dispersal with dilution, tend to assume

some threshold below which a substance is no longer harmful. Neither group, however, typically addresses synergistic effects,38 even though seemingly unimportant pathways of exposure can assume great significance—owing, for example, to biomagnification or food-chain effects.

Moreover, it is rare that a substance is uniformly distributed across pathways and time. Numerous studies have shown that the target dose, pathway, and persistence of a toxic chemical, for example, are difficult to predict, as was illustrated by the 1976 accident in Seveso, Italy; a toxic cloud of TCDD, including dioxin, was released over 5,000 people.39 Similar difficulties plagued epidemiologists tracing the distribution of radiation after the Chernobyl nuclear accident on 25 April 1986. Maps showing the areas of greatest radionuclide deposition look like Rorschach drawings; the distribution did not follow a predictable pattern based on prevailing wind direction and the rotation of the earth. Instead, numerous factors, such as rainfall patterns and the height of the radioactive plume, affected the highly irregular spread of radiation following the accident.40 Moreover, even if pathways of various toxins could be determined exactly, phenotypic variations among populations would mean that one person could be two hundred times more sensitive to a chemical than another, even when both received equal doses41 Given such situations, the assessor is forced to make a number of interpretative, value-laden risk-estimation judgments.

Highly evaluative judgments also are made at the third stage of hazard assessment, risk evaluation. At this stage, analysts typically determine whether a risk is acceptable to society. They employ at least four different methods: risk-cost-benefit analysis (RCBA), revealed preferences, expressed preferences, and natural standards. RCBA is widely used and consists simply of converting the risks, costs, and benefits associated with a particular project to monetary terms and then aggregating each of them, in order to determine whether the risks and costs outweigh the benefits.42 In using RCBA, to make a number of highly evaluative judgments—for instance, that monetary-term parameters can adequately represent the real cost or benefit of a thing or that the magnitude of risks, costs, and benefits is more important than their distribution .43

The second method of risk evaluation, revealed preferences, consists of making inductive inferences about acceptable risk on the basis of the levels of risk that existed in the past.44 The most worrisome value judgment in this method is that past societal risk levels reveal desirable risk levels for the present. This assumption requires one to judge that, where risk is concerned, the present ought to be like the past, and that

hazards accepted in the past ought Both of these highly evaluative judgments may be questionable, either because past levels of risk may be indefensible (for example, they may have been imposed on unwilling victims), or because our ethical obligations regarding present risk may be greater than those of our ancestors regarding past risk.

The third method of risk evaluation, expressed, preferences, consists of using psychometric surveys to determine the acceptability of particular hazards.45 It is built on the questionable assumption that the preferences people express via instruments such as surveys provide reliable indicators of acceptable risks. Obviously, however, preferences are not always authentic indicators of welfare: some persons have irrational fears, and others are too ignorant to realize a serious danger. This method also requires assessors to make a number of evaluative judgments whenever they encounter inconsistencies in survey preferences or whenever the responses fail to correspond with actual behavior regarding environmental hazards.

The fourth method of risk evaluation, natural standards, uses geological and biological criteria to determine hazard levels current during the evolution of the species.46 Like the method of revealed preferences, this method is based on the assumption that, if a particular level of risk was present in the past, then that same level is acceptable at present. Such an assumption seems unwarranted, however, especially if the risk can be avoided, or especially if no overarching benefits arise from allowing the risk to remain high. Moreover, since the method is based on a “natural standard" for each different hazard, it totally fails to take account of synergistic or cumulative effects. These effects could be many orders of magnitude greater than the actual “standard level" of exposure.

All the assumptions, extrapolations, and inferences built into each of the methods in the three stages of hazard assessment arise in large part because nearly every situation of risk identification, estimation, or evaluation is factually incomplete and imprecise, and therefore empircally underdetermined. Because it is empirically underdetermined, risk assessors are forced to make some methodological value judgments, so as to interpret the data available to them. At root, these assumptions and inferences often reduce to the problem of getting a grip on causality. This is difficult, since causes are not seen; they are inferred on the basis of their effects. Cancers, for example, don’t wear tags saying that they were caused by their subject’s smoking, or by use of oral contraceptives, or by diet drinks, or by the subject’s breathing the emissions of the chemical plant next door. Moreover, we know statistically

how much of one agent causes cancer in general, but we can never determine whether a substance caused a particular cancer. We can only infer what caused a certain cancer.

Classic examples of the difficulty of establishing causality of harm, given a certain probability of risk, occur throughout industry and government. Industry analyses of Monsanto and Dow workers, all exposed to chemical accidents and all victims of chloracne for a mean time of twenty-six years afterward, claimed that those exposed had "no heightened mortality." Likewise, a study of GI victims of Agent Orange, thirteen years after exposure, concluded that there were "no unusual findings attributable to the herbicide." The government drew similar conclusions when U.S. servicemen were exposed to above-ground nuclear weapons tests in the western United States and in the South Pacific. Many of them were within five miles of ground zero for as many as twenty-three separate explosions, or were marched to within three hundred yards of ground zero immediately after detonation. Yet, of the half million soldiers so exposed during the 1950s and 1960s, only ten have won benefits when they or their survivors claimed that their cancers were caused by the fallout. Like risk victims everywhere, they were able to show a statistically significant increase in certain types of injuries or deaths, a correlation between increased dose and increased response, but no cause of particular cancers. Without proof of causality, they could claim no benefits. In the absence of proof and complete data, they were forced, like risk assessors everywhere, to rely on value judgments, on inference, extrapolation, and simplification.47

Inhaber's Risk Assessment

One famous example of a highly evaluative, highly controversial hazard analysis is that of Herbert Inhaber. Commissioned by the Canadian Atomic Energy Control Board and summarized in Science, his study estimated the dangers from alternative energy technologies. It concluded that the risk from conventional energy systems, such as nuclear or coal power, is less than that from nonconventional systems, such as solar or wind energy. It also concluded that noncatastrophic risks, such as those from solar power, are greater than catastrophic risks, such as those from nuclear power.48

How did Inhaber arrive at his surprising conclusions? He made some highly questionable evaluative assumptions. In estimating the risk posed by particular energy technologies, he assumed, for example, that all electricity was of utility-grid quality (i.e., able to be fed into a power grid).49 As a result, he ignored the low-risk benefits of solar space heat-

ing and hot-water heating, neither of which can be fed into a power grid. Indeed, he ignored the wide variety of low-temperature forms of solar energy, which could supply 40 percent of all U.S. energy needs at competitive prices and at little risk.50

Another highly evaluative assumption central to Inhaber's risk estimates was that all nonconventional energy technologies have coal backups.51 As a result, for example, in the case of solar thermal electric, 89 percent of the risk that he attributed to it comes not from the solar (involving mainly risk from construction of components) but from the coal backup, which he classified as a solar risk! Moreover, Inhaber assumed that nuclear fission requires no backup, even though these plants have a downtime of 33 percent per year for checkups, refueling, and repairs.52

In the area of risk evaluation, Inhaber's assumptions are just as questionable. When he aggregated and compared all lost workdays, for all energy technologies, he ignored the fact that lost workdays are more or less severe, depending on the nature of the accident causing them and whether or not they are sequential. In Inhaber's scheme, a lost workday due to cancer or the acute effects of radiation sickness is no different from a lost workday due to a sprained ankle.53 Yet obviously the cancer could result in premature death, and the radiation exposure could result in mutagenic effects on one's children. Neither is comparable to a sprained ankle.

Inhaber made a similar questionable assumption in evaluating the severity of risks. Unlike other hazard assessors, he totally ignored the distinction between catastrophic and noncatastrophic risks. He assumed that there is no difference between 1,000 construction workers who fall off a roof and die, in separate accidents, and 1,000 workers who die because of a catastrophic accident in a nuclear fuel fabrication plant.54 Numerous risk assessors, however, typically make a distinction between catastrophic and noncatastrophic accidents. They suggest, for example, that because of increased societal disruption, n lives lost in a single catastrophic accident should be assessed as a loss of n 2 lives.55 Whether or not this n 2 interpretation is a reasonable one, the point is that Inhaber made numerous evaluative assumptions, such as that the distinction between catastrophic and noncatastrophic accidents could be ignored. In so doing, he passed off his results as purely objective hazard assessment. Instead, were one to trace Inhaber's methods, step by step, it would be clear that virtually every assumption he made in estimating and evaluating alternative risks had the effect of increasing his alleged nonconventional risks and decreasing his alleged conventional risks.56

Even if persons such as Inhaber avoided their questionable means of aggregating and estimating risks, values would unavoidably enter the hazard-evaluation process. This is because value judgments are often the only ways to resolve some ethical dilemmas facing risk assessors, especially at the third or risk-evaluation stage. The next chapter will survey some of the dilemmas besetting risk evaluation, showing that they reflect the need for a new account of rational risk acceptance.

Why We Need a New Framework for Rational Risk Evaluation

By late 1987, 25 percent of all Ugandans tested positive for AIDS (acquired immune deficiency syndrome), "slim disease," as it is called there. Since those affected are both heterosexual and homosexual, officials fear that this new retrovirus will wipe out an entire generation of central Africans, from Zaire to Kenya. Indeed, even if the AIDS epidemic does not enter the Asian population, the World Health Organization (WHO) expects one hundred million people, worldwide, to be infected by 1991. If AIDS does travel to Asian countries, then the WHO says that one hundred million "is an excessively conservative figure."1

Even though cases of AIDS in the United States increased from approximately 500 in 1982 to about 40,000 in 1987, managing the disease in the West is not as traumatic as in developing nations. In Uganda and other African countries, there are (for most people) no doctors, no medicine, no suitable hospital facilities, and no one to care for the children of victims. Apart from treatment, the cost of the AIDS test, alone, is more than total annual per capita health expenditures, about a dollar, in Uganda. Consequently, doctors or clinicians diagnosing AIDS in Africa face a dilemma. Ought they to tell the patients about their disease, in the name of honesty, even though nothing can be done (at least in central Africa) to alleviate their condition? Or, to counter "the secondary epidemic" of fear, ought they to lie to AIDS victims about their ailment, on the grounds that the truth would only worsen their mental, and perhaps their physical, state?2

Similar, although less poignant, dilemmas face those who manage other medical, environmental, and technological hazards. U.S. regulators, for example, have discovered that they face a typical risk di-

lemma in the meat-cutting industry. The Occupational Safety and Health Administration (OSHA) reduced meat cutters' risks of accident by requiring construction of guard rails around the cutting machinery. Once they installed the protective equipment, however, officials realized that the unsanitary rails increased consumers' risks from infected meat. 3

Risk management generates pressing dilemmas, in part, because of two characteristics of risk situations: competing opportunity costs and incompatible risk-displacement options. Regarding opportunity costs, if one spends research dollars to help reduce the risk of heart disease, for example, then one has lost the opportunity to use those same funds to reduce cancer hazards, and vice versa. Incompatible risk-displacement options arise because risk can never be removed; it can only be transferred or displaced. One might reduce meat cutters' risk, but only by increasing meat consumers' risk, and vice versa. Or, by lying about the nature of AIDS patients' ailments, a medical doctor might reduce their traumas, but only by increasing the risk to the victims' future sexual partners.

Fully rational people, whether medical personnel in central Africa or OSHA regulators in the United States, often are caught in risk situations for which there are no desirable solutions. Although the situations do not pose logical difficulties, they are dilemmas in the sense that all the options for managing risks lead to unwanted consequences. Often these dilemmas arise at the third, or risk-evaluation, stage of analysis. Discussing five dilemmas of risk evaluation, this chapter shows that several philosophical problems, revealed by these dilemmas, require us to reform risk evaluation and to redefine "rational" risk acceptance. Later in the volume, in Chapters Eleven and Twelve, I shall outline these reforms and redefinitions.

The Fact-Value Dilemma

The first problem, the fact-value dilemma, arises because of the methods used at the third (or risk-evaluation) stage of risk analysis. Recall, from the last chapter, that four methods dominate this stage of assessment: risk-cost-benefit analysis (RCBA), expressed preferences, revealed preferences, and natural standards. None of these four methods of risk evaluation employs any explicitly ethical criteria, even though each of them (as the previous chapter argued) is laden with evaluative assumptions. Therein lies the dilemma: Ought or ought not assessors to employ risk-evaluation criteria that are overtly normative? For example, should they avoid explicitly ethical analyses and instead use (largely factual) risk-cost-benefit analysis? Or should they insert explicit

norms into risk-cost-benefit analysis? Such norms might require, for example, that, regardless of the benefit-cost ratio, all persons ought to be guaranteed equal protection from all risks above a certain threshold.

At the heart of the fact-value dilemma is the assumption that methods of risk evaluation must be either primarily factual and scientific (and accomplished by experts) or predominantly ethical and political (and accomplished by democratic procedures). If they are not mainly scientific (and accomplished by experts), then the evaluation is open to the charge of being subjective and arbitrary. If they are not primarily ethical and political (and accomplished by democratic procedures), then the analysis is open to the charge of ignoring citizens' rights to self-determination and representation.

Admittedly, risk evaluators strive to make their analyses as factual, objective, and scientific as possible, on the grounds that such evaluations will be rational and persuasive tools of public policy.4 Yet, if they are truly factual and scientific, risk evaluations ought to be done only by experts, not by laypersons. But if they are accomplished only by experts, their conclusions probably will not be wholly acceptable to the public. This is because—especially at the third stage of hazard assessment—citizens wish to make their own judgments about how safe is safe enough. There have been too many Chernobyls, Love Canals, Three Mile Islands, Bhopals, and Utah weapons tests for citizens to allow their judgments of acceptable risk to be preempted completely by experts. As one environmental activist put it, deciding to accept a risk "is an entirely human judgment that has nothing to do with whether you're a farmer or an engineer or a mathematician."5 The fact-value dilemma is that risk evaluation cannot be both wholly factual/scientific and wholly sanctioned via democratic procedures.6

The Standardization Dilemma

A second difficulty, the standardization dilemma, raises similar problems. This dilemma arises because, in the name of consistency and avoidance of arbitrariness, many analysts argue that assessment procedures and risk-evaluation theories ought to be standardized.7 For example, one ought to use similar risk-estimation models to evaluate substances and exposure levels that are relevantly similar. Yet, when procedures, models, and theories of risk evaluation are standardized, the assessor is unable to take into account persons' claims about special or unique needs, circumstances, and merits.

Proponents of standardization have often made an equity argument. They argue, for example, that hazard assessors employing risk-cost-

benefit analysis (at the third stage of risk analysis) ought to make the marginal cost of saving lives, across opportunities, the same.8 In other words, they maintain that risk evaluation ought to embody the principle that the same amount of government funds be spent, per life saved, in different risk situations. One could argue that politically powerful persons (e.g., middle-aged, overweight, white, male congressmen who typically die of heart disease) should not be able to have great amounts of risk-abatement and research monies spent to protect them from heart disease, while politically powerless persons (e.g., blacks dying of sickle-cell anemia) have fewer funds spent to protect them from anemia.

The obvious problem with this standardization proposal is that, if the same amount of money were spent to save lives, regardless of different risk situations, then assessors could violate (at least intuitive) standards of rationality. For example, suppose proponents of standardization argued that the same number of taxpayer dollars, per life saved, ought to be spent to protect citizen victims of dam failures as to protect the public from commercial nuclear accidents. Opponents of standardization might object that, since death from radiation is more unknown and more feared than death because of dam failure, more money ought to be spent to protect potential victims of nuclear accidents.

More generally, opponents of standardization could question the equity argument on a number of epistemological and ethical grounds. They might claim that standardization forces one to assume (1) that sameness of expenditures guarantees sameness of protection; (2) that sameness of protection guarantees equality of protection; or (3) that all discrimination (all deviation from standardization or consistency of treatment) is ethically indefensible and that there are never any morally relevant grounds for discrimination. Finally, they might claim (4) that standardization ignores the fact that different persons have different claims of merit, need, incentive, and compensation, any of which might be grounds for "special," rather than "standard," treatment.9 For example, public figures might claim that they need better-than-standard risk protection; or they might argue that they bring about societal benefits and therefore ought to have above-average help. For these and other reasons, Ronald Dworkin has observed that there are no ethical grounds for claiming that everyone ought to receive the same treatment. Instead, he argues, there is an ethical basis only for asserting that everyone ought to have the same concern or respect in the political decision about how to allocate goods (and risk-abatement monies). Not only would it be impossible to accord the same treatment to everyone, even if it were desirable, but there are also strong reasons, some given in

points 1-4 above, for doubting that sameness of treatment is desirable.10

Because of the constraints posed by the standardization dilemma, evaluators must make a value judgment. They must choose between efficiency and equity, between standardization and ethical analysis.11

The Contributors Dilemma

A third difficulty faced by risk evaluators is how to assess numerous small hazards. I call this problem the "contributors dilemma." It consists of the fact that citizens are subject to many small hazards, each of which is allegedly acceptable; yet together such exposures are clearly unacceptable. For example, each of the many carcinogens to which we are exposed—such as asbestos, vinyl chloride, and radiation—is alleged to be acceptable because it is below the threshold at which some statistically significant increase in harm occurs. Yet, statistically speaking, 25 to 33 percent of us are going to die from cancers, 90 percent of which are environmentally induced and hence preventable.12 Many of the cancers are caused by the aggregation of numerous exposures to carcinogens, none of which is alone alleged to be harmful.

The contributors dilemma forces assessors both to assume (in the case of aggregate risks) and not to assume (in the case of individual risks) that the whole risk faced by an individual is greater than the sum of the parts of that risk. Risk evaluators who condone subthreshold hazards, but who condemn the deaths caused by the aggregate of these subthreshold harms, are something like the bandits who eat the tribesmen's lunches in the famous story of Oxford philosopher Jonathan Glover:

The obvious question raised by this example is how risk assessors can say both that subthreshold exposures are harmless, as the data indicate, and yet that the additivity, or contribution, of these doses causes great harm. If they claim that subthreshold risks are unacceptable, then they face the undesirable consequence that government must regulate such risks, an extraordinarily difficult and expensive task, since attaining zero risk is impossible. If they claim that subthreshold risks are acceptable, then they must admit that, although it is immoral to murder fellow citizens, it is moral to allow them to be killed by carcinogens, little by little, much as the bandits stole the tribesmen's lunches, bean by bean.

The De Minimis Dilemma

The de minimis dilemma, named after a Science editorial about de minimis, or negligible, risk, poses many of the same problems as the contributors dilemma.14 It is based on the fact that society must declare some threshold, below which a hazard is judged to be negligible. Often this de minimis level for a given risk is set at what would cause less than a 10^-6 increase in one's average annual probability of fatality.15

The reasoning behind setting such a level is that a zero-risk society is impossible, and some standard needs to be set, especially in order to determine pollution-control expenditures. Choosing the 10^-6 standard also appears reasonable, both because society must attempt to reduce larger risks first and because 10^-6 is the natural-hazards death rate.16 The dilemma arises because no de minimis standard is able to provide equal protection from harm to all citizens. On the one hand, if one rejects the de minimis standard, then pollution-control requirements would be difficult to determine. On the other hand, if one accepts the de minimis standard, then citizen protection would be based on some average annual probability of fatality, not on equal protection for all.

Although this 10^-6 threshold seems acceptable, on the average, it is not necessarily acceptable to each individual. Most civil rights, for example, are accorded not on the basis of the average needs of persons but on the basis of individual characteristics. We do not accord constitutionally guaranteed civil rights to public education, for example, on the basis of average characteristics of students. If we did, then retarded children or gifted children would have rights only to education for children at the average level. Instead, we say that guaranteeing "equal" rights to education means according "comparable education," given one's aptitudes and needs.

If civil rights to education are accorded on the basis of individual,

not average, characteristics, why are civil rights to equal protection from risks not accorded on the basis of individual, rather than average, characteristics? Why is a 10^-6 average threshold accepted for everyone, without compensation, when adopting it poses risks higher than 10^-6 for the elderly, for children, for persons with previous exposures to carcinogens, for those with allergies, for persons who must lead sedentary lives, and for the poor? Blacks, for example, face higher risks from air pollution than do whites, because "average" exposure standards cover up the fact that blacks live in neighborhoods with greater air pollution, while whites live in those with less, even though the average, citywide, exposure is the same.17 The de minimis dilemma forces us to choose between average and equal protection, between efficiency and ethics.

The Consent Dilemma

Another evaluative dilemma arises from the recognition that imposition of certain risks is legitimate only after consent is obtained from the affected parties. The dilemma occurs because those who are genuinely able to give legitimate consent are precisely those who likely will never do so, whereas those who allegedly have given consent are those who are probably unable to do so. Perhaps the best example of the consent dilemma occurs in the workplace. Here there is an alleged compensating wage differential, noted by both economists and risk assessors. According to the theory behind the differential, the riskier the occupation, the higher the wage required to compensate the worker for bearing the risk, all things being equal.18 Moreover, imposition of these higher workplace hazards apparently is legitimate only after the worker consents, with knowledge of the risks involved, to perform the work for the agreed-upon wage.

The consent dilemma appears as soon as one considers who is most likely to give legitimate informed consent to a workplace hazard. It is a person who is well educated and possesses a reasonable understanding of the risk, especially its long-term and probabilistic effects. It is a person who is not forced, under dire financial constraints, to take a job that he knows is likely to harm him. Sociological data reveal that, as education and income rise, persons are less willing to take risky jobs, and that those who do so are primarily those who are poorly educated or financially strapped.19

Consider, for example, those who bear the risks of coal mining in Appalachia. (Appalachia includes much of the states of Kentucky, West Virginia, Virginia, Tennessee, North Carolina, and South Carolina.) It

is common knowledge that poorer workers are typically employed in the most risky jobs,20 and that residents of Appalachia generally have no alternative to working in the mines unless they want to move out of the region. This is because the Appalachian economy is not diversified; because there is no job training for a variety of positions; and because absentee corporations, which control 80 percent of all Appalachian land and mineral rights, also control the only jobs. The situation is one of monopsony, where owners of most of the land also control most of the employment.21 Even if all Appalachian coal miners were generously compensated, and even if they all had perfect information about the dangers of their jobs, very likely the situation of monopsony would prevent their making a wholly voluntary choice to work in the mines. They would essentially have the choice between a risky job and no job at all, unless they wanted to leave their homes. Hence, it is problematic to claim that, in exchange for the higher wages, they have genuinely given free consent to the high-risk mining jobs.

Analogous questions about free, informed consent could be raised in other situations involving environmental hazards. Consider, for example, the type of community most likely to accept a particularly dirty, dangerous factory or a controversial toxic waste dump. It would, very likely, be a town where people needed the jobs, or the tax benefits of the facility, or where levels of education about the relevant dangers of the facility were less known than in other places.22 In other words, the communities least able to give free, informed consent to a risk are typically those that allegedly, consent.

Risk evaluators who face the consent dilemma may have something to learn from medical experimenters who face analogous constraints. They taught us, long ago, that prisoners who are promised early release in exchange for being subjects in risky medical experimentation very likely are unable to give free, informed consent to the risk.23 This is because the prison situation provides a highly coercive context that could jeopardize inmates' free, informed consent. Analogous reasoning has led to the prohibition of medical experimentation, not only on prisoners but also on medical and psychiatric patients, on recipients of welfare, and on those who have not given free, informed consent, as in the celebrated Tuskegee syphilis study, which has been called both racist and "ethically unjustified."24 Following these examples from medical ethics, one could argue that a workplace where high wages are paid to economically desperate workers who take hazardous jobs provides a highly coercive context that could jeopardize their alleged free, informed consent. As one labor representative (at government hearings to set occupational exposure standards for lead) put it: "We just cannot

expect a group of employees to be put on a sacrificial table because we happen to need lead and because it is a vital metal and we cannot do without it."25

If we cannot sacrifice workers, then risk evaluators are faced with a value judgment. They must either deny the traditional doctrine of free, informed consent, or they must provide extra compensation to persons who are otherwise unlikely to consent to societal and occupational risks. Either choice involves a value judgment that is itself not risk free.

Where We Go from Here

How might one avoid these five dilemmas? More generally, how might one avoid the unfairness, inequity, or loss of freedom following from various value judgments made as part of risk evaluation? Before moving to some proposed solutions for the dilemmas outlined in this chapter, we must be clear on the precise nature of the difficulties we face. Bound up with each of the five dilemmas are a whole host of biases that cause risk assessors to employ faulty methodological strategies in analyzing and evaluating risks. In the next five chapters, we shall examine some of these problematic strategies. Once that is accomplished, we shall have a clearer idea of the ways in which we need to redefine "rational" risk acceptance and to reform risk evaluation and management.

The Case for a Negotiated Account of Risk and Rationality

Until recently, the study of risk behavior—what some have called "motivated irrationality"—has suffered from a lack of attention by any single group of researchers. Brain scientists at the National Institute of Mental Health recently sought to remedy this deficiency. They gave evidence that those who voluntarily increase their individual (not societal) risk—rock climbers and hang gliders, for example—have lower levels of the brain enzyme monoamine oxidase. They also have lower levels of a brain chemical called DBH and higher levels of gonadal hormones.1

One problem with attributing risk taking simply to biochemical factors, however, is that many people delight in taking certain kinds of risks but detest others. American visitors to China, for example, are appalled at the high percentage of Chinese men who smoke, while most Chinese are surprised at Americans' chronic overconsumption of alcohol, especially when it is associated with half of the fatal automobile accidents in the country and with many types of violence.2 How do we explain such phenomena? Perhaps everyone employs subjective criteria for rational risk behavior. Or perhaps the criteria are not subjective but are so complex that they are impossible to formulate.

If the arguments in the last five chapters have been correct, then at least one fact about responses to societal hazards is certain: they are highly value laden, often in ways that discount lay views of risk, and often in ways that disenfranchise the very persons likely to be risk victims. Now that we have analyzed the value judgments in hazard assessment, and especially in risk evaluation, we are ready to examine some of the more specific techniques and methods used to define ac-

ceptable risk. Fundamental to many risk techniques and methods is what I call the "expert-judgment strategy."

The Expert-Judgment Strategy

Assessors who subscribe to the "expert-judgment strategy" assume that one always can make a legitimate distinction between "actual risk" calculated by experts and so-called "perceived risk" postulated by laypersons.3 They assume that experts grasp real, not perceived, risk, but that the public is able only to know perceived risk.

To understand more precisely what is meant by the expert-judgment strategy, recall that one of the most fundamental sources of divergence, in technology-related risk assessments, is whether evaluation of specific hazards is in part a function of a real, probabilistically described risk or wholly a product of the societal processes by which information concerning the danger is exchanged. More succinctly put, do people fear nuclear power, for example, because they assume that it is an inherently dangerous technology or because they are victims of a sensational media campaign that has played on ignorance and paranoia about technology?4

The expert-judgment strategy consists of the belief either that risk can be reduced to some characteristic of a technology, determined only by experts, or that it is possible for experts alone to distinguish "actual risk," as a property of a technology, from so-called "perceived risk" postulated by laypersons. Once they make the distinction between perceived and real risk, many assessors assume that the perceived risks of laypeople cause most controversy over technology.5 As a consequence, they ask how to mitigate the impact of risk perceptions (which they assume to be erroneous), rather than how to mitigate the impact of risk itself. They assume that public relations, "risk communication," is their only problem.6

The expert-judgment strategy can also occur when assessors presuppose that a technological risk is defined purely in terms of physical impacts. For example, if researchers speak of the "social amplification of risk" as a process whereby hazards produce social impacts (e.g., fear) that exceed their health and safety effects,7 then they appear to presuppose that the risk itself is purely physical and measurable by experts. On the contrary, however, it is arguable that social impacts (such as increased trauma or decreased civil liberties) are part of the risk itself.

Some researchers who fall victim to the expert-judgment strategy also speak, for example, of "perception-caused impacts."8 But there are no impacts that are purely perception induced unless one has uncon-

troversial measures of hazards (versus perceptions of hazards) and uncontroversial measures of risk impacts (versus impacts of risk perceptions). We have wholly exact measures for neither. What we have, instead, is a quantitative, "expert" definition of risk, as opposed to a qualitative, allegedly subjective notion of lay risk perception. Those who fall victim to the expert-judgment strategy typically assume that risk can be defined purely probabilistically, as an average annual probability of fatality.9 They likewise assume that anyone (for instance, a layperson concerned about consent, equity, and the like) who does not subscribe to this purely probabilistic definition has an erroneous risk perception, rather than an accurate, alternative conception of risk. But one cannot stipulatively define one type of risk (that of laypeople) merely as a misperception of another type of risk (that of experts) unless one can distinguish precisely and completely between the risks and perceptions of them.

All Risks Are Perceived, Although They Are Not Wholly Relative

One can distinguish between risks and risk perceptions only if one is able to establish that a perception about a risk (and not the risk itself) caused a particular impact. Yet this is virtually impossible to do, since behavior has multiple causes; sometimes neither a researcher nor the actor knows what those causes are. In particular, it is not enough to establish correlations between particular impacts (such as an aversion to a particular danger) and specific risk perceptions, because this would not show that the perceptions caused the alleged effects. For example, there might be a correlation between catastrophic risks and the impact of high risk aversion. Yet this correlation might be accidental. Instead, the real cause of high risk aversion might be the lack of control over the hazard, not its catastrophic nature. If so, then it may be difficult to distinguish between impacts of risks and impacts of risk perceptions.

To understand why this distinction is problematic, consider another example. Suppose a person becomes an antinuclear activist because he believes that a commercial reactor accident might occur, and because nuclear liability is limited to approximately 1 percent of the total possible losses in the event of a catastrophe.10 The problematic question raised by this case is whether the impact, the person's activism, is caused by the real risk (the risk of not being compensated fully) or by the perceived risk (the perception that he will not be compensated fully).

If one argues that the risk causes the activism, then to mitigate this impact, the threat of noncompensation ought to be removed and the

Price-Anderson Act (which limits nuclear liability) ought to be repealed. If one argues that erroneous risk perceptions (that noncompensation is likely) cause the activism, then mitigating the impact requires removing these allegedly erroneous perceptions. But one can do this only by guaranteeing full compensation, should an accident occur, and hence only by repealing the Price-Anderson Act. Therefore, whether risk or risk perception causes an impact (lay aversion or activism), strategies for mitigating the impact are often the same. But this case suggests that it is often difficult to determine whether risks or risk perceptions cause a particular impact; whether one can distinguish between strategies for mitigating impacts of risks, versus risk perceptions; and therefore whether one can distinguish impacts of risks, versus perceived risks. These difficulties raise the question of whether one can differentiate hazards from perceptions of them.

Our inability to distinguish risks from risk perceptions does not, however, force us into the cultural relativism that we criticized in Chapter Three. Even though all risks are perceived, many of them are also real. The risk of death, for instance, is real. But it is also, in part, a probability, and such probabilities can rarely be known with certainty. Until death becomes a certainty, the risk of death is purely perceived, theoretical, or estimated. Indeed, the occurrence of death, in a particular case, reveals how accurate our perceptions or estimates of the risk of death were. More generally, although the risk that some X will occur is real, the exact degree and nature of this risk are not, in principle, confirmable until X actually occurs. Prior to this occurrence, risk perceptions can be judged as more or less accurate only on the basis of nonempirical and theoretical criteria such as explanatory power, simplicity, or internal coherence. Nevertheless, risk perceptions are often real and objective, at least in the sense that empirical evidence (e.g., accident frequency) is relevant to them and is capable of providing grounds for amending them. All risks (the probability p that some X will occur), then, are both perceived and real. Their exact nature and magnitude become more fully knowable, however, insofar as more instances of X occur. The relativists (discussed earlier) erred in believing that, because all risks are perceived, all risks are therefore relative. (Later in the volume, I shall provide a number of more specific criteria for securing the objectivity and rationality of hazard evaluation.)

Now that we have sketched at least a partial basis for showing how one might argue that this account is not completely relativistic, let's examine why actual risk is not typically distinguishable from perceived risk. One reason is that risk probabilities often do not reflect risk frequencies. This is in part because there are numerous difficulties of haz-

ard estimation that do not admit of analytic, probabilistic resolution by experts. Often the risk problem is not well enough understood to allow accurate predictions, as the use of techniques like fault-free analysis shows. Hence, experts are forced to rely on subjective or perceived risk probabilities, instead of on actual empirical accident frequencies established over a long period of trial. Even if assessors based their notions of probability on actual, empirical, accident frequency, this move would not always deliver their estimates of risk from the charge of being "perceived." Since there are reliable frequencies only for events that have had a long recorded history, use of historical accident/hazard data for new technologies likely results in underestimating the danger, because certain events may not have occurred between the inception of a technology and the end of the period for which the risk information is compiled. Moreover, low accident frequency does not prove low accident probability. Only when the period of observing accident frequency approaches infinity would the two, frequency and probability, converge.

Another reason why one cannot distinguish actual from perceived risk, in any wholly accurate way, is that actual risk estimates are always very rough and imprecise, as the first two sections of Chapter Four indicated. The estimates typically vary from two to six orders of magnitude. Such imprecision is unavoidable, whether the estimates are based on probabilistic calculations or on actual experience. On the one hand, if they are based on probabilities, then assessors are forced to employ a number of value-laden theoretical assumptions and mathematical models. On the other hand, if the risk estimates are based on actual experience, or accident frequency, they are likewise "perceived," because probability does not equal frequency, as has already been argued. Moreover, even actual frequencies do not provide a precise measure of a particular risk, because this number is always formulated as an average, and such averages, by definition, do not take particular, perhaps site-specific, deviations into account.

Also, some of the most important aspects of hazards, whether real or perceived, are not amenable to quantification. (What experts call) "actual" risk estimates are based on the presupposition that risk is measured by probability and consequences, and that both can be quantified. Yet most laypeople would probably claim that what makes a thing most hazardous are factors that are not susceptible to quantification, factors such as a risk's being imposed without consent, or being unknown, or posing a threat to civil liberties.11

Distinguishing risks from risk perceptions is likewise impossible because both are theoretical concepts and hence not amenable to precise empirical

prediction or confirmation. If risks were empirically confirmed or determined, they would be certain, and hence not risks. In general, "risk" is defined in terms of expected-utility theory and hence is a theoretical concept carrying with it all the baggage of this specific decision theory.12 In particular applications, "risk" is always defined on the basis of a whole host of theoretical assumptions, many of which are often controversial. For example, a number of incompatible "cancer models" (dose-response models), each with attendant assumptions, have been used to estimate the incidence of tumors in populations exposed to formaldehyde. In 1987, the U.S. Environmental Protection Agency (EPA) called formaldehyde a "probable human carcinogen." The EPA said that those exposed to formaldehyde-treated pressed wood could face a cancer risk, over ten years, of 2 in 10,000. Experts at the Harvard School of Public Health, however, criticized the EPA risk assessment as premature and said that the true formaldehyde risk was uncertain. Other experts, including scientists at the American Cancer Society and the Consumer Product Safety Commission, argued that the EPA models were incorrect, but in the opposite direction. They said that the EPA findings underestimated the cancer risk. The formaldehyde case—as well as those of ethylene dibromide (EDB), dioxin, and methylene chloride—illustrates that, even as late as the 1980s, particular accounts of risk have been highly controversial and theory laden.13 But if all risk is defined in accordance with the categories of a particular scientific or modeling theory, then there is no actual hazard apart from some particular theoretical account of it. Hence, there is no uncontroversial way to distinguish "actual" from "perceived" risk.

Moreover, because risk perceptions often affect risk probabilities, and vice versa, it is frequently impossible to distinguish hazards from perceptions of them. This phenomenon is well known to social scientists as part of the "self-fulfilling prophecy." For example, if I perceive my chances of getting cancer to be high, then my perceptions can exacerbate stress and therefore increase the probability that I actually do develop cancer. Hence, there is often no clear distinction between actual and perceived risk.

There are also a number of reasons for arguing that the distinction between actual and perceived risk cannot be based on the alleged objectivity of expert estimates, as opposed to the alleged subjectivity of lay risk estimates. Admittedly, laypersons typically overestimate the severity of many technological hazards.14 However, even if it could be established that the public exaggerates the accident probabilities associated with some technology, such as liquefied natural gas (LNG), this fact alone would be a necessary, but not a sufficient, condition for establishing the thesis that laypersons erroneously overestimate risks.

Their risk perceptions could be said to be erroneously high only if they were based solely on incorrect accident probabilities. These perceptions might also be based, for instance, on the potentially catastrophic consequences of an accident, or on the fact that such consequences are involuntarily imposed.

Also, apart from whether probabilities alone explain risk judgments, there is reason to believe that, at least in some areas, expert estimates of probabilities are not necessarily superior to those of laypeople. In their classic studies of the heuristic judgmental strategies that often lead to error in probability estimates, Kahneman, Tversky, and Oskamp concluded that experts, whenever they have merely statistical data, are just as prone as laypeople to judgmental error regarding probabilities. Experts are particularly susceptible, for example, to the fallacy known as representativeness, the gratuitous assumption that a particular sample is similar in relevant respects to its parent population and that it represents the salient features of the process by which it was generated. Kahneman and Tversky showed not only that experts are just as prone to this probabilistic bias as laypeople but also that, even after the error is explained to the experts, the bias cannot be "unlearned."15

Another common judgmental error of mathematically trained professionals is overconfidence. Experts' trust in their probability estimates is typically a function of how much information they have gathered, rather than a function of its accuracy or its predictive success. Since everyone, even those highly trained in probability and statistics, must make simplifying assumptions in estimating probabilities, and since experts are just as prone as laypeople to these judgmental errors, there is little reason to believe that experts are always able to calculate actual or real risk, while laypeople are merely able only to construct perceived or subjective risk.16

Another difficulty is that those who attempt to distinguish "actual risk" and "perceived risk" are wrong to assume that the latter is merely an erroneous understanding of the former.17 They are wrong because there is no universal definition of risk underlying the two concepts. For one thing, experts disagree on whether (and when) to employ concepts such as "individual risk," "relative risk," "population risk," and "absolute risk."18 Moreover, as was already suggested, the term 'risk' in "actual risk" and "perceived risk" has neither the same referent nor the same meaning. What Hafele, Okrent, Jones-Lee, Morgan, and others (see Chapter Three) call "actual risk" is the probability of a particular hazard's occurring, times the magnitude of its consequences. What they call "perceived risk," alleging that it is an incorrect view of actual risk, however, is not merely (an incorrect) perception of probability times consequences. Rather, most laypeople would claim that (what

typical risk assessors call) "perceived risk" includes more than mere probability. Hence, when laypeople say that something is a "high risk," they do not necessarily mean only that it has a high probability of causing death. (See the next chapter for more discussion of this point.) And if so, then "actual risk" is not mere probability times fatality, and "perceived risks" are not merely perceptions of probability times fatality.

In sum, there is no distinction between perceived risks and actual risks because there are no risks except perceived risks. If there were hazards that were not perceived, then we would not know them. That we do know them, in some sense, proves that even real risks are perceived and that even real risks must be known via categories and perceptions. This is related to the earlier point that all risks are, in part, theoretical constructs, not completely empirical, not wholly capable of confirmation.

If we did accept the distinction between actual and perceived risk, such acceptance could lead to undesirable consequences. One negative effect is that the distinction misidentifies the source of controversy over risk judgments and misattributes it to accurate, versus inaccurate, knowledge of probability times accident consequences. Hence, it provides policymakers with little insight on controversy over hazards, little basis for investigating the important ethical, methodological, and evaluative reasons for divergent risk judgments. Moreover, if policy-makers assume that there is a distinction between perceived risk and real risk, then there is less reason for them to take account of lay views, since error could be said to have no rights. As a result, laypersons and other democratic decisionmakers might be disenfranchised.

But if accepting a sharp distinction between risk and risk perception is both politically dangerous and epistemologically confused, then what follows? It makes no sense to talk about risks versus perceived risks, as if experts had some magic window on reality. Instead, we must deal with all hazards as they are perceived, even though (as was just argued) they are not purely relative. We must focus on disputes over risk perceptions and attempt to ameliorate them and the controversy surrounding them.

The EPA: Negotiating Consent and Compensation

If we ought to deal with all risks as risk perceptions, then we might rely on what Liebow and Herborn call an "institutional memory" in order to learn how to use past conflicts as a basis for resolving current

risk controversies.19 One case is particularly instructive in this regard. It has been the nation's most celebrated, lengthiest, and perhaps most costly environmental controversy.

The conflict began in 1964 and pitted the Environmental Protection Agency (EPA) against five New York utility companies. It focused on the potential environmental impacts of Con Ed's proposed Cornwell Project, a pumped-storage facility to be built on a mountain overlooking the Hudson valley. Central to the debate was the impact of the facility's water withdrawals on the Hudson River's striped bass population. As several authors in BioScience put it:

The settlement of the Hudson River case, negotiated between the EPA and the five companies, called for the utilities to deviate from the outage schedule, provided that the overall degree of mitigation of impacts was not reduced. The credit allowed for shutting down a specific generating unit during a given week was determined by the contribution of that unit to the conditional entrainment mortality rate for striped bass.

What scholars learned (from tens of millions of dollars of research spent on biological processes that were impossible to define in the Hudson River case) is that science, even science regarding risk perception and behavior, is often at an impasse. In the Hudson River case, a resolution was reached only after all affected parties decided not to try to define particular impacts and ascertain their causes. Likewise, I suspect that the controversy over risk and risk perception will not be resolved until people stop trying to distinguish perceived risk from real risk. Aristotle astutely noted that the wise person recognizes the degree of certainty appropriate to particular types of inquiries and does not seek a level of certainty inappropriate to the specific kind of investigation. If Aristotle's insight is applicable to the question of hazard

evaluation and management, then perhaps assessors ought to stop the interminable haggling over who is correct, the experts or the public. Instead, perhaps we and they, the experts, ought to shift our focus, in part, from the scientific to the ethical dimension. We ought to attempt to formulate a procedural or negotiated solution as a means of solving the problem of defining, evaluating, and managing risks.

But how does one negotiate among persons so as to resolve some of their conflicts about acceptable risk? Elizabeth Peelle, sociologist at Oak Ridge Laboratories, has argued that creating a citizens' task force to specify mitigation, compensation, and incentive measures for hazardous technologies could help resolve controversies over risk. In Tennessee, for example, such task forces have already succeeded in negotiating consent to hazardous facilities. The "net local balance" of the DOE's (Department of Energy's) proposal to site a monitored retrievable storage (MRS) facility in the state changed from negative to positive after a citizens' task force was organized.21 What the Tennessee experience suggests, and what Chapters Eleven and Twelve of this volume argue in more detail, is that there are both pragmatic and ethical grounds for rejecting the expert-judgment strategy. Instead, we ought to focus on negotiating citizen consent and compensation, regardless of how a risk is perceived or defined.

Conclusion

If any single lesson can be learned from the arguments of this chapter, it is that experts do not always know best—particularly if the problem at issue, distinguishing risks from risk perceptions, is not wholly amenable to theoretical resolution. Like successful adjudication of the Hudson River controversy, adjudicating conflicts over hazard evaluation often requires that we supplement the theoretical task of defining the risk problem in some purely scientific way. Instead, we need also to focus on the practical task of negotiating its resolution in a procedural way.

The Case for a Populist Account of Risk and Rationality

From the fourteenth through the sixteenth centuries, courts executed over a million "witches." According to risk assessor Alvin Weinberg, witch hunts subsided only after the Inquisitor of Spain convened a group of savants who proclaimed that there was no proof that "witches" caused misfortunes. Our current "environmental hypochondria," says Weinberg, is like the hysteria that drove witch hunts, and only savants realize that there is no proof that "environmental insult" causes "real health problems." Weinberg concludes that risk assessors need a new Inquisitor who is able to bring the public to its "senses.1

Like Weinberg, many scientists, risk analysts, and government policymakers have not dealt kindly with the public's distrust of high technologies and industrial toxins. A well-known energy spokesperson has condemned laypersons and environmentalists as victims of "pathological fear" and "near-clinical paranoia."2 Others have said that lay-persons would not fear certain scientific and industrial activities if they only understood that catastrophic accidents are extremely unlikely.3 As the previous chapter argued, many risk assessors subscribe to the expert-judgment strategy, perhaps in part because they see themselves as the experts, the Inquisitors, who ought to bring ignorant and fearful laypersons to their senses.

Many risk assessors believe that laypersons will "come to their senses," in evaluating environmental risks, if they can learn to base their risk aversion on accident probabilities calculated by experts, rather than on their feelings. In other words, many assessors subscribe to the "probabilistic strategy."

The Probabilistic Strategy

The probabilistic strategy is the belief that, for any rational and informed person, there is a linear relationship between a risk, defined as an actual probability of fatality (associated with a particular technological activity), and the value of avoiding the risk posed by that technology.4 Following this strategy, many hazard assessors "explain" a societal aversion to certain low-probability technological risks by alleging that the public does not know the accident probabilities in question. These assessors maintain that, given knowledge of the actual likelihood of death, rational persons always are more averse to high-probability risks than to low-probability ones.

In thus subscribing to the probabilistic strategy, risk analysts likely err, in part, because the restriction of risk to "probability of fatality" is highly questionable. There are obviously many other cost burdens (for example, "decreasing the GNP by a given amount") whose probability also determines the value of avoiding a given risk. Another problem is that the value of avoiding a given risk often depends on whether there are benefits to be gained from it, or whether it is distributed equitably.5 In fact, if Fischhoff and other assessors who employ psychometric surveys are correct, then risk acceptability is more closely correlated with equity than with any other factor.6

Catastrophic potential and the fact that low-probability/high-consequence situations are often the product of societally imposed (as opposed to privately chosen) risks may also explain risk aversion. There is evidence that the psychological trauma (feelings of impotence, depression, rage) associated with the imposition of a public hazard is greater than that associated with the choice of a private risk of the same probability. One author even suggests that widespread despair and an increasing suicide rate may be attributable to the hazards and fatalities caused by "industrial cannibalism."7 If so, then there may be good reason why society's risk aversion is not proportional to the probability of fatality. Moreover, although according to utility theory, a high-probability/low-consequence event (10,000 accidents, each killing one person) and a low-probability/high-consequence situation (one accident killing 10,000 persons) may have the same expected value, reasonable persons are typically more averse to the low-probability/high-consequence situation. One explanation may be that the high-consequence events, such as catastrophic global warming, are often more difficult to quantify.8

If it is true that a risk's importance is measured not only by its probability, but also by other factors, then it makes sense for people

to value the same level of safety differently in different settings. Although this valuing may be economically inefficient, it is neither inconsistent nor irrational. Why, then, do risk assessors believe that societal aversion to allegedly low-probability risks is a consequence of false beliefs about the relevant probabilities?

Probabilities Are Not the Problem

To support their claims, many assessors maintain that laypersons view low-probability nuclear accidents as quite likely. Starr and Whipple, for example (citing research by Otway; Lawless; and Fischhoff, Slovic, and Lichtenstein), argue that "the bulk of disagreement" over nuclear power is over different beliefs about accident probabilities. Likewise, Cohen criticizes the public as being "uninformed" about the real risk probability from hazardous waste and obsessed with regulating risks that are "trivial."9 Starr points out that "society has not learned to place . . . hypothetical man-made events [like nuclear catastrophes]" in the same perspective that it places natural catastrophes such as earthquakes.10 Generalizing on the basis of the controversy over nuclear power, he criticizes public concern with "imaginary large catastrophes."11 Finally, Cohen, Starr, and Whipple suggest that conflicts over technology arise "because of intuitive estimates of unreasonably high risk," and because the public is "emotional" in its risk evaluations.12 This chapter will show that all these claims are questionable.

Starr and Whipple assert that Otway substantiates their claim about the public's misperception of nuclear accident probability.13 But Otway's studies show that, in his view, the real controversy is over values and over incompatible views of the benefits attributed to nuclear power, not over different beliefs about accident probabilities. For example, Otway says that "in general the con [nuclear power] group . . . assign high importance to the risk items while the pro group view benefit-related attributes as most important."14 Otway also claims that, although both pro and con groups "strongly believe that nuclear power is in the hands of big government or business, . . . the pro group evaluates this attribute positively, the con group evaluates it negatively."15 In fact, Otway says explicitly that his research confirms the existence of only three statistically significant differences between pronuclear and anti-nuclear persons, all of which concern "the benefits of nuclear power." The pro group, Otway says, "strongly believed" that nuclear power offers three benefits: an essential good for society, economical energy, and a higher quality of life. On these three points, "the con groups tended to be uncertain to somewhat negative.16 Regarding probabilistic

evaluation of nuclear power, Otway explicitly states: "There were no significant differences between the [pronuclear and antinuclear] groups on the eb [evaluation-belief] scores of any items related to risk."17 This conclusion appears to be a fiat contradiction of the claim that Otway's research supports the probabilistic strategy.

Lawless's work does not seem to support it either. In fact, Lawless never mentions that misperceived probabilities cause disagreements over environmental risk. He argues instead that conflict over technology is greatest where proof of harm is uncertain, not where there is incorrect public perception of certain hazards.18

In the recent controversy over methylene chloride (dichloromethane, DCM), for example, the dispute was clearly a function of uncertainty in scientific knowledge, not a result of incorrect public perception of certain knowledge. DCM is a multipurpose solvent used in paint stripping, metal cleaning, foam blowing, electronics, chemical processing, and in certain aerosol propellant mixtures. Because of its many applications, citizens are exposed to DCM in the workplace, through use of consumer products, and from emissions to ambient air. In 1987, the U.S. Environmental Protection Agency listed DCM as a probable human carcinogen, although industry groups disagreed. The EPA cited the fact that DCM was carcinogenic in mice (although not in hamsters and rats), whereas industry cited the fact that the two studies of humans exposed to DCM in the workplace showed no significant increase in cancer deaths.19

At the heart of the controversy over DCM between industry and environmentalists is the fact that industry tended to use pharmacokinetic models. These permitted the calculation of internal doses of DCM through integrated information on administered dose, the physiological structure of the species involved, and the biochemical properties of DCM. As a result, industry (e.g., Dow Chemical) predicted an average annual probability of fatality, for lifetime exposure to DCM, of 3.7 × 10^-8 . (This risk is below the level of those typically regulated by government.) Other scientists, however, predicted the same risk as 4.1 × 10^-6 for lifetime exposure to DCM. (Two orders of magnitude higher, this latter risk is at the level typically requiring government regulation.) Assessors obtained the higher risk figure by means of more conventional (than the pharmacokinetic) models. These involved linear extrapolation of external DCM dose and an interspecies factor based on body surface area.20

Although numerous other risk cases exemplify the same point, the DCM controversy shows clearly that the dispute was not caused because the public incorrectly perceived the real, or probabilistic, risk. Indeed,

even now (in the early 1990s), the "real" level of risk is unknown (and cannot be calculated) because the carcinogenic mechanism, in the animals in which it has occurred, is unknown, according to the EPA.21 Hence, scientific uncertainty, not faulty public knowledge, appears to be driving the conflict.

Lawless also argues that the nuclear controversy, in particular, has been caused not only by scientific uncertainty, but also by the lack of credibility of federal regulators and by apparent government failure to consider environmental values. He notes, in general, that disputes over risk arose (in more than 50 percent of the cases studied) because technologies were allowed to grow, despite evidence that they were beset with problems, and because they were used irresponsibly.22 All this suggests that laypersons' risk aversion may be reasonable, rather than merely a product of their erroneous risk perceptions.

As with the Otway and Lawless studies, there is little clear evidence—contrary to the claims of Starr, Whipple, and others—that the work of Fischhoff and his associates supports the thesis that environmental controversy arises because of lay misperception of risk probabilities. Their research with the League of Women Voters indicates that the commercial nuclear risk, for example, was not perceived as worth the benefits accruing from it.23 A number of risk assessors, however, appear to dismiss the importance of the question of whether the nuclear benefit is worth the risk.24 Since these same assessors claim that risks and benefits are not evaluated independently, however, it is not clear how they can be so sure that the debate over nuclear safety is primarily over probabilities, rather than whether the benefit is worth the risk.

Fischhoff and his colleagues specifically note that, in their surveys, the public (students and members of the League of Women Voters) judged nuclear power to have "the lowest fatality estimate" for the thirty activities studied, but the "highest perceived risk."25 As a consequence, they conclude, "we can reject the idea that lay people wanted to equate risk with annual fatalities, but were inaccurate in doing so. Apparently, laypeople incorporate other considerations besides annual fatalities into their concept of risk."26

According to the Fischhoff study, the key consideration influencing judgments of high risk was not perceptions of high accident probability but the fact that certain technologies represent an unfamiliar (as opposed to a common) risk; an inequitably (as opposed to equitably) distributed risk; and a hazard with severe consequences in the unlikely event that an accident were to occur.27 In fact, the assessors found that perceived risk could be predicted almost completely accurately solely on the basis of the single variable "severity of consequences," even

though the probability of those consequences' occurring was quite small and was perceived as quite small.28 If this is true, then the suggestion that environmental controversy is fueled primarily by incorrect probability estimates of laypersons is less helpful than the suggestion that, for high-magnitude events, it is the possible consequences that are important to societal evaluation. Several authors have proposed that n lives lost simultaneously in a catastrophic accident should be assessed as a loss of n 2 lives. They argue that the risk-conversion factor for catastrophic accidents should be exponential.29

Admittedly, the studies by Fischhoff and others show only that, contrary to the assertions of other assessors, controversy over hazardous technologies very likely arises because of the value placed on consequences, not because of overestimated risk probabilities. The studies do not show that consequences ought to be valued in this way. Moreover, the Atomic Energy Comission (AEC), the Nuclear Regulatory Commission (NRC), and the courts generally have not attributed much importance to consequences. Courts "have consistently taken the position that probabilities are determinative of risk, regardless of potential consequences."30 Nuclear risk assessments have also consistently adopted the newer nuisance rule that probabilities alone determine risk, probably because of society's interest in technological development. Historically, the rule has owed its inspiration to the reluctance of the nineteenth-century courts to allow the traditionally restrictive law of nuisance to hinder economic progress.31

There are, however, a number of reasons for arguing that, in certain cases, risk consequences are more important than the accident probabilities. For one thing, greater social disruption arises from one massive accident than from a number of single-fatality accidents, even though the same number of people may be killed. The law of torts also recognizes the heightened importance of high-consequence events, apart from their probability of occurrence. In fact, for the rule of strict liability, risk is based almost totally on grave potential consequences, regardless of the associated probability.32 Part of the justification for this judicial emphasis on accident consequences is apparently the fact that the parties involved in litigation over catastrophic accidents—the injured persons and those liable for the injury—are not equal in bargaining power.33 The representative of some technological or industrial interest usually has more clout than the person damaged by it. Moreover, as Chapter Twelve will argue in more detail, a person is more deserving of compensation according to strict liability when she is victimized by an impact that she did not voluntarily accept or help to create. For all these reasons, societal risk evaluation of potentially cata-

strophic technologies ought to focus on the accident consequences as well as on their probabilities.

This point is clear if one considers a rational response to the invitation to play Russian Roulette. Suppose the probability that a bullet is in a chamber when the trigger is pulled is 1 in 17,000—the same likelihood, per reactor-year, as a nuclear core melt. Even with such a small probability, a person could still be rational in her refusal to play the game. She could even maintain that the probability in question is irrelevant. Any probability of fatality might be too high if the benefits deriving from taking the risk were not great enough. And if so, then probabilities might not be as important, in environmental risk evaluation, as proponents of the probabilistic strategy suggest. As one expert expressed it, current debate over whether a given technology has a particular risk probability is a spurious issue. "Risk assessors tend to choose methods and data that support the position to which they are already committed."34 If so, then debate over environmental risks is likely to be over many factors other than probability.

Probabilities Are Often Uncertain

The claim that probabilities are central to risk evaluation, and that "the bulk of disagreement" over environmental hazards has been caused by "intuitive estimates of unreasonably high risk," also errs in ignoring reasonable disagreement over risk probabilities. Risk assessment has been repeatedly criticized as an "arcane expert process" that is over-dependent on probability estimates of assessors.35 Many risk assessors appear to believe that it is "perfectly valid to base public policy on expert estimates and data," but that, once a risk expert has spoken, any disagreement is unreasonable and intuitive.36 Such a notion is doubly questionable.

It is in part implausible because it presupposes a far more objective picture of probabilistic risk data than is now available. Even the authors of the most complete hazard analysis ever accomplished, WASH-1400, cautioned that their probability estimates were deficient, unprovable, possibly incomplete, assumption laden, and saddled with "an appreciable uncertainty." They said that "the present state of knowledge probably will not permit a complete analysis of low-probability accidents in nuclear plants with the precision that would be desirable."37 More generally and more recently, risk assessors have pointed out that "uncertainties of six orders of magnitude are not unusual" in any probabilistic risk analysis.38 In the face of such caveats, alleged certitude about which risk probabilities are correct, and which are incorrect, may be

doubtful. As this chapter has noted, often the scientific mechanisms causing a hazard are unknown, as in the case of methylene chloride.39 And, as the previous chapter also argued, accident probability often cannot be determined on the basis of observed accident frequency. 40 On the one hand, very low values of an accident probability per LNG trip, or per reactor-year, for example, are consistent with an assumed record of zero accidents in 800 voyages, or zero core melts in 17,000 reactor-years. On the other hand, an annual accident probability as high as 1 in 100 or 1 in 200 would still be consistent with the current LNG accident-frequency record, just as a yearly probability as high as 1 in 2,000 would be consistent with the existing record for nuclear accidents. Even though an accident record may be consistent with very low risk-probability values, this frequency alone "does not prove that the values are low."41

Proponents of the probabilistic strategy also err, in emphasizing risk probabilities, because they are unable to account for the reasonable controversy—among Nobel Prize winners, the American Physical Society, the Environmental Protection Agency, the Nuclear Regulatory Commission, and the American Nuclear Society—over various risk probabilities.42 Reputable assessors affirm that many of the most serious environmental risks, such as global warming from burning of fossil fuels, are "highly resistant to quantification.43 Moreover, for a number of reasons, nuclear probabilities, for example, are especially resistant to accurate estimation. Compound events, sequential component failures, sabotage or human error, and weapons proliferation are not amenable to quantification.44 Rasmussen computed the probability of having a Three Mile Island type of accident as anywhere from I in 250 to 1 in 25,000 reactor-years.45 All this suggests that certain accidents are not really "impossible," because many low probabilities are not believable. For example, the probability for a royal flush is 1 in 464,000. Yet, in a card game the probability is actually much higher, since the probability of cheating is likely to be as high as 1 in 10,000. Likewise, although the probability of a given environmental or technological accident may be only very slight, the higher probability of sabotage or terrorism is likely to increase this number by several orders of magnitude. Real risks, then, are likely to include so-called "outrageous events" or "rogue events," which are difficult to handle in probabilistic risk assessment.46 Indeed, human error causes a majority of most industrial, marine, and transportation accidents.47

In claiming that the public overestimates many risk probabilities, such as those for nuclear accidents, many assessors assume that, in some of the most controversial, untested, and potentially catastrophic areas

of technology, it is possible to judge clearly when a risk probability is accurate and when it is not. This is an appeal to authority, an appeal that (given the history of science) simply does not hold up.

Risk assessors' emphasis on the importance of probability estimates is especially vulnerable because the characteristics that, according to various experts, influence judgments of perceived and acceptable risks are highly intercorrelated; involuntary hazards, for example, "tend also to be inequitable and catastrophic."48 .Therefore, as noted in Chapter Six, it is especially difficult to determine whether society's expressed concern about involuntary risks, for example, is merely an artifact of the high correlations between involuntariness and other undesirable risk characteristics. There are numerous allegedly causal explanations, all consistent with the same "observed" phenomena. Kasper makes an analogous observation:

Or, as Cox and Ricci put it, "multiple models, having quite different implications at low doses, may all adequately 'fit' the observed dose-response data.50

Precisely because their hypothesis about misperceived probabilities is consistent with other explanations, proponents of the probabilistic strategy are not warranted in singling out the public's alleged misperceived probabilities as the cause of its high aversion to societal risks. Rather, as the previous chapter argued, the distinction between expert/ objective and lay/subjective determination of environmental risks will not hold up. Because of problems of actual risk calculation (prior to any alleged evaluation), many hazard estimates are merely the intuitive guesses of individuals. Authors of a recent study conducted at the Stanford Research Institute admitted, for example, that analytic techniques could not handle probability estimates for certain human-caused events. They concluded: "We must rely on expert judgment, quantified, using subjective probabilities."51 Likewise, the loss of the astronauts in the Challenger disaster, as well as the death of three astronauts on the ground at Cape Kennedy, demonstrated that even the best systems-analytic approaches cannot anticipate every possibility. In fact, one of the most famous nuclear risk probabilities, widely touted as "objective," is highly value laden. This is the reactor-year probability of a core melt in a nuclear plant, 1 in 17,000. As defended in WASH-1400, this proba-

bility is notoriously laden with value judgments about the effectiveness of evacuation in the face of catastrophe, the probability of weather stability, and the Gaussian Plume rise of radioactivity.52 The problem, however, is not that such "objective" probabilities (as given by experts) are value laden but that they are apparently not recognized as such by proponents of the probabilistic strategy.

Lessons Learned from Experts' Claims about Societal Risks

The tendency of proponents of the probabilistic strategy to overemphasize the importance of risk probabilities and to condemn the public's alleged "misperceptions" of societal risks reveals an important flaw in contemporary environmental risk analysis. Assessors presume that, if there is a public preference for a risk whose probability of fatality is statistically higher than that of an alternative, then this preference is a result of misperceived probabilities, not a legitimate value system.53 Their failure to recognize the value components of allegedly objective probability estimates goes hand in hand with assessors' tendencies to define ethical and political issues as merely technical ones, as the naive positivists are prone to do.54 They assume, incorrectly, that agreement about technical matters is sufficient for resolving normative disputes.

Apparently, they make this assumption because they are afraid of damaging "the scientific pretenses of their work."55 As a consequence, their emphasis on the importance of abstract, "objective" science helps both to disguise the often exploitative way in which technology is used and to condone a passive acceptance of the status quo. It allows assessors to dismiss as irrational or unscientific (as Okrent, Starr, Whipple, Maxey, Cohen, Lee, and others have done) any attempts to challenge our contemporary ethical or political values.56 But, as Dickson has argued, "the use of supposedly objective models of . . . social behavior serves to legitimate the imposition of social policy," because the real risk concerns of laypersons can then be dismissed as subjective.57 As one critic puts it, this is "like playing Monopoly with the Mafia: they always start the game owning Boardwalk."58

Conclusion

Because experts often define risk only in terms of probability of fatality, and consequently neglect ethical and political concerns, they fail to attack an essential problem of risk evaluation: how to make the decision

process more democratic. Later chapters will suggest ways to democratize risk evaluation and management. As Thomas Jefferson warned, the only safe locus of societal power is in the people: "I know of no safe depositor of the ultimate powers of the society but the people themselves; and if we think them not enlightened enough to exercise their control with a wholesome discretion, the remedy is not to take it from them, but to inform their discretion."59

The Case for a "Maximin" Account of Risk and Rationality

A recent U.S. National Academy of Sciences study pointed out that the estimates of increase in bladder cancer, caused by consuming normal amounts of saccharin over seventy years, differed by seven orders of magnitude. Despite these uncertainties, U.S. officials have sanctioned the use of saccharin, justifying their decision on the basis of a liberal risk assessment. As a result of a very conservative risk analysis, however, they banned cyclamates. If the two risk-assessment methodologies were consistent, experts have argued, cyclamates could easily have been shown to present a lower relative risk than saccharin. In Canada, for example, cyclamates are permitted, and saccharin is banned, making Canadian regulations in this area exactly the reverse of those in the United States.1

As the saccharin-cyclamates controversy illustrates, risk evaluations may be uncertain, not only because of the wide range of predicted hazard values, but also because of inconsistent risk models and rules of evaluation. This chapter will assess several of the prominent rules for evaluating risks. To see how alternative rules can generate different risk-management decisions for the same hazard, consider the following case.

The probability of a core melt for U.S. reactors is about I in 4 during their lifetime.2 Risk assessments conducted by both the Ford Foundation-Mitre Corporation and the Union of Concerned Scientists (UCS) agree on the probability and consequence estimates associated with the risk from commercial nuclear fission, but they disagree in their recommendations regarding the advisability of using atomic energy to generate electricity. The UCS risk analysis decided against use of the technology; the Ford-Mitre study advised in favor of it.3

How could hazard assessments and evaluations that agree on the probability and consequences associated with a particular technological accident reach contradictory conclusions about the advisability of using the technology? In many such cases, including that of the UCS-Ford-Mitre controversy, the reason is that the two studies used quite different methodological rules at the third (risk-evaluation) stage of assessment. The Ford-Mitre research was based on the widely accepted Bayesian decision criterion that it is rational to choose the action with the best expected value or utility, where "expected value" or "expected utility" is defined as the weighted sum of all possible consequences of the action, and where the weights are given by the probability associated with each consequence. The UCS recommendation followed the maximin decision rule that it is rational to choose the action that avoids the worst possible consequence of all options.4

To know whether a particular policy analysis provides the best recommendation for rational behavior regarding risk, we obviously need to know under what circumstances a particular decision rule is to be preferred. Ought we to be technocratic liberals and choose a Bayesian rule? Or ought we to be cautious conservatives and follow a maximin strategy?5 The "prevailing opinion" among scholars, according to John Harsanyi, is to use the Bayesian rule,6 even in conditions of uncertainty.7

In this chapter, I argue that the prevailing Bayesian or utilitarian rules are often wrong. Like the expert-judgment strategy and the probabilistic strategy (Chapters Six and Seven), the Bayesian strategy dismisses the views of ordinary people as conservative, pessimistic, and irrational (rather than liberal, optimistic, and rational).8 This chapter argues that there are compelling reasons for rejecting the Bayesian or utilitarian strategy under uncertainty, and that it is often more rational to prefer the maximin strategy.

Harsanyi versus Rawls: Expected Utility versus Maximin

Perhaps the most famous contemporary debate over which decision rules ought to be followed in situations of risk and uncertainty is that between Harvard philosopher John Rawls and Berkeley economist John Harsanyi. Following the utilitarian strategy, Harsanyi takes the Bayesian position that we ought to employ expected-utility maximization as the decision rule in situations of uncertainty, certainty, and risk.9 Typical conditions of uncertainty occur when we have partial or total ignorance about whether a choice will result in a given outcome

with a specific probability; for example, if we use nuclear power to generate electricity, we have partial ignorance about whether that choice will result in the probability that at least 150,000 people will die in a core-melt accident. Under conditions of certainty, we know that a choice will result in a given outcome; for example, if we use nuclear power to generate electricity, we are certain to have the outcome of radwaste to manage. Choices between bets on fair coins are classical examples of decisions under risk, since we can say that we know, with a specific probability, whether a choice will result in a given outcome. (Let's call the Bayesian sense of risk, involving specific, known probabilities, "risk_B " and the sense of risk, often involving uncertainty, with which hazard assessors deal, simply "risk.")

Most technology-related decisionmaking probably takes place in situations of uncertainty. We rarely have complete, accurate knowledge of all the probabilities associated with various outcomes of taking technological risks (e.g., from hazards such as pesticides, liquefied natural gas facilities, and toxic wastes), since very risky technologies are often new. The U.S. National Academy of Sciences confirmed, in a 1983 report, that the basic problems of risk assessment stem from the "uncertainty of the scientific knowledge of the health hazards addressed."10 This statement suggests that many of the difficulties facing risk evaluation concern uncertainty, not "risk_B ." If that is so, then one of the most important questions in the Harsanyi-Rawls expected-utility—maximin debate is what decision rule to follow under the conditions of uncertainty that characterize various technologies.

Harsanyi believes that, under conditions of uncertainty, we should maximize expected utility, where the expected utility of an act for a two-state problem is

u₁p + u ₂ (1 - p ),

where u ₁ and u ₂ are outcome utilities, where p is the probability of S ₁ and (1 - p ) is the probability of S ₂ , and where p represents the decisionmaker's own subjective probability estimate.11 More generally, members of the dominant Bayesian school claim that expected-utility maximization is the appropriate decision rule under uncertainty.12 They claim that we should value outcomes, or societies, in terms of the average amounts of utility (subjective determinations of welfare) realized in them.13

Proponents of maximin maintain that one ought to maximize the minimum—that is, avoid the policy having the worst possible consequences.14 Many of them, including Rawls, take the maximin principle as equivalent to the difference principle. According to a simplified version

of this principle, one society is better than another if the worst-off members of the former do better than the worst-off in the latter.15 As previously noted, the obvious problem is that often the maximin and the Bayesian/utilitarian principles recommend different actions. Consider an easy case:

Although this simplistic example is meant merely to illustrate how proponents of Bayesian utilitarianism and maximin would sanction different social decisions, its specific assumptions make maximin (in this case) appear the more reasonable position. Often, however, the reverse is true. There are other instances, especially situations of risk_B or certainty, in which the Bayesian position is obviously superior. In this chapter, we shall attempt to determine the better decision rule for cases of societal hazard decision under uncertainty (not personal hazard decisions under uncertainty or risk_B ), since environmental hazards are often typified by uncertainty.17 A reasonable way to determine whether the Bayesian/utilitarian or maximin position is superior is to examine

carefully the best contemporary defenses, respectively, of these evaluation rules. The best defenses are probably provided by Harsanyi, an act utilitarian who defends a Bayesian version of utilitarianism,18 and Rawls, a contractarian.

Harsanyi's Arguments

Harsanyi's main arguments in favor of the Bayesian/utilitarian, and against the maximin, strategy under uncertainty are as follows: (1) Those who do not follow the Bayesian/utilitarian strategy make irrational decisions because they ignore probabilities. (2) Failure to follow this strategy leads to irrational and impractical consequences. (3) This failure also leads to unacceptable moral consequences. (4) Using the Bayesian/utilitarian strategy, with the equiprobability assumption, is desirable because it allows one to assign equal a priori probability to everyone's interests.

Rawls's Arguments

Admittedly, discovering difficulties with Harsanyi's arguments for Bayesian/utilitarian rules is not a sufficient condition for rejecting them. We also need to assess maximin, perhaps the best alternative rule for certain classes of cases under uncertainty. To assess this option, let's evaluate Rawls's analysis. His main arguments to support the maximin strategy in situations of uncertainty (his "original position") are as follows: (1) It would lead to giving the interests of the least advantaged the highest priority. (2) It would avoid using a utility function, designed for risk taking, in the area of morals, where it does not belong. (3) It would avoid the Bayesian/utilitarian use of interpersonal comparisons of utility in defining justice. More generally, (4) it would avoid making supererogatory actions a matter of duty, as do utilitarian theories. And (5) it would avoid the Bayesian/utilitarian dependence on uncertain predictions about the consequences of alternative policies.

Maximin and Practical/Prudential Considerations

These arguments about the merits of maximin and Bayesian/utilitarian strategies for decisions under uncertainty suggest several general reasons why the maximin principle might be superior in certain situations. For one thing, Bayesianism/utilitarianism might demand both too much and too little of agents. It might demand too much of them in committing them to supererogatory acts, as necessary for maximizing utility. It might demand too little of agents in committing them to maximizing average utility, even in instances when maximum average utility supports a tyranny of the majority against the minority. Furthermore, although the maximin position is also susceptible to its own types of difficulties (for example, it assumes that we can specify ordinal preferences), these problems often appear less troublesome, especially in situations of uncertainty, than those inherent in the Bayesian approach.145

There are also at least three practical arguments for preferring maximin. For one thing, the wording of the 1969 U.S. National Environmental Policy Act (NEPA) makes it clear that federal policymakers should ensure that every individual enjoys safe and healthy surroundings, not merely that government maximizes average safety or utility.146 In requiring safety for "all Americans" and for "each person," NEPA clearly rejects the Bayesian/utilitarian notion that it is acceptable to deprive a minority of persons of their health or safety.

Common sense, as well as specific laws such as NEPA, likewise enjoin us to use maximin in situations of societal decisionmaking under un-

certainty. One of the most commonsensical of notions is that bureaucracies are notoriously slow moving and inefficient. For this reason, bureaucracies ought not to be trusted, in the absence of specific requirements, to make reliable and timely decisions about life-and-death matters, especially worst-case situations. Perhaps those decisions should instead be governed by prima facie rules, such as "In cases of technological risk under uncertainty, especially cases of potentially catastrophic risk, follow maximin strategies."147

Pearl Harbor illustrates the inefficiency, irresponsibility, and inability of the bureaucracy to "handle the unexpected":

The example suggests that society ought to have procedures for implementing protections, such as the maximin rule, to guard against the inadequacies of bureaucracy. Society might also need a way to avoid the catastrophic consequences of human error. The case of commercial nuclear fission is a classic instance of the need to guard against the results of poor risk evaluations and operator error. "Nowhere are issues of perceived risk more salient or the stakes higher than in the controversy over nuclear power."149 The Chernobyl accident occurred at least partly because assessors called the disaster "highly improbable" before it happened.150 In a Bayesian/utilitarian scheme, highly improbable accidents are not a significant concern in decisionmaking. But in a maximin scheme, the fact that a worst-case nuclear core melt could kill 150,000 persons151 would be a significant cause for concern. Had regulators implemented a maximin strategy, with appropriate procedures for ensuring it, then the Chernobyl accident might have been less likely.

If technological rulemaking created a "climate" of maximin, a climate in which decisionmakers aimed at avoiding worst cases, both they and society would likely be more aware of potential accident conse-

quences, more aware of erroneous subjective probabilities, and more aware of human errors in risk assessment. In their classic works, as was already mentioned, Tversky and Kahneman showed that virtually everyone falls victim to a number of characteristic biases in the interpretation of statistical and probabilistic data. These biases are the result of following erroneous intuitions, and they often result in disastrous consequences. For example, as was mentioned earlier, people often follow an intuition called "representativeness," according to which they believe that various samples are similar to one another and to the population from which they are drawn. In subscribing to this bias, both experts and laypeople are insensitive to such matters as the prior probability of outcomes, sample size, the inability of sampling to obtain a good prediction and to correct itself, the inaccuracy of predictions based on redundant and correlated input variables, and regression toward the mean—despite the fact that training in elementary probability and statistics warns against all these errors.152

Both risk assessors and statistics experts also typically fall victim to a bias called "availability"153 and to the "anchoring" bias.154 These systematic and predictive errors of experts are significant because risk assessment is based on complex theoretical analyses that "include a large component of judgment. Someone, relying on educated intuition, must determine the structure of the problem, the consequences to be considered, and the importance of the various branches of the fault tree."155 According to Kahneman and Tversky, "people do not follow the principles of probability theory in judging the likelihood of uncertain events,"156 and "the same type of systematic errors . . . can be found in the intuitive judgments of sophisticated scientists." These findings indicate that risk assessment, especially Bayesian/utilitarian risk assessment, is likely to produce many erroneous analyses.157 After all, the experts were wrong when they said that irradiating enlarged tonsils was harmless. They were wrong when they said the X-raying feet, to determine shoe size, was harmless. They were wrong when they said that irradiating women's breasts, to alleviate mastitis, was harmless.158 And they were wrong when they said that witnessing A-bomb tests at close range was harmless.159

More specifically, psychometric researchers have concluded that risk experts typically overlook six common "pathways to disaster": (1) They fail to consider the ways in which human error can cause technical systems to fail, as at Three Mile Island. (2) They have too much confidence in current scientific knowledge, even though inadequate scientific knowledge caused such catastrophes as the 1976 collapse of the Teton Dam. (3) They fail to appreciate how technical systems, as a

whole, function. For example, engineers were surprised when cargo-compartment decompression destroyed control systems in some airplanes. (4) Experts also do not take into account chronic, cumulative effects, as in the case of acid rain. (5) They fail to anticipate inadequate human responses to safety measures, such as the failure of Chernobyl officials to evacuate immediately. (6) They fail to anticipate "common-mode" failures simultaneously afflicting systems designed to be independent. A simple fire at Brown's Ferry, Alabama, for example, damaged all five emergency core-cooling systems for the reactor.160 The fact that experts typically overlook these six "pathways to disaster" suggests that they are often unable to model risk situations correctly. As Kahneman and Tversky point out, however, "the usefulness of the normative Bayesian approach to the analysis and the modeling of subjective probability depends primarily . . . on whether the model captures the essential determinants of the judgment process."161 It is important, therefore, both to avoid the errors frequently made by experts and to correct our risk models. By definition, we cannot avoid errors and correct models under uncertainty. Therefore, prudence dictates that we use conservative risk-evaluation rules, such as maximin.

Some Objections to Using Maximin

An important objection to using maximin is that it violates a rationality axiom. This objection loses some of its force if one recalls that Bayesian utilitarians have a similar problem. They ignore ethical and democratic processes and emphasize only outcomes. As was argued earlier, this emphasis provides grounds for doubting the adequacy of the sure-thing principle (one of the three main rationality axioms underlying Bayesianism),162 especially in cases of societal risk under uncertainty.

Another objection to maximin is that it appears to sanction an anti-progress, antiscience notion of risk assessment. To this claim, one could respond that use of Bayesian methods for evaluating societal risks under uncertainty has itself resulted in highly publicized scientific failures, such as the Chernobyl disaster. Such disasters, in turn, have thwarted nuclear progress. If the dependence of Bayesian/utilitarian rules on subjective decisionmakers has contributed to such catastrophes, then restricting the use of Bayesian rules to cases in which they are successful—namely, individual hazards under uncertainty and risk —is likely to promote respect for science, for technical progress, and for Bayesian methods.163

Utilitarians also complain that maximin approaches are more expensive than Bayesian/utilitarian approaches to hazard assessment,

since it costs more to protect society against worst-case, highly improbable hazards than against more probable lesser harms.164 The obvious response to this objection, however, is that worst-case occurrences, such as Bhopal and DES liability, are extremely costly. Moreover, one knows that preventing worst cases is more costly only if one has the assurance that they are highly improbable, which, by definition, one cannot know in a case of uncertainty.

Apart from cost considerations, some experts also allege that maximin is not obviously the only moral choice under uncertainty. This objection, formulated by a statistician,165 is that positive utility attaches to being moral. Hence, for example, the Union Carbide chemical spill could not be justified on Bayesian grounds under conditions of uncertainty. However, there is no in-principle duty under Bayesian utilitarianism to consider moral or legal obligations and prima facie rights. Hence, Bayesian utilitarianism permits, but does not guarantee, avoidance of situations like the Union Carbide chemical spill.

The Case for Minimizing Type-II Errors in Rational Risk Evaluation

In 1973, U.S. assessors established that the fungicide ethylene dibromide (EDB) is carcinogenic. Gathering epidemiological data and following the procedures required for regulation, however, took more than a decade. Meanwhile, EDB began showing up in bread, flour, and cereal products in such quantities that risk assessors predicted that, based on lifetime consumption, EDB would cause up to 200,000 cases of cancer per year in the United States. Immediately after the Environmental Protection Agency announced regulations limiting its presence in food, industry switched to using methyl bromide instead. The main benefit of the new chemical is that it is unregulated, although it is a very close chemical relative of EDB and almost certainly highly carcinogenic. Policymakers expect it to take another decade to develop regulations to cover methyl bromide. Meanwhile, it is being used in bread, flour, and cereal products.1

Apart from whether industry ought to have regulated itself and not begun use of methyl bromide after EDB was banned, this case raises another question. What is appropriate risk behavior when methyl bromide carcinogenicity is not absolutely certain? If the arguments presented in Chapter Eight are correct, then in situations of uncertainty, especially those involving potential catastrophes, policymakers ought to act so as to avoid the worst case. In this fungicide instance, the worst case is probably something like 200,000 annual cancers caused by the chemical. In order to avoid the "worst case," we might need to protect the public from high-consequence technological dangers that are less likely to occur than more ordinary, lower-consequence events. But—because it is more expensive to protect against improbable, worst-case accidents than against more likely ones—this higher level of protection

typically leads to potential financial losses for industry and to the possibility that a certain technology will be rejected on grounds of safety.2

All this indicates that choosing a maximin strategy, in cases of uncertainty (see the previous chapter), typically minimizes public risk (to citizens) and maximizes industry risk (to those responsible for the dangerous technology). But this raises the question of whether, in a situation of uncertainty where we must do one or the other, we ought to minimize industry risk or public risk. This chapter argues that rational risk evaluation and management often requires us to minimize public risk.

Type-I (Industry) Risks and Type-II (Public) Risks

The concepts of industry and public risks are related to those of type-I and type-II statistical error. In a situation of uncertainty, errors of type I occur when one rejects a null hypothesis that is true; errors of type II occur when one fails to reject a null hypothesis that is false. Statistics dictates that we make assumptions about the size of each of these types of error that can be tolerated, and, on this basis, we choose a testing pattern for our hypothesis. This concept of significance, for example, is often defined in terms of a type-I risk of error of either 0.01 or 0.05, where there is not more than a I in 100 or a 5 in 100 chance of committing the error of rejecting a true hypothesis.

Determining significance, however, is not a sufficient basis for answering an important question. Given a situation of uncertainty, which is the more serious error, type I or type II? An analogous issue arises in law. Is the more serious error to acquit a guilty person or to convict an innocent person? In assessing technological impacts, ought one to minimize type-I risk, which Churchman terms the "Producer Risk," or ought one to minimize type-II risk, which he calls the "Consumer Risk"? (I also call the producer risk and the consumer risk, respectively, "industry risk" and "public risk.") That is, ought one to run the risk of rejecting a true null hypothesis, of not using a technology that is really acceptable and sate; or ought one to run the risk of not rejecting a false null hypothesis, of using a technology that is really unacceptable and unsafe? To decrease industry risk might hurt the public, and to decrease public risk might hurt industry.3

Why Risk Assessors Tend to Minimize Industry Risk

Just as most experts (risk assessors) probably tend to pursue the dominant Bayesian strategy, even in situations of uncertainty, they also

probably tend to minimize producer or industry risk and to maximize consumer or public risk.4 This tendency likely arises because preferences for type-II errors and for minimizing type-I risks appear more consistent with scientific practice. Hypothesis testing in science operates on the basis of limiting false positives (assertions of effects where none exists), or limiting incorrect rejections of the null hypothesis. In order to minimize type-I errors, scientists design studies to guard against the influence of all possible confounding variables, and they demand replication of study results before accepting them as supporting a particular hypothesis. They apply tests of statistical significance, which reject results whose probability of occurring by chance (whose p value) is greater than, for example, 5 percent. Moreover, it is difficult to see how the scientific enterprise could function without such rigorous reluctance to accept positive results. As Abraham Kaplan put it: "The scientist usually attaches a greater loss to accepting a falsehood than to failing to acknowledge a truth. As a result, there is a certain conservatism or inertia in the scientific enterprise, often rationalized as the healthy skepticism characteristic of the scientific temper."5

The preference for type-II errors (for consumer or public risks) and for minimizing type-I errors (industry risks) is also consistent with the standards of proof required in criminal cases, as opposed to cases in torts. Our law requires the jury in a criminal case to be sure beyond a reasonable doubt that a defendant is guilty before deciding against him; standards of proof in criminal cases thus also reveal a preference for type-II error, a preference for accepting the null hypothesis or innocence, a preference for the risk of acquitting a guilty person. In a case in torts, however, our law requires the jury to believe only that it is more probable than not that the defendant is guilty; standards of proof in civil cases thus reveal no preference for either type-I or type-II error.

As Judith Jarvis Thomson points out, "our society takes the view that in a criminal case, the society's potential mistake-loss is very much greater than the society's potential omission-loss."6 Presumably, this difference in the standards for proof in civil and criminal law stems at least in part from the fact that the consequences to the defendant in the criminal case are more likely to lead to grave harms (such as death or life imprisonment) than are the consequences to society.7 Moreover, the state needs to protect its moral legitimacy by minimizing type-I risks in criminal cases. If it fails to convict the guilty, the state commits wrong in a more passive (less reprehensible) sense than if it errs by convicting the innocent. For these reasons, if standards of proof in cases of industry (or type-I) risks were analogous to those in criminal

cases, then this analogy might provide a good reason for minimizing industry, rather than public, risk. (As the next section of this chapter will argue, however, they are analogous neither to hypothesis testing in pure science nor to determination of guilt in criminal cases.)

Preferences for type-II error or public risk might also arise from the fact that many risk assessments and impact analyses are done by those who are closely associated with the technology being evaluated and who are therefore sympathetic to it and to those who implement it.8 In such cases, assessors typically underestimate risk probabilities,9 at least in part because it is difficult to identify all hazards, and because unidentified risks are usually assumed to be zero. Minimizing industry risk probably also arises because technical experts almost always use widely accepted Bayesian decision rules based on expected utility and subjective probabilities, rather than the maximin principle.10 As a re-suit, even when everyone agrees that the probability of a high-consequence impact is uncertain but low, using a Bayesian decision rule typically generates a choice in favor of the potentially catastrophic technology or environmental impact, whereas using a maximin principle likely produces a verdict against the technology or environmental impact.11

Why Analogies from Scientific Practice and Criminal Law Are Not Applicable

Contrary to most assessors, I shall argue that there are prima facie grounds for minimizing public, rather than industry, risk. Before doing so, however, I will attempt to explain why standards of proof in assessing technological and environmental risks under uncertainty are analogous neither to hypothesis testing in pure science nor to determination of guilt in criminal cases. If my explanation is correct, then neither science nor criminal law provides arguments for minimizing industry, over public, risk.

Researchers doing pure science apparently prefer to minimize type-I error and to risk type-II error (consumer or public risk) because it is a more conservative course, epistemologically speaking, than is risking type-I error (industry risk). It is more conservative in the sense that it avoids positing an effect (e.g., that a substance causes cancer above a given level). Instead, it presupposes that the null hypothesis is correct (e.g., that a substance causes no cancer above a given level). Hence, it is reasonable to claim that one ought to follow a model of "epistemological rationality," rationality concerned with maximizing truth and avoiding positive error, when one is engaged in pure science.

Societal decisionmaking under uncertainty, however, is arguably not analogous to decisionmaking in pure science. "Epistemological rationality" is an insufficient basis for making decisions about whether to minimize industry and public risks affecting the welfare of many people. This is because, as was argued in the previous chapter, judgments about societal welfare involve "cultural rationality," and hence an assessment of the democratic justifiability of the risk imposition. Individual risk cases, or cases of purely scientific decisionmaking, might involve a substantive concept of epistemological rationality; but cases involving societal risk decisions require a procedural concept of rationality, one able to take account of a process for recognizing ethical and legal obligations. When one moves from pure science to applied science affecting policy, the question of what is rational moves from epistemological considerations to both ethical and epistemological ones. Therefore, the fact that pure scientists minimize type-I errors provides no compelling reasons for arguing that societal decisionmakers ought to minimize type-I errors.

Likewise, the legal system provides no strong reasons for minimizing type-I errors. For one thing, civil law exhibits no preference for minimizing type-I errors. Although criminal law does reveal such a preference, it does not appear to be analogous to the societal situation of technological or environmental risk under uncertainty. In protecting the accused criminal, presupposing his innocence, and therefore minimizing type-I errors, the criminal law is protecting the most vulnerable person. Since an accused criminal is more vulnerable than his potential or alleged societal victims, harm to him is more serious than harm to society, in the event of a mistake. In a case of societal risk from technology, however, harm to the public is more serious than harm to industry. If decisionmakers err in assuming that a given technology is safe, then this type-II error could result in loss of life by members of the public. However, if decisionmakers err in assuming that a given technology is harmful and therefore err in rejecting it, then the main losses for industry are economic. Hence, if the aim of decisionmaking under risk is to avoid the more serious harms, this case is not analogous to that of the criminal being tried, since in the risk case the greatest threat is to the public, and in the criminal case the greatest threat is to the defendant.

Moreover, if Thomson is correct in claiming that criminal law shows a presupposition in favor of minimizing type-I error, industry risk, because alleged criminals are more vulnerable than are societal victims of criminals, then this reason also shows that the criminal case and the societal risk cases are disanalogous. The public is far more vulnerable

than are industrial producers and users of risky technologies. The public is privy to less information about the alleged risks, and the public typically has fewer financial resources to use in avoiding the risks. Also, it is extremely difficult for victims of societal/industrial hazards (for example, people who have cancer) to exercise their due-process rights, since such an exercise requires that they prove causality in the case of a technological risk.12 Hence, if the public is more vulnerable than industrial producers of risk, protecting the more vulnerable persons requires that one minimize public or type-II risks (rather than the type-I risks minimized by the criminal law).

The Prima Facie Case for Minimizing Public Risk

In arguing that there are prima facie grounds for reducing public (rather than industry) risk, I intend to show that, all things being equal, one ought to take this approach. Obviously, however, the decision to minimize either public or industry risk must be decided in part on a case-by-case basis (so that, for example, the particular benefits at stake are considered). Arguing that there are prima facie grounds for reducing public risk therefore amounts to arguing that the burden of proof (regarding risk acceptability) should be placed on the person wishing to reduce industry, rather than public, risk. If that person cannot provide evidence to the contrary, one ought to minimize public risk.

There are at least eight different reasons for holding that assessors' prima facie duty is to minimize the chance that an unsafe technology is implemented; that is, to minimize public risk. The first five arguments stress that the dangers faced by the public (rather than by industry) represent the kind of risk most deserving of reduction. The last three arguments focus on the public as the locus of decisionmaking regarding societal hazards, since laypeople typically argue for reducing public risk.

The Case for an Egalitarian Account of Rational Risk Management

Between 300,000 and 400,000 of the 1,000,000 current and former U.S. asbestos workers are expected to die of occupation-induced cancer. As a result, the U.S. Occupational Safety and Health Administration (OSHA) has issued new workplace standards, and manufacturers have developed cleaner technologies for processing asbestos. Rather than installing the safer technologies, however, many U.S. corporations are continuing to use the dirtier methods and moving their asbestos operations to other countries, notably Mexico. For example, Amatex, a Norristown, Pennsylvania, firm, closed its U.S. asbestos facilities and opened plants in Agua Prieta and Ciudad Juárez, Mexico, both just across the border. There are no Mexican regulations to protect workers from asbestos, dust levels in the Mexican plants are not monitored, and workers wear no respirators. Employees receive minimum wage and are told nothing about the hazards they face. Asbestos waste covers the factory floor and clings to the fence and the dirt road, behind the factories, where Mexican children walk to school.1

Asbestos processing is not the only hazardous technology typically transferred from the United States. Recently the Nedlog Technology Group, Inc., of Arvada, Colorado, offered the president of Sierra Leone up to $25 million to dump millions of tons of toxic chemical wastes in his west African nation. Each year U.S. companies offer nations in the Caribbean and in west Africa hundreds of dollars for every fifty-five-gallon barrel of toxic waste that can be dumped legally.2 One of the greatest problems with transfer of hazardous technologies arises in connection with pesticides. Massive advertising campaigns by corporations such as Dow and Chevron have turned the third world into both a market and a dumping ground for dangerous chemicals, es-

pecially DDT. For example, Ortho (a division of Chevron and an arm of Standard Oil of California) in Costa Rica is the main importer of eight banned or heavily restricted U.S. pesticides: parathion, DDT, aldrin, dieldrin, heptachlor, chlordane, endrin, and benzene hexa-chloride (BHC). In Ecuador, Shell, Velsicol, Bayer, American Cyanamid, Hercules, and Monsanto are the main importers of pesticides banned in the United States. In Colombia, fourteen different corporations import virtually every U.S. pesticide banned since 1970.3

The consequences of third-world use of banned pesticides are not small. According to the U.S. General Accounting Office (GAO), 29 percent of all U.S. pesticide exports are products that are banned (20 percent) or not registered (9 percent) for use in the United States. The World Health Organization (WHO) estimates that there are approximately half a million cases of pesticide poisoning annually, with a death-to-poisoning ratio of one to ten. This means that about 49,000 persons, many in developing nations, die annually from pesticides. One person is poisoned by pesticides every minute in developing countries.4

The fundamental moral problem raised by each of these cases is whether corporations have an obligation to guarantee equal protection from risks across national boundaries. Or do they simply have an obligation to provide whatever protection is legally required in the country to which they export? Perhaps the dominant attitude toward transfers of hazardous technologies is that the ethics of risk evaluation in developed nations is isolated or separate from analogous moral requirements in developing countries. I call this view the "isolationist strategy." It sanctions corporate transfers of hazardous technologies to other countries, provided only that the risk imposition meets whatever conditions are imposed by the host nation. For third-world peoples, these conditions are typically minimal or nonexistent.

Advocates of the isolationist strategy characteristically reject risks close to them in space or time but sanction those that are distant from them. The object of this chapter is to provide some grounds for challenging the isolationist strategy, for questioning the view that we are not morally responsible for hazards that are spatially or temporally distant from us. In order to evaluate this strategy, I discuss four main arguments used to justify transfers of hazardous technologies to third-world countries. Next l show why all these arguments, except the last, can be defused easily. If my analysis is correct, then effective action to protect citizens of third-world nations may demand not only individualistic efforts but also coordinated political activity. Indeed, a rational risk response may require political activity that is nothing less than revolutionary.5

The Social-Progress Argument

Transfers of hazardous technologies often are defended on the grounds that one is not ethically bound to accept any principles of equal protection for all persons. Many utilitarian moral philosophers, especially act utilitarians, for example, are opposed to accepting principles of equal protection, whether within a nation or across nations.6 For this reason, many act utilitarians would likely hold some variant of what I call the "social-progress argument." They would maintain that, although they do not wish to see third-world peoples killed or injured by asbestos, hazardous wastes, or banned pesticides, nevertheless adopting a principle of equal treatment or equal protection for all persons could jeopardize social progress. Act utilitarians, such as J. J. C. Smart, also typically believe that more human suffering is caused by following principles of equal treatment than by attempting to maximize the well-being of the majority. They believe there is no "right" to equal treatment and equal opportunity because, if there were, then economic and social improvements for the majority of the people would be delayed.7

Good act utilitarians, pursuing the social-progress argument, might point out, for example, that worker fatalities during the building of the U.S. transcontinental railroad reached a peak of approximately three per thousand per year.8 Although this death rate is three orders of magnitude greater than the current, allegedly acceptable level of regulated risk in the United States, they might view it as a necessary evil.9 It might be seen as the price paid to ensure a greater good— western expansion and consequent economic growth bringing an extraordinary level of prosperity to a majority of U.S. citizens. If failure to provide equal protection to these railway workers was a partial cause of widespread prosperity that otherwise would not have been possible, a number of act utilitarians would likely justify it.

The main problem with the social-progress argument, however, is its presupposition that there is no in-principle obligation to recognize individual rights, and that there are ethical grounds for sacrificing the welfare of significant numbers of human beings for the sake of the majority. This presupposition is questionable in part because it is inconsistent with individualistic, democratic, liberal traditions that are embodied, for example, in the U.S. Bill of Rights. The argument is also problematic because act utilitarians admit that, in their view, not every individual would be protected from capricious or expedient denials of justice.10 This admission renders their argument problematic, for reasons outlined in Chapter Eight—namely, that discrimination is

unjustified unless it works to the advantage of everyone, including those discriminated against.11

The Countervailing-Benefits Argument

Another argument for transfer of hazardous technologies to third-world nations is the "countervailing-benefits" argument. It amounts to the claim that, although it would normally be wrong to transfer technologies known to cause injury and death, recipients of risky technologies are better off than they would have been without them. Proponents of the countervailing-benefits argument admit that, although there are health costs, for example, to third-world asbestos workers or victims supplied with banned U.S. pesticides and toxic wastes, there are also benefits that are part of the same package. And the crucial point, in their view, is that the benefits outweigh the costs. The Mexican asbestos worker might not have a job if he did not work in substandard asbestos-production facilities, and the African village might have neither a local school nor clean water were it not for the revenues supplied by storing toxic wastes from the United States. The argument is that a bloody loaf of bread is sometimes better than no loaf at all, that a dangerous job is preferable to no job, and that food riddled with banned pesticides is better than no food at all.12

Perhaps the greatest presupposition of the countervailing-benefits argument is that any cost is allowable as long as the benefits are greater,13 One could easily challenge this assumption by arguing that some costs are preventable evils that should never be allowed, even for countervailing benefits. One might argue that not everything "has its price." Since utilitarians would typically be the moral philosophers most likely to claim that "everything has its price," one could invalidate this claim by showing that not even all utilitarians support the counter-vailing-benefits argument. Mill, for example, would not support it.

The Consent Argument

Even if transferred risks do threaten individual security, one might argue that the recipients of the technology consented to them. Moreover, goes the objection, unless we want to deny the autonomy of native peoples and their rights to make their own choices, we are bound to allow them to have the technology transfers that they request. Even if such transfers involve substandard asbestos processing or the import of pesticides banned in the United States, third-world peoples have a right to determine their own fate.36 In a nutshell, this "consent argument" is that corporations are not morally responsible for inflicting harm through technology transfer because the recipients agreed to it.

The plausibility of the consent argument rests in part on the classical economic theory of the compensating wage differential: when people accept risky jobs for higher pay, they implicitly consent to the hazards. As Adam Smith expressed it, "the whole of the advantages and disadvantages of the different employments of labor" continually tend toward equality, because the wages vary according to the hardship of the occupation.37 Analogously, proponents of the consent argument might claim that imposition of greater public health risks is acceptable because citizens voluntarily agree to trade some societal safety for greater public benefits, such as a stronger economy or a higher standard of living.

Clearly, the acceptability of the consent argument depends on whether recipients of technology transfer (such as Mexican asbestos company employees, Central American farm workers who use pesticides, and African workers in hazardous waste facilities) accepted these technological risks in situations of informed consent. That is, it depends on whether (1) the workers were informed of the severity and probability

of harm, and, after being given this knowledge, they freely chose to work; and (2) the governments allowing imports of hazardous technologies, such as banned pesticides, also gave free, informed consent to the risks. But both the amount of information and the level of freedom likely to be present in cases of technology transfer to many third-world countries are open to question.

Just because a third-world worker, for example, holds a particular risky job, it cannot be assumed that that occupation is an expression of the worker's authentic preferences. Many people engage in certain work, not because they voluntarily and autonomously choose to do so, but because they have no alternatives. Moreover, in the absence of minimum standards for occupational safety, and without alternative opportunities for employment, one could hardly claim that the worker's occupation and its attendant risks were the result of autonomous or free choice.

Several years ago in the United Kingdom, for example, the government's Advisory Committee on the Safety of Pesticides (PAC) was locked in battle with the National Union of Agricultural and Allied Workers (NUAAW) over the spraying of 2,4,5-T by farm workers. The PAC asserted that the pesticide was safe when used properly. The NUAAW responded that worker consent, safety, and distribution of the pesticide were precisely the areas of concern. Its argument was that "the organizational realities of farm life often do not allow a farm worker to refuse to spray just because the climate is not correct, or because specified protective equipment is defective or nonexistent. Chemicals called 'adjuvants' that speed up the action of the main chemical are often added . . . and new spraying technologies designed to improve economic efficiency have had marked effects on exposures." In other words, the cultural reality of a low-paid, "dispensable" farm worker does not allow him to say that he is concerned about risks. Hence, such workers are not likely to give free, informed consent to the risks they incur.38

A similar example concerns the recent conflict over hormones fed to beef cattle. In 1985, a scientific committee of the European Commission said that certain "growth promoters" were safe if used (1) by means of earlobe injection, (2) with a specified dose threshold, and (3) in connection with a ninety-day waiting period before sale of the cattle. The Council of Ministers rejected the alleged safe use of the hormones on the grounds that such conditions of use are not enforceable in reality. Consequently, the public is unable to give informed consent to the beef hormone injection.39

Market constraints or greed often militate against the conditions

necessary for free, informed consent. For example, after the 1985 Bhopal disaster, a French board of inquiry discovered numerous improprieties in the handling of the toxin methyl isocyanate (MIC) in France. MIC was imported through Marseilles and sent to a plant in Béziers. At the Marseilles docks, however, because of the economics of unloading operations (e.g., piece rates being paid to increase productivity), and the necessity to fill shifts productively, barrels of MIC were being thrown, lifted, and hauled as if they were bales of straw. The cultural and economic realities of the dock situation made free, informed consent (among workers and residents living near the docks) highly questionable.40

One reason why an occupation and its associated risks are not necessarily the result of a free decision is that job choices are often not made in the context of what John Rawls might call ethically desirable "background conditions." Such background conditions might include the operation of a free market, lack of coercion by employers, and the existence of alternative employment opportunities. I f background conditions necessary for procedurally just, voluntary employment decisions are not met, an appeal to the theory of informed consent cannot justify exposing persons to workplace hazards imposed because of technology transfer.41

Consider a farm worker, for example, hired to apply pesticides in a third-world country. It is well known that such jobs are very risky and also that, as education and income rise, employees are far less likely to remain in hazardous occupations. Therefore, workers in high-risk jobs are likely to be both financially strapped and poorly educated. Moreover, the situations in which third-world peoples would be most in need of work are precisely those in which background conditions are likely to preclude genuine free consent to accepting those jobs. In Mexico, for example, the unemployment rate is typically 50 percent. This suggests that, in rural third-world countries likely to employ pesticides, for instance, there is probably no diversified economy that would provide a variety of alternative employment options. Hence, the situations in which one most needs to take risky work are precisely those in which free consent could not likely be given to the job.42

Indeed, for half the world's population, free, informed consent may not even be possible. About 800 million people, one-fifth of the humans on the planet, are deprived of all income, goods, and hope. They live primarily in India, Bangladesh, Pakistan, Indonesia, sub-Saharan Africa, the Middle East, Latin America, and the Caribbean. Another one-fifth to two-fifths of the world's people, above the one-fifth that Robert McNamara called the "absolute poor," are chronically malnourished.

Given pervasive disease, malnutrition, illiteracy, and squalor—not to mention few job alternatives and an economy likely not diversified—it is questionable whether, even with perfect information on the risks involved, employees could be said to freely choose to work with hazardous technology.43

Often consent is not likely to be truly informed, since the same third-world conditions that militate against free consent also militate against education. An isolated African or Latin American region where banned pesticides are used, for example, is unlikely to have an educated populace to help make citizens aware of pesticide danger. It is also unlikely to have a local chapter of the Sierra Club or of Ralph Nader's Public Interest Research Group or any other social institution that could help remedy the citizens' inability to give free, informed consent.

Moreover, even in some of the most developed countries of the world, such as the United States, where societal institutions are in place, free, informed consent is still rare. When the Binghamton (New York) state office building caught tire recently, it was highly questionable whether the accident victims gave free, informed consent to the risk of reentering the building. The fire spewed about 180 gallons of coolant containing polychlorinated biphenyls (PCBs), from the electrical transformers in the mechanical room, throughout the building. Later, despite the fact that the state office building's garage was contaminated with PCBs, officials opened the garage because of "the shortage of parking space in downtown Binghamton." It was opened only because officials withheld crucial information about testing the garage and about the toxicity of PCBs. The director of health for the state "intentionally concealed important information . . . to mollify public concern."44 If even highly developed nations cannot guarantee free, informed consent to their citizens, it is questionable whether the consent argument is able to justify transfer of hazardous technologies to less developed countries.

The Reasonable-Possibility Argument

If the analysis thus far has been correct, all three arguments enlisted to support transfer of hazardous technologies face serious objections. Someone could still maintain, however, that such transters are legitimate because it is impossible to prevent them. I call this the reasonable-possibility argument. It is based on the ethical maxim "Ought implies can "; if corporations ought to be required not to transfer banned technologies to third-world countries, then this requirement must be one that can be achieved. If it is not achievable, then it is not required.

The main reason for believing that it might not be possible for a corporation to introduce safer technology on its own, in the absence of mechanisms to control the behavior of competing firms, is that such an action could financially destroy a company. Moreover, goes the argument, governments (not individual corporations) are in the business of promoting and regulating worker and citizen safety. To expect a firm to introduce safer technology, and thus be undercut by other corporations with fewer moral qualms, is thus ethically questionable. It is to impose an unrealistic, heroic, and self-sacrificial burden on a corporation. And morality does not require heroism, only justice.45 That is why Gewirth, for example, in his classic argument for the absolute right not to have cancer inflicted on one, argues that it is necessary for the state to regulate and enforce this right. One cannot expect voluntary compliance with strict environmental and technological standards.46

Attorney Richard Stewart likewise has recognized that strong federal regulation, rather than heroism, is necessary to restrain technology. Stewart points out that states cannot afford to impose more stringent environmental standards than their neighbors, unless they want to hurt their economy. Similarly, it could be argued that corporations cannot accept more stringent standards than other firms, unless they want to go out of business.47

A Possible Moral Response to the Reasonable-Possibility Argument

Despite the plausibility of the suggestions that morality cannot rest on heroism, many philosophers are likely to respond that it is both reasonable and possible—not heroic—to cease transfer of hazardous technologies. In subsequent paragraphs, I argue that this response is questionable. Henry Shue, for example, claims that corporations are morally bound to cease transfer of hazardous technologies because (1) no institution has the right to inflict harm, even to hold down production costs; and (2) underdeveloped countries, alone, cannot be expected to impose strict environmental and technological standards because they are competing with other countries for foreign investment.48 Although Shue's first argument is correct, that one ought not inflict harm in order to hold down production costs, a critical problem is knowing how to define "infliction of harm." At what point does inflicting a higher probability of damage constitute infliction of harm?49 Contrary to an assumption behind Shue's argument, we do inflict harm in the form of increased probability of risk, in order to hold down pro-

duction costs in the United States. Our pollution-control regulations are specifically designed to trade a certain amount of safety for a given amount of production savings. The typical norm, adopted by the Environmental Protection Agency, a National Academy of Sciences panel, the Nuclear Regulatory Committee, and other government groups, is that safer technology is not required unless the risk to the public is greater than a one-in-a-million increase in the average annual probability of fatality. Moreover, allowable worker risk is typically ten times greater than that for the public, in part because permitting higher workplace risks is sometimes more cost-effective than prohibiting them.50

Many U.S. corporations often are merely required to keep environmental hazards "as low as is reasonably achievable," on the basis of a "favorable cost-benefit analysis." If it costs the licensee of a nuclear power plant, for example, more than $1,000 to avoid an additional person-rem of exposure to the public, then the licensee is not required to do so. If it costs less, then the licensee must aim at reducing maximum radiation exposure to the public to 0.0005 rem per person per year. Hence, according to current law, there is no (and indeed there cannot be an) absolute prohibition against harm (where "harm" includes increased probability of risk), in part because such a prohibition would be impossible to achieve in a technological society.51 Therefore, Shue's first argument, as it stands, may sanction a morality (absolute prohibition of harm) that is impossible to fulfill.

Shue's second objection, that underdeveloped countries cannot be expected to impose strict environmental standards because they are competing with other nations for foreign investment, also makes a reasonable point, but it contains a flawed assumption: that because countries compete with one another for foreign technological-investment dollars, just as corporations compete with one another for profits, therefore nations have no more responsibility than do private industries to protect their citizens' health and safety by regulating technology. This assumption is flawed because it puts countries and corporations on the same level, so to speak, as far as protecting citizens is concerned. Putting them on the same level is problematic, because firms are concerned primarily with promoting private interests (maximizing shareholders' profits), whereas nations are obliged to promote the public welfare. Moreover, a strong case could be made for the claim that citizens, by virtue of their citizenship, share an explicit contract with their country. In exchange for citizens' paying taxes and remaining loyal, for example, the country performs many services, such as protecting the health and welfare of the public. Except in the employer-employee relationship, there is no comparably strong contract between

a corporation and citizens in a given nation. Therefore, it could easily be argued that the greater responsibility for protecting public health and welfare belongs to the country, and not to the corporation that imports pesticides or agrees to manage toxic wastes. Moreover, at least in part, the nation appears to have the stronger obligation to protect citizens because corporations often fail to do so.

Consider the consequences that would follow if one were to accept Shue's notion that corporations have more responsibility to force use of safe technology than do host countries. Corporations that did not willingly accept this responsibility would be able to act with impunity, knowing that they, not government, had primary responsibility for citizen safety. And government would be powerless to "right" corporate wrongs, since it would have no mandate to protect citizens working in risky facilities. Indeed, one of the most common industry arguments against government regulation is that it is "not needed" and that corporations themselves can do the job. This is exactly the argument made by Henry Shue. Obviously, however, industry cannot police itself completely, as our earlier examples of risks suggest.52 If it could, then it would have nothing to lose through government regulation. Because firms have something to lose, they oppose governmental regulation; therefore, such regulation must be needed.

A Practical Response to the Reasonable-Possibility Argument

If government regulation is typically needed to protect citizens and workers from environmental/technological hazards, and if industry alone cannot do the job, then it may be neither reasonable nor possible to expect corporations to cease transfer of banned technologies, especially if government does not require them to do so. Since "ought implies can," corporations are morally obliged to use safer technologies only if they can do so without heroic sacrifices. Moreover, even if it were reasonable to argue that firms are morally obliged to make heroic sacrifices, they are unlikely to do so, at least for long, because they will not survive. Hence, even if corporations are morally required to use safer technologies, they are not likely to do so if the safety threatens their competitive advantage. Apart from what is ethically desirable, one cannot realistically expect certain companies to cut their profits, in the name of safety, unless governments, corporate employees, and consumers force them to do so. We are therefore faced with an interesting practical problem, one quite different from the one with which we

began: Do we have any ethical obligations, as consumers in developed nations, to force transfer only of the safest technologies? We consumers in developed countries may have the greatest power, and perhaps also the greatest obligation, to help solve the problems of transferring hazardous technologies.

Conclusion

This chapter has argued that we have an obligation to "make a difference," to make it difficult for nations and corporations to subject unwitting third-world peoples to transfers of hazardous technology. But

the only clear way that we can "make a difference" is through coordinated political activity, not primarily through individual efforts. We need to put pressure on U.S. agencies such as OPIC. We need to recognize that we have a moral obligation to political activism.

Such recognition has been in short supply in recent years, especially in the United States. We no longer seem to believe that, as Plato argued, each of us is an essentially political animal, concerned with the affairs of the polis or state; or that, as Thucydides argued, all citizens ought to spend time concerning themselves with public affairs as well as with their own personal activities. As Thucydides reminded us, if we succumb to the belief that "someone else" will carry our political responsibilities for us, the common cause will imperceptibly decay.76 We, as citizens and consumers, owe it to the common cause to help prevent many risks of technology transfer.

Methodological Reforms

A naturally occurring decay product of radium 226, radon 222 has an important impact on human health. Because it can leave the soil or rock in which it is present and enter the surrounding air or water, radon gas is ubiquitous. When it decays into a series of radioisotopes collectively referred to as "radon daughters," two of these daughters (polonium 218 and polonium 214) give off alpha particles. These alpha particles, when emitted in the lung, can cause lung cancer.1

Until recently, radon daughter exposure was associated with lung cancer in uranium miners. Now we know that radon daughters are present in the air of buildings, building materials, water, underlying soil, and utility natural gas. Although radon concentrations in American homes have not been systematically surveyed, available data show that some dwellings have concentrations greater than the control levels in underground mines2 This means, for example, that the lung cancer risk for lifetime exposure from age one is approximately 9.1 x 10^-3 assuming an average life span of seventy years, or 70 WLM.3 This is a risk of lung cancer of nearly 1 in 100, as a result of lifetime radon exposure. Despite such risk figures, the present state of scientific knowledge does not allow one to specify the best values characterizing the dose in a home or mine; there are fundamental uncertainties surrounding the radon dosimetry factors.4

Risk assessment and evaluation are beset with typical uncertainties, similar to those in the radon case. Indeed, as was already mentioned, uncertainties of six orders of magnitude "are not unusual.5 As a result of these uncertainties, hazard assessors often fall victim to a variety of methodological errors, such as the probabilistic strategy and the isolationist strategy. In this chapter and the next, I shall outline several proposals for avoiding some of the worst effects of these uncertainties and erroneous methodological strategies. This chapter will examine

several methodological suggestions designed to improve quantitative risk analysis, especially risk evaluation. The last chapter will outline some regulatory and procedural solutions for reforming hazard management. These suggestions should address many of the problems criticized earlier in the volume; they should fill out the new account of rational risk evaluation, scientific proceduralism, begun in Chapter Three.

Policymakers have attempted for some time to improve various methods of risk analysis and evaluation, but their proposed methodological reforms have been fraught with controversy—notably, the debate between the cultural relativists, who overemphasize values in hazard evaluation, and the naive positivists, who underemphasize them.6 Decisionmakers and evaluators likewise disagree over how to resolve some of the more practical problems associated with risk management. At one end of the "solutions spectrum" are environmentalist policy-makers, such as Howard Latin. They argue for uniform environmental standards, for prohibitions against agents suspected of causing harm (carcinogens, for example), and against economic incentives and benefit-cost techniques for reducing risk.7 At the other end of the spectrum are industry-oriented policymakers such as Bernard Cohen and Lester Lave. They argue for situation-specific environmental standards, for negotiation regarding agents suspected of causing harm, and in favor of economic incentives and benefit-cost techniques for reducing risk.8

In this and the next chapter, I shall continue to argue for a middle position, between the industry and the environmentalist solutions to the problems of hazard evaluation and risk management. Agreeing to some extent with the environmentalists, I have argued throughout this book that quantified risk assessment (QRA), risk evaluation, and risk-cost-benefit analysis (RCBA) are seriously flawed, in part because of questionable methodological strategies associated with them.9 However, siding also with the industrial experts, I shall show (in this chapter) that although QRA is in practice deficient, there are in-principle reasons for continuing to use it.

More specifically, this chapter argues for a number of methodological claims regarding risk evaluation: that we need to use QRA and RCBA; that RCBA could be significantly improved by means of ethical weighting techniques; that hazard evaluation ought to be accomplished by means of alternative analyses designed to take account of different methodological, ethical, and social assumptions; that expert opinions on risk estimates ought to be weighted on the basis of their past predictive successes; and that assessors generally ought to give up both the naive positivist and the cultural relativist views, as well as the en-

vironmentalist and the industry-oriented accounts of risk. Instead, the chapter argues that experts ought to adopt a middle position, scientific proceduralism, that emphasizes increasing the analytic precision of hazard assessment and the democratic control of environmental risks.

Admittedly, the conclusion of many persons who have followed my repeated criticisms of methodological strategies associated, with QRA and RCBA might be that we ought to discontinue use of both. Such a conclusion would follow, however, (1) only if the problems with QRA and RCBA were essential to the use of these quantitative and economic methods and (2) only if there were alternative, superior tools capable of achieving better risk analysis and management. Since I believe that both (1) and (2) are false, more modest and reformist conclusions follow from the criticisms leveled earlier in the book. The case for using QRA can probably be made most clearly by considering RCBA, a method described briefly in Chapter Four.10 Since most of the methodological criticisms of QRA in this volume have been directed at the risk-evaluation stage, any plausible defense of the claim that we ought to continue to use QRA ought to focus on this third stage and on the most prominent tool of risk evaluation, RCBA. Moreover, since most people object to QRA because of its quantitative, reductionistic approach, their criticisms typically focus on RCBA. Let's examine RCBA, in order to see why, although these complaints are partially correct, they fail to show that policymakers ought to abandon either RCBA or QRA.

Why We Need QRA, despite Its Flaws: The Case for RCBA

Currently, nearly all U.S. regulatory agencies (with the exception only of the Occupational Health and Safety Administration) routinely use RCBA to help determine their policies. Although the National Environmental Policy Act (NEPA) of 1969 requires that RCBA be used to evaluate all proposed environment-related federal projects, opponents of this technique often view its practitioners as dehumanized numerators engaging in a kind of economic Philistinism.11 Amory Lovins compares RCBA to the street lamp under which the proverbial drunkard searched for his wallet, not because he lost it there but because that was the only place he could see.12 The most common objections to RCBA generally focus on four alleged problems: there is no accepted theory of rationality to undergird RCBA; the democratic process, as well as mathematical-economic techniques, ought to determine risk policy; RCBA ignores factors such as the equity of distribution and the

incommensurability of various parameters; and its data base is inadequate.13

Rather than focus on each of these objections, I want to ask a more basic question: What arguments ought to be counted as decisive in the case for and against RCBA? Unless proponents and opponents of this technique agree on the criteria for resolving their disputes, no consensus on the methods for making public policy seems possible. Currently, much of the debate is at cross-purposes. Both opponents and proponents of this technique are arguing for theses which those who disagree with them do not regard as decisive on the issue of whether or not to use RCBA.

One argument used by opponents of RCBA appears to be both misguided and central to their objections. I would like to expose its flaws and to point out what alternative reasons might be decisive grounds for arguing for or against use of RCBA. This flawed argument is that, since RCBA is deficient in a number of serious ways, it should not be used routinely for societal decisionmaking regarding environmental projects. More descriptively oriented variants of this argument, such as those by Dreyfus, focus on the claim that RCBA cannot model all instances of "human situational understanding."14 More normative variants, such as those by MacIntyre, maintain that RCBA exhibits some of the same defects as "classical utilitarianism."15

Two Main Attacks on RCBA

Although those who attack RCBA do not formulate their arguments in terms of explicit premises, they all appear to be employing a simple, four-step process. For the more normative philosophers, these steps are as follows:

1.     RCBA has a utilitarian structure.

2.     Utilitarianism exhibits serious defects.

3.     Typical applications of RCBA—for example, in risk assessment and in environmental impact analysis—exhibit the same defects.

4.     Just as utilitarianism should be rejected, so should the use of RCBA techniques.16

The more descriptive attacks on RCBA are similar:

1.     RCBA is unable to model all cases of human situational understanding and decisions.

2.     The inability to model all cases of human situational understanding and decisions is a serious defect.

3.     Typical applications of RCBA—for example, in risk assessment and in environmental impact analysis—exhibit the same defect.

4.     Just as faulty models of human situational understanding and decisions should be rejected, so should RCBA.

This four-step argument is significant, not only because persons such as Dreyfus, Gewirth, MacIntyre, and MacLean appear to be employing it, but also because it seems to reflect the thinking of many (if not most) philosophers who discuss RCBA and applied ethics. After taking a closer look at both variants of the argument, I shall attempt to establish two claims:

1.     Even if all the premises of both versions of the argument were true, because the inference from them to the conclusion is not obviously valid, it is not clear that the conclusion (4) ought to be accepted.

2.     Although the second and third premises of the normative variant of the argument are likely true, because the first premise is not obviously true,it is not clear that the conclusion (4) ought to be accepted.

Let's examine first the descriptive variants of the argument. Although its proponents would probably admit that RCBA is useful for some cases of individual decisionmaking (for example, for determining whether to repair one's old car or buy a new one), they claim that, in most instances, individuals do not use RCBA to do a "point count."17 Dreyfus and Tribe maintain, instead, that they use intuition. Socolow and MacLean claim that they employ open discourse, argument, or debate.18 In any case, all critics agree that any formal model such as RCBA is unable to capture what goes on when someone understands something or makes a decision. They maintain that such formal models fail in being too narrow and oversimplified.19

As Dreyfus put it, much policymaking is "beyond the pale of scientific decisionmaking." It requires "wisdom and judgment going beyond factual knowledge," just as chess playing and automobile driving require "expertise" and "human skill acquisition" not amenable to RCBA.20 The expert performer, so the objection goes, plays chess or drives the automobile "without any conscious awareness of the process." Except during moments of breakdown, "he understands, acts, and learns from results," without going through any routine such as RCBA. Indeed, he not only does not go through any such routine; he could not. And he

could not, according to critics of RCBA, because he is often unable to tell a cost from a benefit; much of the time, he doesn't know either the probability or the consequences of certain events.21 Hence, goes the argument, because RCBA cannot model many cases of individual decisionmaking, it cannot model societal decisionmaking.22

More normative policymakers likewise reject use of RCBA for societal risk decisions, but their emphasis is on the claim that RCBA should not, rather than that it cannot, succeed. The gist of their objections to RCBA is that the technique forces one to assume that "everybody has his price." Lovins and MacLean are quick to point out that some things are priceless and hence not amenable to RCBA calculation.23 Gewirth argues that certain rights cannot be costed and then traded for benefits. Critics of RCBA all argue that moral commitments, rights, and basic goods (such as the sacredness of life) are inviolable and incommensurable and hence cannot be bargained away for any benefits revealed in an RCBA.24 In a nutshell, they allege that RCBA shares the same defects as classical utilitarianism; it cannot account for crucial values such as distributive justice.25 Hence, they conclude, in a society where decisionmaking ought to be based on rights and justice, hazard analysis and management ought not to be based on RCBA.

Improvement of RCBA by Ethical Weighting Techniques

Because RCBA is often misinterpreted in purely utilitarian ways, it typically fails to take account of egalitarian values, social obligations, and rights. One way to remedy these deficiencies would be to employ a weighting system for RCBA. Many of the dilemmas of risk evaluation, discussed in Chapter Five, could be ameliorated by a weighting scheme. For example, in cases where free, informed consent to risk imposition was in question (the "consent dilemma"), policymakers could assign a negative weight to those risks. In cases where dangers to hazard victims were likely to combine to some unacceptable level (the "contributors dilemma"), decisionmakers likewise could assign a negative weight to those risks. Ethically weighted RCBA could also be used to counteract some of the risk-analysis strategies criticized in Chapters Six through Ten. For example, if a potentially catastrophic technology imposed a greater risk on consumers, rather than on producers, then its costs could be weighted more negatively.

Admittedly, neither RCBA nor risk assessment can tell us what weights to impose; RCBA is a formal calculus amenable to the weights dictated by any ethical system. Its purpose is to help us clarify our ethical values as a society, not to dictate them. Hence, perhaps the best way to use RCBA would be to have different interest groups prepare alternative RCBAs, each with different ethical weights. The public or its representatives could then decide which weighting scheme best represented its values. I have argued elsewhere that RCBA ought to be weighted by various parameters, but that teams of analysts, including ethicists and members of various public-interest groups, ought to be responsible for devising alternative, ethically weighted RCBAs.47 Once these were completed, normal democratic procedures could be used to choose the desired RCBA. Before I go into that proposal, however, it makes sense to explain why any ethical weighting at all appears desirable.

Because no necessary connection exists between Pareto optimality, the central concept of RCBA, and correct policy,48 it would be helpful if there were some way to avoid the tendency to assume that RCBA alone reveals socially desirable policy. Alternative, ethically weighted RCBAs would enable persons to see that good policy is not only a matter

of economic calculations but also a question of ethical analysis. If people follow Gresham's law and thus persist in their tendency to accord primacy to quantitative results, such as those of RCBA, then using ethically weighted RCBAs might keep them both from ignoring ethical parameters, which are not normally represented in a quantitative analysis, and from identifying un weighted RCBA results with a prescription for desirable social policy.

Ethically weighted RCBAs would also provide a more helpful framework for democratic decisionmaking. Policy analysis would be able to show how the chosen measures of social risks, costs, and benefits might respond to changed value assumptions.49 Likewise, ethically weighted RCBAs might bring values into policy considerations at a very early stage of the process, rather than later, after the RCBA conclusions were completed. In this way, citizens might be able to exercise more direct control over the values to which policy gives indirect assent. Acceptance of the willingness-to-pay measure, for example, implies acceptance of the existing scheme of property rights and income distribution, since risks, costs, and benefits are calculated in terms of existing market prices and conditions.50 To employ a system of alternative, ethically weighted RCBAs, among which policymakers and the public can decide, would be to assent to the theses (1) that existing RCBAs already contain ethical weights (which are probably unrecognized) and (2) that proponents of the ethics implicit in current analyses ought to be required to plead their cases, along with advocates of other ethical systems, in the public court of reason. Given that weighting is already done anyway, both because of the implicit presuppositions of economists who practice RCBA and because of the political process by means of which the public responds to RCBA, it makes sense to bring as much clarity, ethical precision, openness, and objectivity to it as possible, by explicitly weighting the RCBA parameters in a variety of ways.51

Another reason for using ethically weighted RCBAs is that it appears more desirable than alternatives such as the use of revealed preferences or the current market assignments for income distribution, risks, costs, and benefits.52 Employing revealed preferences requires one to make a number of implausible assumptions, as Chapter Four argued.53 These could be avoided if, instead of employing revealed preferences, one followed the practice of preparing a number of alternative, ethically weighted RCBAs. Following Mill's lead,54 several economists have also proposed that RCBA can take account of the nonutilitarian distinction between right and expediency by using a system of weights.55 Economists Allen Kneese, Shaul Ben-David, and William Schulze, for example, working under National Science Foundation funding, have developed

a scheme for weighting risks, costs, and benefits by means of alternative ethical criteria.56 They also suggest that each ethical system be represented by a general criterion rather than by a list of rules, such as the Ten Commandments.57 They claim that the requirement that an ethical system be represented as a transitive criterion for individual or social behavior leaves at least four ethical systems (and probably more) for use in reweighting RCBA parameters: Benthamite utilitarianism, Rawlsian egalitarianism, Nietzschean elitism, and Paretian libertarianism. On their scheme, the Benthamite criterion is that one ought to maximize the sum of the cardinal utilities of all individuals in a society. The Rawlsian criterion is that one ought to try to maximize the utility of the individual with the minimum utility, so long as that individual remains worst off. According to the Nietzschean weighting scheme, one ought to maximize the utility of the individual who can attain the greatest utility. Finally, according to the Paretian criterion, say Kneese, Ben-David, and Schulze, one ought to act in such a way that no one is harmed and, if possible, that the welfare of some persons is improved.58

The merit of the ethical weighting criteria described by Kneese and his associates is that they could theoretically provide a unique sort of nonmarket information. For example, one could ask an individual how much he would be willing to pay for redistributing income to less fortunate members of society. The purpose of asking about such ethical beliefs, of course, would not be to provide a prescription for policy, but to allow the public and policymakers to see how different assumptions about the desirability of given distributions change the overall ratio of costs to benefits. They would be able to examine alternative, ethically weighted options, and not merely marginal costs or benefits across opportunities.

The work of Kneese and his associates illustrates dramatically that what is said to be feasible or unfeasible, in terms of RCBA, can change dramatically when different ethical weighting criteria are employed. Using case studies on helium storage, nuclear fission, and automobile-emission standards, they showed how alternative ethical weights can be used to generate contradictory RCBA conclusions. For example, when used with a Benthamite weighting criterion, RCBA reveals that using nuclear fission for generation of electricity is not feasible. When either the Paretian, Rawlsian, or Nietzschean criterion is used, however, nuclear power may be said to be feasible or unfeasible, depending on the value attributed to factors such as compensation for damages and the utility attributed to future generations.59

Admittedly, the weighting schemes outlined by Kneese and his colleagues have a number of limitations. Most notably, because they em-

ploy simple criteria, they fail to capture the complexity of ethical systems. They are allegedly unable, for example, to represent a priority ordering of different ethical claims within the same ethical system.60 Since I have argued elsewhere that one can order the ethical claims of individuals and societies, and then use them to weight RCBA parameters, I shall not repeat those arguments here.61 The main objection to them is that any sort of weighting scheme would be "at variance with the allocative principles by which the competitive economy is vindicated."62

However, government often makes policy decisions inconsistent with certain presuppositions underlying the competitive economy. Whenever government takes account of an externality such as pollution and imposes restrictive taxes or outright prohibitions, it is clearly making decisions inconsistent with particular presuppositions underlying a purely competitive economy. Indeed, if it did not, then grave harm, such as pollution-induced deaths, could occur. Also, it is well known that economists "correct" their (competitive) market parameters for risks, costs, and benefits.

Another classic objection to the use of any weighting scheme in RCBA is that weightings ought to be left to politicians and the democratic process and not taken over by economists.63 On the contrary, although economists and ethicists might help to formulate lexicographic rules and to weight RCBA parameters in accordance with these rules,64 they need only be responsible for helping to formulate a number of alternative, ethically weighted analyses. The public and policy-makers would decide among the alternative RCBAs. Indeed, one reason why this proposed weighting scheme would not lend itself to control by experts, who have no business dictating policy in a democracy, is that it calls for preparation of alternative RCBAs, each with different ethical weights.65 This amounts to employing an adversary means of policy analysis (see the next chapter).

Use of Alternative Risk Analyses and Evaluations

Weighting QRAs and RCBAs so as to reflect different methodological, ethical, and social preferences, however, is valuable only to the degree that one is able to see how alternative evaluative assumptions generate different risk-assessment conclusions. In other words, weighting schemes are valuable as sensitivity analyses. They make it possible to observe the effects of different assumptions on the same hazard assessment. Therefore, several risk analyses ought to be done for any single

societal or environmental threat, and each of these analyses should contain different methodological assumptions.

One reason for mandating that several alternative hazard assessments be done is that successful decisionmaking depends in part on knowing all the relevant facts and seeing all sides of a given "story." It is more likely that all sides to a story will be revealed if different groups conduct hazard analyses.66 Moreover, the public deserves a role in the process of determining rational risk choices, if the locus of public decisionmaking ought to be with the people, rather than merely with scientific experts. Public control and consumer consent can hardly be obtained if only one risk evaluation, controlled merely by scientific experts, is performed. Also, because all risks are both perceived and value laden,67 and because all hazard evaluations employ judgmental strategies, some of which are highly questionable,68 there are no value-free risk assessments. Consequently, hazard analysis, especially risk evaluation, is highly politicized and therefore should be accomplished in a political and legal arena, so that all persons affected by a given risk will receive equal consideration.

Only if the naive positivists were correct in their views about risk would it make sense to perform only one hazard evaluation.69 Since they are not correct, assessors ought to perform alternative evaluations, so as to spell out some of the controversial social and political dimensions of risk policy.70 In the account that I have been defending, increasing the degree of analytic sophistication is not sufficient for resolving risk conflicts. Policymakers must rely on procedural and democratic, rather than merely scientific, methods of evaluating and managing risk.71

Purely scientific methods of risk evaluation are inadequate because there are numerous uncertainties in hazard assessment. Scientists generally are unable to evaluate long-term environmental changes, in part because of the complexity of ecological science.72 Likewise, epidemiological aspects of risk analysis are often problematic owing to factors such as lack of exposure data, small sample size, recall bias, chance variation, long latency periods, and control selection. Exposure assessments are often uncertain because of synergistic effects, reliance on average doses, and dependence on particular mathematical models that oversimplify the actual situation. Dose-response assessment is uncertain because of the unreliability of particular experimental conditions, the difficulty of estimating whether there is a threshold, incorporation of background hazard rates, and extrapolation from observed doses.73 Because of these uncertainties, assessors fill gaps in information with inference and judgment. Because they do so, it is important to perform

several different analyses of the same risk.74 Moreover, because the various institutions performing assessments often seize upon whatever data and value judgments aid their cause,75 it is not wise to rely only on one study. In fact, a variety of government policy groups has already recognized the importance of doing alternative assessments.76

Use of Weighted Expert Opinions

Another methodological device for improving hazard evaluation is to weight expert opinions. This procedure amounts to giving more credence to experts whose risk estimates (probabilities) have been vindicated by past predictive success. It is a way to exercise probabilistic control over expert opinions and to make them more objective. Weighting such opinions would reveal whether a hazard assessor (who provides a subjective probability for some accident or failure rate) is "well calibrated." (A subjective probability assessor can be said to be well calibrated if, for every probability value r in the class of all events to which the assessor assigns subjective probability r , the relative frequency with which these events occur is equal to r. 77 ) The primary justification for checking the "calibration" of risk assessors is that use of scientific methodology requires testing problem solutions. Not to test them is to reduce science to ideology or metaphysics.78

Admittedly, obtaining empirical control of subjective probability assessments is very difficult in the practice of risk analysis. The very reason for resorting to the use of such subjective estimates, based on degrees of belief of experts, is that actual accident frequencies, for example, cannot be determined in any uncontroversially objective way. As was already mentioned, the accidents for which there are no objective probability estimates are typically those involving new technologies. Recognizing this, risk assessors have used subjective probabilities, based on expert opinion, since about 1975. Decision theorists have used them since the 1950s.

It is difficult both to arrive at subjective probabilities and to calibrate them because the numerical values in question are often very small; for example, the per-reactor-year (subjective) probability of a nuclear core melt, as was mentioned earlier in the volume, is 1 in 17,000.79 For such subjective probabilities, calibration is ruled out; it would take too many years to test the frequency of actual core melts. However, there are many subjective probabilities that go into the calculation of the likelihood of a core melt, and weighting expert opinions in risk assessment may be possible if we are able to use these other, lower-level

subjective probabilities. It is also extremely important to do so, because there is a wide divergence of opinion among experts as to their actual values. In the famous WASH-1400 study of nuclear reactor safety, for example, thirty experts were asked to estimate failure probabilities for sixty components. These estimates included, for instance, the rupture probability of a high-quality steel pipe of diameter greater than three inches, per section-hour. The average spread over the sixty components was 167,820. (The spread of these thirty expert opinions, for a given component, is the ratio of the largest to the smallest estimate.) In the same study, another disturbing fact was that the probability estimates of the thirty experts were not independent; if an expert was a pessimist with respect to one component, he tended to be a pessimist with respect to other components as well.80

Recently, a group of hazard assessors in the Netherlands used empirical frequencies obtained from a study done by Oak Ridge National Laboratories to calibrate some of the more testable subjective probabilities used in WASH-1400, one of the best and most renowned risk assessments ever accomplished.81 Obtained as part of an evaluation of operating experience at nuclear installations, the frequencies were of various types of mishaps involving nuclear reactor subsystems. The Oak Ridge study used operating experience to determine the failure probabilities for seven such subsystems (including loss-of-coolant accidents, auxiliary feedwater-system failures, high-pressure injection failures, long-term core-cooling failures, and automatic depressurization-system failures for both pressurized and boiling water reactors). Amazingly, all the values from operating experience fell outside the 90 percent confidence bands in the WASH-1400 study. However, there is only a subjective probability of 10 percent that the true value should fall outside these bands. Therefore, if the authors' subjective probabilities were well calibrated, we should expect that approximately 10 percent of the true values would lie outside their respective bands. The fact that all the quantities fall outside these bands means that WASH-1400, allegedly the best risk assessment, is very poorly calibrated. It also exhibits a number of flaws, including an overconfidence bias.82

One can calibrate subjective probabilities and thus correct, in part, for these biases in probability estimates.83 There is extensive psychometric literature on calibrating probability estimates,84 as well as long-term meteorological experience with calibration techniques. National Weather Service forecasters have been expressing their predictions in probabilistic form since 1965, and numerous analysts have collected and evaluated weather data for predictive accuracy.85 Nearly all these

analyses show excellent calibration. Even more important, calibration has improved weather forecasts since the probabilistic forecasts were introduced.86

In response to these arguments for calibrating probability assessors, at least two objections are likely. One is that calibration does not guarantee reliable risk decisions. Admittedly not, but this is not an insurmountable problem, since no process can ensure that it will lead to the right decisions. We cannot command results (since that would require having a crystal ball), but we can command methods. Hence, our responsiblities in decisionmaking must focus primarily on methods, not on results.87

Another objection to calibrating experts is that there is no firm relationship between assessors' performance on the smaller probability estimates, used for checking calibration, and on the larger ones. In probability estimates of a nuclear core melt in particular, there are no good checks or controls for the larger estimates. This objection fails, however, because policymakers ought to try to calibrate expert opinion, in part because it is apparently so often wrong, in part because meteorologists have been successful in doing so, and in part because the consequences of failure to calibrate could be catastropic (e.g., 145,000 people killed in a nuclear core melt).88

Given these reasons to calibrate, the key question is how to go about doing it in the best way. The best calibration data we have are the many observed frequencies whose calculated probabilities together comprise larger probabilities (such as those for core melt). Although these lesser figures are not exactly the same as the core-melt probability, they are a necessary part of obtaining the larger probabilities and are hence better than no data. To argue against using a second-best check on expert opinion, merely because it is not as good as the best check (knowing the actual frequency), is unrealistic, since knowledge of frequencies is in principle unavailable. If it were available, one would not be relying on expert opinion in the first place.

How Scientific Proceduralism Guarantees Objectivity

All these methodological suggestions (calibration, alternative assessments, ethical weights) for improving risk evaluation and hazard management are predicated on two principles, both defended earlier in this volume. One principle is that assessors ought to give up both the rigid, naive-positivist assumption that experts' risk estimates are completely value free and the erroneous relativist assumption that risk assessment is not objective in any sense. It is objective, in at least the

three senses discussed in Chapter Three. The second principle is that contemporary hazard evaluation needs to become more democratic, more open to control by the public, and more responsive to procedural accounts of rational risk behavior.

We must now show how to safeguard scientific rationality and objectivity, even though risk-evaluation methods need to take account of democratic, ethical, political, and procedural factors, factors allegedly not capable of being handled in purely rational and narrowly objective ways. The procedural account of hazard evaluation (outlined in Chapter Three) presupposes that rationality and objectivity ultimately require an appeal to particular cases as similar to other cases believed to be correct, just as legal reasoning requires. Aristotle recognized that there are no explicit rules for such judgments, but that inexplicit ones guide moral reasoning. These inexplicit rules or judgments rely on the ability of a group of people, similarly brought up, to see certain cases as like others. This recognition of cases is also what Wittgensteinians are disposed to believe about all instances of human learning. At the final level of risk evaluation, after subjecting the assessment to the tests of predictive and explanatory power and to the rigors of debate over alternative assessments, there can be no specific algorithms to safeguard objectivity, no infinite regress of rules. Ultimately, even rules must give way, not to further appeals to specific (risk-evaluation or risk-assessment) rules, as the naive positivists presuppose, but to a shared appreciation of similarities.89

As such, this Popperian and Wittgensteinian account (scientific proceduralism) anchors objectivity to (1) criticisms made by the scientific and lay community likely to be affected by risk judgments and (2) empirical control over expert opinion (obtained by pursuing the goal of explanatory power, tested by prediction; by calibrating probability assessors; and by performing sensitivity analyses). The criticisms would help to protect the procedural and democratic aspects of risk evaluation, and the empirical control would help to safeguard its predictive and scientific components—that is, its rationality and objectivity.

This account of scientific objectivity is premised on the assumption that open, critical, and methodologically pluralistic approaches to hazard analysis and evaluation (via alternative studies, sensitivity analyses, calibration, and ethical and methodological weighting schemes) can in principle reveal the theoretical, linguistic, and cultural invariants of reality, much as a plurality of experimental perspectives reveals the true invariants of quantum mechanical systems.90 In the view that I am suggesting, the relevant variance principles applicable to hazard analysis, and especially evaluation, dictate that risk behavior is rational and

objective if it survives scrutiny and criticism by different, well-calibrated communities of theory holders, each with different transformations or ethical assessments of the same hazard.91

In this view, the risk assessments that we ought to call "rational" are those that have been subjected to systematic critical inquiry regarding explanatory and predictive power, as Popper proposed for science. Risk evaluations that we ought to call "rational" are likewise those that have been subjected to systematic, democratic, and procedural constraints.92 But systematic, critical inquiry requires a plurality of techniques: sensitivity analyses, calibration, alternative assessments, and so on. According to the epistemology outlined here, what characterizes science-related activities, such as hazard analysis, is that they are susceptible to empirical control. They are evaluated, in part, on the basis of explanatory power and predictive successes. The salient features of science and related activities such as risk assessment are thus empirical and predictive control, criticism of one's own perspectives, and recognition that there are alternative perspectives and methods. Indeed, if objectivity has anything to do with invariance under different theoretical perspectives,93 then scientists and risk assessors ought to be strongly committed to retaining only those beliefs that survive critical scrutiny. Such a notion of objectivity, however, suggests that the ultimate requirement for improved risk assessment is a whole new theory of rationality, one that is critical, procedural, populist, egalitarian, and democratic, as well as objective, scientific, and controlled in part by prediction and calibration. It requires, in other words, an epistemology in which what we ought to believe about risk analysis is bootstrapped onto how we ought to act. 94 For example, we ought to act in ways that recognize due-process rights, give equal consideration to the interests of all persons, and so on.

Conclusion

Since this chapter, like the rest of the volume, is primarily philosophical, it is not meant to provide a precise account of the methodological solutions to problems of risk assessment and evaluation. These details need to be given by statisticians, epidemiologists, psychometricians, economists, and risk assessors. Precise methodological techniques (e.g., for calibration) need to be spelled out, and more examples and more developed arguments need to be provided, in order to support responses to these and Other objections to the position of scientific proceduralism. The argument sketches given thus far, however, suggest some of the ways that we might attempt to improve our methods of risk assessment and evaluation, so as to address the problems outlined earlier in this volume. The argument sketches also show that the notions of scientific rationality and objectivity outlined here deserve further investigation; indeed, much of the epistemological work for this account of hazard evaluation remains to be done. Let's move to the next chapter to examine some of the ways in which we might improve risk management.

Procedural Reforms

During any period in almost any region of the country, one hears reports of local opposition, either to siting unwanted technological facilities or to bearing the environmental risks associated with already existing developments. Names like "Love Canal," "Three Mile Island," "Baltimore Canyon," "Minimata Bay," and "Seabrook" have been etched in our newspapers, our courtrooms, and our fears.

According to some experts, many cases of public opposition to risk are created by self-serving attitudes and unrealistic public expectations, rather than by catastrophic accidents and accelerating cancer rates. These other controversies, they say, represent instances of an "inverse" tragedy of the commons. In Garrett Hardin's "tragedy of the commons," people misuse a "public good," such as clean water, simply because it is free.1 The "inverse tragedy of the commons," however, occurs when people avoid their share of responsibility for "public bads" that fulfill essential public purposes—for example, when people try to avoid having airports, roads, or toxic waste facilities near them.2 One of the main tasks of hazard assessment and management is to know when a risk imposition represents a tragedy of the commons, which ought to be avoided, and when it is an inverse tragedy of the commons, which ought to be accepted responsibly.

One reason why it is difficult for the public to determine whether a given risk ought to be accepted or rejected is that evaluators often do not make an adequate case for accepting the risks that are likely to be imposed. They do not do so, in large part, because their hazard analyses often have two main deficiencies. They typically employ a number of questionable evaluative assumptions,3 epistemological strategies,4 whose effect on policy is to disenfranchise the public, the potential risk victims.

This chapter will provide a brief overview of several procedural ways to improve hazard evaluation and risk management. It will argue that, in order to address some of the problems with equity of risk distribution, compensation, consent, and scientific uncertainty, we need to investigate ways to reform statutes dealing with societal hazards. Moreover, because of the deepening loss of public trust in most institutions, carefully structured citizen participation and negotiation in making risk decisions are the only ways to legitimate them to the public. Policy decisions, especially when they affect health and safety, stand little chance of being accepted, for example, if government and industry continue their current DAD (decide, announce, defend) strategy for siting hazardous facilities and imposing environmental risks. Such a procedure relies on an unstructured, industry-controlled siting process and limited citizen input; the result is that public controversy is virtually guaranteed.5

This chapter will not provide specific solutions to problems of managing environmental risks, since these and the arguments for them are best accomplished by specialists in fields such as environmental law and welfare economics. Nevertheless, the general proposals provided in subsequent sections of this chapter should be enough to establish the prima facie plausibility of several proposals: (1) to achieve statutory reform in areas such as toxic torts; (2) to aim at obtaining free, informed consent from all potential risk victims and to include them in decisionmaking about societal hazards; (3) to provide ex ante and ex post compensation for imposed public risks; (4) to negate all liability limits that protect risky technologies at the expense of the public; and (5) to begin to resolve environmental controversies through negotiation and adversary assessment. Since the distribution of public risks has rarely been fair and has often fallen disproportionately on the powerless and the poor,6 the proposals aim at a more equitable distribution both of societal hazards and of decisionmaking power regarding them.

Statutory Reform and Risk Management

One of the central ways of reforming risk management is to amend the laws governing environmental or technological hazards, such as the laws governing toxic torts. Statutory improvements in this area are difficult to achieve, as Chapters Four, Five, and Eleven outlined, because there are numerous scientific uncertainties in identifying, estimating, and evaluating societal risks. As a result, litigation related to

public safety and health is complex, time consuming, and expensive; toxic tort plaintiffs, for example, must explain how and why a hazard caused a specific condition, even though any of numerous factors could be the real cause. Moreover, the relative contribution of genetic makeup, life-style, and the external environment to human health is difficult to determine. As a consequence, industry and government repeatedly invoke scientific uncertainty as grounds for the failure to reform laws dealing with hazardous substances.7

The reason why these legal mechanisms need to be amended is clear if one considers the costs that environmental hazards force on citizens. Public risks typically impose at least three burdens on society: costs of harm (e.g., medical bills, pain); costs of avoiding harm (e.g., pollution-control mechanisms); and transaction costs incurred in allocating harm (e.g., litigation, negotiation, and regulation). If there were no transaction costs, then society could allocate resources so as to minimize the sum of the costs of harm and avoiding harm. In such a situation, the party imposing risk or damage would have to strike a bargain with the (potential) victims, a bargain leading to an economically efficient outcome.8 Since our society has transaction costs, however, its laws of liability often lead to economically inefficient outcomes. Also, they favor those who impose risks, rather than those who are their victims, because those who cause risk and harm are not penalized until the damage to victims exceeds their (the victims') transaction costs (e.g., exceeds the costs of the victims' initiating litigation). The net effect is that high transaction costs (often caused by scientific uncertainty) beset attempts to establish liability; trying to prove that someone who imposes a risk (e.g., a toxic dump licensee) is at fault may thus encourage inefficient and unethical allocations of resources.9 The purpose of this chapter is to examine some efficient and ethical ways of managing societal risks.

To understand the typical difficulties of using the existing legal system to protect victims of environmental risks, we can consider several characteristics of toxic tort litigation. (1) Often victims do not know how or that they have been put at risk or harmed by toxic substances. (2) The time, effort, and expense of bringing suit often exceed the probable damage award, as is evidenced by the small number of compensation awards in cases of occupational cancer. (3) Toxics victims typically have less information than the defendants, usually companies producing hazardous chemicals. These difficulties are illustrated by the fact that, even outside the tort system, workers' compensation boards are typically unable to determine causes of harm. The boards face the same difficulties as plaintiffs within the legal system when they attempt

to establish causes of harm despite limited information. In the state of Colorado, for example, less than two plaintiffs per year typically receive workers' compensation awards for occupational cancer, even though epidemiologists estimate that many times this number deserve such benefits.10

Private damage actions for victims of environmental hazards are also problematic because many states have statutes of limitations, and most courts require victims to prove that the defendant's behavior unreasonably exposed them to a hazardous substance. Most courts also require victims to quantify costs and benefits in attempting to establish fault, to show causation of harm, and to prove that it was "worth" avoiding in a cost-benefit sense. Proving causation is especially difficult, as was noted in Chapter Four, because studies of environmental harm address causality in terms of statistical probabilities, a form of evidence typically not recognized by courts as establishing causation. Even if victims can establish the agent causing their harm, they are still faced with showing which of many exposures or risks are responsible for the actual harm— for example, whether they were exposed to chlorine in the drinking water or in the workplace.11

Regulation by administrative agencies, such as the Environmental Protection Agency or the Nuclear Regulatory Commission, likewise often fails to protect victims of environmental hazards, just as do private damage actions. This is in part because, in most such cases, the risk perpetrator is considered innocent until proved guilty. Hence, even though the government possesses resources enabling it to overcome many scientific uncertainties associated with environmental risks and harms, well-financed business interests often dominate regulatory decisions. The Atomic Energy Commission, for example, was so embroiled in lawsuits, because of its catering to vested interests, that it had to be abolished in 1974.12

Another difficulty with regulatory agencies is that statutory control is often fragmented among several different commissions. For example, five different federal agencies are responsible for the regulation of toxic chemicals. This duplication of effort creates confusion, inefficiency, and inconsistency. Regulatory agencies also typically have insufficient funds to bring lawsuits in more than 1 or 2 percent of the cases that ought to be tried. Since current regulations do not encourage private-sector research on avoiding unknown environmental dangers, only the government is able to determine the responsibilities of those who impose risks on the public. Moreover, rulemaking is time consuming and cumbersome, so that, even if something is known to cause a risk, it takes more than a century to establish regulatory standards for it.13

Informed Consent and Citizen Negotiation

To some extent, protecting those who are most vulnerable to environmental risks (e.g., those who work with toxic chemicals) is a matter of guaranteeing free, informed consent to public and occupational hazards. Statutory reform, as suggested in the previous section, would provide some of these guarantees, since the new statute would include provisions for limiting liability as an incentive for full disclosure of possible hazards.37 To achieve explicit consent, however, we need actual citizen participation in negotiating solutions for problems of risk. Moreover, once one realizes that the process of hazard assessment and management is highly value laden and politicized,38 then negotiation (rather than mere expert decisionmaking) becomes a virtual necessity for ensuring free, informed consent in situations of controversial risk.

If the purpose of negotiation is to defuse highly politicized risk situations, so as to ensure citizens' free, informed consent, such negotiation will need to presuppose that several conditions have been met. First, ideally the bargaining parties ought to be roughly equal in political and economic power, in order to ensure free, procedurally just transactions. This means that both citizens and industry groups

need equal funding so as to obtain access to experts, attorneys, and staff assistance. In particular, citizens need to have access to taxpayer monies to fund their negotiations, so that they will not be disadvantaged in any way in bargaining with industry. Second, alternative points of view, different evaluative assumptions, and a variety of risk methodologies ought to be considered. Government would fund the completion of alternative assessments and hazard-management plans,39 ensuring that all sides to a controversy were represented and well informed. Consideration of alternative positions would be a requirement of democratic decisionmaking, rather than a luxury accessible only to those who are financially able to participate in administrative hearings or legal appeals. Third, the negotiation process would be controlled not by a regulatory agency with discretionary powers, but by a group of citizens and experts with no conflict of interest in the matter under consideration. Hence, unlike regulatory and administrative decision-making, the negotiation would be less likely to be co-opted by unrealistic environmentalists or by unscrupulous industry representives.

On the ethical side, negotiation is often able to equalize the costs and benefits associated with hazards, even though risk imposition usually results in dissociation of costs and benefits. For example, one group of citizens might receive the benefits associated with using toxic chemicals, while another set of persons (those who live near the manufacturing plant) might bear the costs associated with the hazard. Those affected by risky technologies or environmental actions realize that, without compensation or consent, it is not fair for them to bear the dangers of some activity that benefits the whole society.40 Economists recognize that such inequities cause a misallocation of resources and a misrepresentation of policy choices.41 Typically, however, the social costs of this inequitable risk distribution are not taken into account in decisionmaking. Most economists generally consider these social costs or inequities only as "externalities," external to the market processes with which economics is concerned.

As was argued in Chapter Eleven, however, justice demands that we internalize these social costs, perhaps by using risk-cost-benefit analysis that is weighted to reflect considerations of fairness and distributive equity. One way to assess which of a variety of weighting schemes and alternative hazard analyses to employ is to use negotiation among citizens, the relevant industry officials, and policymakers. Since it enables local citizens and potential victims to have a voice in hazard assessment and management, negotiation can be an important means of avoiding antipopulist biases (Chapter Two), reductionist approaches to risk (Chapter Three), as well as the expert-judgment (Chapter Six),

probabilistic (Chapter Seven), and producer (Chapter Nine) strategies criticized earlier.

On the practical side, negotiation is one way of avoiding costly, time-consuming litigation, as well as local opposition to environmental risks. Many observers are convinced that a hazardous facility, for example, cannot be sited over a community's objections. A local group has many tactics that enable it to get its way: delay, litigation, political pressure, legislative exemption, and gubernatorial override.42 Individuals and communities have a keen sense of the inequities that result when one group bears the risks and another reaps the benefits. Consequently, they often subscribe to the NIMBY ("not in my backyard") syndrome and, at least in the siting of toxic waste dumps, have usually succeeded in averting the hazard.43

One practical way of avoiding the NIMBY syndrome is to negotiate with citizens, in order to determine what actions, if any, might be required to mitigate hazards, to promote safety, or to compensate them for accepting a given societal risk.44 Project siting generally fails when the government agency attempts to "sell" its premade decision, when it redefines public concerns, when it sponsors no prior public education efforts, and when it merely holds citizen hearings. It is often successful when the agency seeks data on the public's attitudes and needs, when it holds small-group meetings, when it sponsors prior education efforts, and when it exchanges written information with citizens or environmental groups.45

Admittedly, proponents of a risky technology might try to avoid negotiation with the public on the grounds that it is costly and time consuming. They might prefer instead to have state governments preempt local decisionmaking. Using legal and regulatory devices to avoid negotiation, impact mitigation, and compensation, however, is ultimately not a practical way to move toward siting a hazardous facility or imposing a particular risk on a community. It is impractical because, apart from what the law and the courts say, opponents of an environmental risk (unless they are won over through mitigation and negotiation) can resort to civil disobedience to accomplish their goals. In Michigan, for example, as was already mentioned, local residents put nails and tacks on the highways in order to prevent the state from burying cattle contaminated by polybrominated biphenyls. In other jurisdictions, residents have threatened to dynamite existing risky facilities; they have also taken public officials hostage to vent their anger over policymaking processes. All this suggests that, apart from the ethical and legal questions involved, people will resort to extreme measures to avoid a risk if they believe that their lives or their homes are threat-

ened. In such a situation, only negotiation, and neither law nor force, stands a chance of winning them over.46

Since there are a number of different models of risk participation and negotiation within local communities, I shall attempt neither to defend a particular one here nor to answer various questions that might arise in a given model (e.g., who should negotiate and how they should be chosen).47 Those tasks are better left to policymakers, arbitrators, and sociologists.48 To illustrate the prima facie plausibility of negotiation, I need only sketch several cases in which it has been used successfully. Environmental sociologist Elizabeth Peelle, a member of the research staff of Oak Ridge National Laboratory, has done much of the pioneer work on successful citizen negotiation in the face of risky technologies and undesirable environmental impacts. (Both she and I count citizen participation as "successful" if it leads to risk decisions acceptable to the informed, affected public.)

By analyzing some of the successful instances of public participation, Peelle was able to determine their common characteristics and hence to suggest criteria for successful negotiation regarding hazards. In one recent essay, she examined a local citizens' task force for a monitored retrievable storage (MRS) facility for nuclear waste in Tennessee; a two-state citizen forum for a Hanford, Washington, nuclear waste project; two local citizens' task forces for toxic waste in North Carolina; and local and state participation in a hazardous waste project in New Mexico. Her conclusion was that, when citizens participate in risk assessment and management and are involved in the decision process, they are more likely to accept the ultimate actions taken in such projects. When decisionmaking includes negotiation, say Peelle and numerous other experts, the possibilities increase for legitimated, stable public policy.49

In another essay, Peelle detailed three impact-mitigation plans, two for nuclear power plants in Tennessee and Washington State and one for a coal-burning facility in Wyoming.50 Faced with opposition from the local county (which refused to rezone the site) and from the Sierra Club and the Farm Bureau (which contested the application for a license to generate electricity), the three utilities (involved in each case) negotiated with citizens on how to promote safety, how to reduce hazards, and how to compensate them for their potential risks and losses.51 The Wyoming plan, the most comprehensive of the three discussed by Peelle, provided compensation for the risk by guaranteeing the community funds for mental health and social services, recreation, roads, and law enforcement. All three of the risk-abatement plans included some hazard-monitoring provisions.52 Their annual costs, for each utility, are between $125,000 and $900,000. In each case, the utility gives

direct payments to the community, either in the form of prepayments of future taxes or by means of loan guarantees and grants.53 Because they internalize (through compensation) many of the social costs associated with the imposition of societal risks, and because they are the result of negotiation with the relevant local communities, these impact plans have had at least three desirable effects. They have lessened uncertainties concerning facility siting; they have reduced delays re-suiting from unresolved issues, such as safety; and, through monitoring impacts, they have provided an urgently needed data base for improving future impact predictions and risk-mitigation plans. The three cases also illustrate a difference in risk-mitigation approaches: the Tennessee and Wyoming plans offer compensation for undesirable impacts, such as inequitable risk distributions,54 whereas the Washington plan provides fiscal incentives for acceptance of a risky technology.55

Perhaps the best argument for use of incentives and compensation is that offering them to citizens within the context of negotiation is far superior to most existing means of deciding whether to impose environmental risks on communities. Public participation within these existing means of decisionmaking often takes the form of intervention; that is, environmentalists and consumer activists appear, speak, and hence "intervene" on behalf of their cause, before the agency (for instance, the Nuclear Regulatory Commission) that has the power to impose a risk on a community. The process is fatally flawed, however, both because intervenor status is hard to obtain and because, once granted, the status does not guarantee adequate funds for representing the point of view of citizens or environmentalists. Moreover, intervenors are typically limited to raising only those issues which the law already requires the agency to take into account, even though other questions may be far more important. In the siting of nuclear power plants, for example, intervenors are not allowed to challenge the adequacy of current standards for radiological protection, but only to question whether a given utility or plant will meet those standards.56 Consequently, much of the intervenor money that is raised must be used on expensive lawyers who know how to fight highly stylized battles. All this is to little purpose if intervenors cannot challenge the decision-making process itself, a procedure that typically disenfranchises the public. Furthermore, decisions to site risky facilities are often a fait accompli by the time an intervenor has his say. Hence, intervenors are often viewed more as obstructionists than as representatives of any reasonable point of view. Use of incentives within a framework of negotiation, however, provides a possible opportunity for consensus building rather than mere obstructionism.57

For incentives to be successful in increasing public support for risky

environmental projects, social scientists have determined that at least three prerequisites must be satisfied: (1) The incentives must guarantee that local criteria for public health and safety will be met; (2) they must provide some measure of local control; and (3) they must legitimate negotiations as a viable mechanism for building consensus and resolving disputes.58

Of all three prerequisites, the first is probably the most important. Some of the most successful negotiations have taken place only when developers and manufacturers of risky technologies adopted a comprehensive, integrated program of risk reduction. For example, the New Jersey Hazardous Waste Facilities Siting Commission has been successful because it has sought not merely to obtain financial compensation for local citizens but also to reduce the pollutant stream, recycle waste, and detoxify it.59

The theory behind such negotiation and risk reduction is that, through the use of neutral or self-interested negotiators, all interested parties can come together, identify areas of common interest, and reach a solution that is fair and mutually acceptable. Besides New Jersey, Massachusetts, Rhode Island, and Wisconsin are also pioneers in negotiating to resolve local risk disputes. All three states preempt local land-use controls, but then throw the owner and operator of the proposed or existing hazardous facility into a procedure that combines methods of mediation, negotiation, and arbitration. (Mediation or conciliation, as defined by many policymakers, is a subset of negotiation; an approach adapted from labor relations, it uses negotiation to identify the real cooperative actions possible for interdependent parties whose interests and objectives differ. Negotiation typically involves two parties, whereas mediation is handled by a third party seeking to resolve a two-party dispute; neither procedure is ordinarily binding. In arbitration, however, two parties sit before a judge, and the outcome is binding.60 ) Some of the elements recognized by social scientists as necessary for successful negotiation include the following: identifying all the relevant parties early in the process, involving affected persons in early information development and evaluation, providing incentives to residents to accept a risk, taking steps to see that all citizens are informed about the proposed action or facility, and ensuring that the benefits exceed the costs for the host communities.61

Objections to Negotiation

Admittedly, there are a number of objections to negotiation, both on the part of the affected communities and on the side of the developers or industries seeking to gain risk acceptance from a local group. The

industry objections typically have to do with the political or economic costs likely to be required because of negotiation. These objections can often be answered by some of the practical considerations advanced in the previous section.62 A common citizen objection to negotiation is that it allows situations in which persons trade health and safety for financial rewards. Citizens claim that negotiation either allows developers and risk imposers to "bribe" local groups, or else it condones communities that want to "extort" compensation from developers.63

It is important to put these deficiencies in perspective, however, in order to see why negotiation is needed. Negotiation has all the same flaws that democracy generally has: when citizens are allowed to make choices, they often make mistakes. The only way to guarantee that citizens will never make mistakes is never to allow them to make decisions. Admittedly, neither negotiation nor democracy is a sufficient condition for avoiding mistakes such as trading lives for dollars. Negotiation with potential victims is needed, however, because it is a necessary condition for avoiding error in hazard evaluation and management; only those affected have the right to consent to the imposition of risk.

The argument throughout this chapter and the preceding ones has not been that negotiation has no problems. It has been a much more modest claim: that negotiation (in which citizens have a voice) is preferable to hazard assessment and risk management in which they have no role. The argument here is that negotiation is preferable to the typical current procedure for siting facilities, a strategy that often disenfranchises the public. Negotiation is preferable to preemption of local control, both because citizens have a right to exercise control over activities that threaten their well-being, and because preemption does not silence the risk aversion, civil disobedience, and opposition of persons who believe that their lives are in danger. Only rarely, for example, can states prevent a town from amending the maximum weight limit on a bridge it maintains in order to restrict truck access to a proposed risky facility. Instead, attempted preemption of local control of risks may simply force opponents to turn either to guerrilla tactics of opposition or to costly and time-consuming litigation. Or, as happened in Massachusetts, preemption could force local governments to use their power in state governments to defeat or prevent the operation of a preemption statute.64

Another response to criticisms of negotiation is to admit that many hazardous facilities (such as nuclear power plants) probably ought not to be sited, because there are less dangerous ways of providing the same benefits to society, and that any potential risk victims can be bribed

if they are somehow disadvantaged in or by society. That they are vulnerable to "bribes," and that they seek to remedy this disadvantage by trading safety for financial rewards, however, is not the fault of negotiation.65 It is not the fault of negotiation if people in Oregon, for example, vote for a copper smelter because they need the jobs, even though the smelter threatens the lives and safety of all citizens.66 It is not the fault of negotiation that any humans are so desperate that they feel forced into such choices. As Judith Jarvis Thomson puts it, "It is morally indecent that anyone in a moderately well-off society should be faced with such a choice . . . a choice between starving on the one hand, and running a risk of an incompensable harm on the other."67 It is the fault of society, not negotiation, if people face such choices. When societal balances of power are inequitable, risk negotiation will reflect those institutional imbalances. It cannot correct injustices already sanctioned by society. Risk negotiation can merely enfranchise those who heretofore had little voice in what hazards were imposed on them. Hence, negotiation does as much as can be expected of it.

Even though negotiation cannot necessarily prevent "bribing" the consumer, there is some evidence that such bribes are unlikely. Environmental sociologists report that noneconomic incentives—such as independent monitoring of risks and hazards, rights to information, and access to decisionmaking—are just as important to most citizens who are negotiating about local environmental risks as are financial incentives.68 If these findings are correct, then "buying off" citizens, in exchange for lessened safety requirements, may not be very likely.69

Moreover, apart from whether negotiation leads to what some might view as "bribes," consumer sovereignty and democracy require that people be allowed to make their own choices.70 As one philosopher put it, so long as they do not harm others, people ought to be allowed to choose cigarettes and saccharin; they ought to be able to choose nuclear fission and big cars, instead of being forced to accept solar power and small cars.71 As Mill and many other liberals recognized, people must be allowed to run whatever risks they wish, so long as the consequences affect only them.72

To oppose negotiation because it can lead to faulty citizen decisions or to consumers' being co-opted is unrealistic and narrow-minded. It is unrealistic because a technological society cannot avoid all risks. It is narrow-minded because it refuses to recognize that, in a democracy, everyone must make trade-offs. Everyone must have a mutual willingness to share power, make compensatory arrangements, and negotiate regarding risks, whether he is a developer of hazardous technology or a potential citizen victim. Not to negotiate would be to bring society to

a technological standstill. The objection that not everything is negotiable, that not everything has its price, is, of course, correct. Not all harms are compensable. That is why the federal government sets minimum environmental standards. Negotiation, however, is not proposed as a way to attain less stringent standards. Nothing in this proposal for negotiation suggests that we should abandon uniform environmental standards or that we should allow some communities to be victimized by less stringent risk requirements simply because a given industry has "bought them off." Rather, negotiation presupposes that a whole system of uniform environmental standards is in place, that such standards define what level of risk is minimally acceptable, and that people cannot "negotiate" their way to less stringent requirements. All that negotiation does it to compensate potential risk victims, give them more control over hazards, and reduce their risks.

Negotiation (with concomitant compensation or incentives) is at least in principle plausible because it is consistent with the legal and philosophical bases of neoclassical economic theory and the compensating wage differential. According to the theory behind the wage differential, persons can accept occupational risks provided that they are compensated and give free, informed consent to the risk imposition.73 Hence, if negotiation is wrong because it presupposes that compensation can sometimes be justifiable, then so are the foundations of much of our law, philosophy, and economics.

Another objection to negotiation is that it involves the unrealistic presupposition that all persons ought to give free, informed consent before risks are imposed on them. If society always adhered to this presupposition, an objector might claim, then many useful programs, such as public vaccination, could not be implemented.74 Arguing for universal vaccination, however, is not a case of attempting to justify denial of free, informed consent on grounds of expediency. The justification instead is the common good. Presumable, people do not have the right to refuse vaccination, because, if many do so, they will put other citizens at risk. Their right to free, informed consent ends where other persons' rights to bodily security and freedom from injury begin. Therefore, consent to technological risk is not analogous to consent to vaccination. Instead, the limits on one's free consent to vaccination are analogous, for example, to the limits on one's free consent to having a firearm taken away when it is likely that one will use it wrongly.75 And if so, then neither the vaccination nor the firearm case presents a counterexample to the requirement for consent.

Yet another objection to the negotiation scheme, with its attendant presuppositions about compensation, is that it would be both impossible

and impractical to compensate persons for all the technological and environmental risks that they face.76 But even though compensation may be "impractical," to dismiss it as such when there are ethical grounds for requiring it may be unjust, unless the impracticality is so great as to approach impossibility. If rights to compensation were recognized only when it was convenient to do so, there would be no rights. It would not work, for example, to tell a victim of racial discrimination that compensation is impractical, since presumably she has a right to compensation. Likewise, it will not do to tell a potential victim of environmental risk that compensation and, therefore, her rights are "impractical."

Compensation is also not impractical, as earlier arguments in this chapter have suggested, if it is the one way of negotiating community acceptance of a societal risk. Moreover, although compensation is difficult, because there are many hazards, we can nevertheless begin negotiating about, and compensating for, the worst environmental risks. Finally, the practicality of compensation and negotiation is established by the fact that our courts and our city councils have already employed them effectively. Compensation has been used to achieve successful community projects, as the Wyoming, Tennessee, New York, and Washington cases (cited earlier in this chapter) illustrate. Hence, it is at least prima facie plausible to argue that we can simply extend the record of these successes by devising federal statutes for risk negotiation and compensation.

Adversary Assessment and Risk Management

Perhaps one of the most disturbing objections to the negotiation and compensation scheme outlined earlier is that it presupposes a benign and cooperative regulatory climate. Yet, if those who disagree about environmental risk will not cooperate, then negotiation will not work. As one legal expert puts it,

If either party to a negotiation sees that more is to be gained from formal legal proceedings or obstructionist tactics than from negotiation, that party will abandon negotiation. The only alternative then would be to use some sort of adversary proceeding. This proceeding would include many of the same guarantees already discussed (earlier in the chapter) in connection with negotiation: (1) It would involve funding citizens' groups, so as to ensure that all parties in the adversary process had equal bargaining power. (2) Consistent with the previous chapter, it would guarantee consideration of different hazard assessments and alternative points of view about risk imposition and management. Finally, (3) the process would be controlled by disinterested parties, not by a regulatory agency.

Since I have written at length elsewhere about adversary proceedings and the "public jury" as vehicles for resolving environmental controversies, I shall not repeat those arguments and objections here.78 Instead, I shall merely sketch some of the main ideas involved in the notion of adversary assessment. Both scientists and laypeople would take part in adversary proceedings. The procedure would involve scientific experts presenting different technical positions, and social scientists and humanists arguing for alternative evaluative points of view. The final risk decision would be left to some democratic determination, probably a representative process, rather than to experts.

The precedent for adversary assessment of societal risks has been set by a number of citizen panels throughout the country. These are composed almost entirely of laypeople, not scientists, and many of them are responsible, for example, for the formulation and enforcement of scientific research guidelines.79 City councils in Cambridge (Massachusetts), San Diego, and Ann Arbor, for instance, have taken a number of initiatives in forming such citizen boards. In Cambridge, the city council authorized its city manager to appoint a citizen review board to evaluate the safety procedures required by the U.S. National Institutes of Health (NIH) for recombinant DNA research. Both the city council and the city commissioner unanimously approved the recommendations of the citizen review board.80

Need for Procedural Reform of Risk Management

Adversary proceedings would provide several benefits over the current system of decisionmaking regarding societal risk. First, an adversary system would require that funding be given to all sides involved in a controversy. Second, the adversary proceedings would make consideration of alternative positions a requirement of democratic decision-

making, rather than a luxury accessible only to those financially able to participate in administrative hearings or legal appeals. Third, unlike administrative and regulatory hearings, as well as negotiations, adversary procedures would be decisive. They would also be less likely to be co-opted by environmentalists or developers with vested interests, since they would not be dominated by a regulatory agency capable of exercising discretionary powers. Instead, they would be controlled by a group of citizens chosen because they had no apparent conflict of interest.

Admittedly, a few states and communities have had limited experience in using some of the procedural improvements in risk management suggested in this chapter (e.g., statutes requiring negotiation with, and compensation for, potential victims of societal risk). As a result, industry spokespersons have often alleged that these procedural reforms are not needed, or that they are already being accomplished. Both claims are false, as was argued in Chapter Eleven.

There are no federal statutes guaranteeing negotiation with, and compensation for, potential victims of all forms of societal risk. Currently, negotiation and compensation are not rights guaranteed by due process, but privileges accessible only to the moderately wealthy. In practice, these privileges are limited to those who are able to bear the transaction costs associated with adjudication through tort law. The whole point of this volume is that public consent to, and control over, risk evaluation and management are not the prerogatives of the rich, but the rights of all. They must therefore be protected by federal statute and administrative law, not subjected to the vagaries of circumstance.

Conclusion

Despite the alleged benefits of negotiation and adversary assessment, both procedures face many obstacles. For one thing, they are expensive (but perhaps not as expensive as the political unrest and loss of lives possibly resulting from erroneous societal decisions about environmental risk). Moreover, people may not be willing to pay the price, either for greater safety or for citizen negotiation. If they are not, they should not be forced to do so, as long as existing protections are equitably distributed.

Even if people would accept both negotiation and adversary assessment as vehicles for mitigating environmental risks, and even if they were willing to pay for them, however, there might still be objections. For one thing, I have not indicated specifically how negotiation and adversary assessment might work—for instance, who would take part

in the process. Although I have addressed some of these particulars elsewhere,81 most of them are better left to policymakers and social scientists. My purpose here has been to sketch some of the reasons why both negotiation and adversary assessment seem to be prima facie plausible.

Another possible objection to mitigating risks via negotiation and adversary assessment might be that reducing hazards is technically unworkable, allegedly because there are no economical, safe, feasible alternatives to existing risky technologies. If we decide to avoid use of commercial nuclear fission, for example, less risky energy options must be both developed and workable. Admittedly, all my remarks are predicated on the supposition that less risky technologies are both possible and economical. This supposition needs to be defended, on a case-by-case basis, for each environmental hazard. Obviously, there is no time to do so here, although such a defense conceivably could be given.82

Even without a defense of the thesis that less risky technologies are feasible, hazard evaluation could be more rational than it is. If the arguments in this volume are correct, we need to reform our methods, our statutes, our procedures, and (most important) our philosophies for making decisions about risks. On the ethical side, we need to recognize that persons have rights to compensation, to informed consent, and to due process; therefore, they have rights to negotiate about, or perhaps even prohibit, the hazards others wish to impose on them.

On the epistemological side, we need to recognize that risk evaluation and management are irreducibly political (normative), in much the same way that quantum mechanics is irreducibly statistical (nondeterministic). In physics, we have come to realize that quantum-mechanical measurements "interfere" with the state of the system being measured. In applied science and environmental policy, we have been slower to realize that the human components of societal risk evaluation cannot be removed, even though they "interfere" with positivistic measures of risk.

This book has been a first step in suggesting how we might adopt a more democratic and procedural account of rationality, so as to reflect the human dimensions of hazard assessment and evaluation. It has argued that the public is frequently rational in its risk evaluations and that, even when laypersons are wrong about risk, they often have the right to be wrong. Even, and especially, victims have choices.