Risk Acceptability According to the Social Sciences

4. Choice and Risk

Cllapter 4 Cllolee and Rlslc Summary: Nineteenth-century theorizing about Iisk separated gambling from other Iisk-taking. Within the modern theory of choice, Iisk aversion and riskseeking are now presented within an integrated theory. This chapter points out some limitations of the theory of choice for the question of ascertaining public levels of tolerance for Iisk. The theory of choice applies logic to the act of choosing. The rational argument is one that is not self-contradictory and likewise the rational choice. To be rational, one choice does not negate another . Rational behavior implies some ordering of alternatives in terms of relative desirability. The logic of choice concerns noncontradictory or ordered preferences. In science, probabilities are assessments of the reliability of expectations about events. Probabilities also figure prominently in the theory of choice. It makes a lot of difference to a decision if the alternatives involve choosing between a certainty and an uncertainty, or between a low and a high probability. The variance of the probability constitutes the risk element. In the seventeenth and eighteenth centuries theorizing about risk concerned the mathematics of gambling, and so the focus was precisely on the structure of probabilities as a whole. In the nineteenth century the theorizing about risk shifted from gaming to the risks of economic enterprise and particularlv to the probabilities of loss. Inevitably, the theory of risk has COl i Ie to be emphasized as the probability of not reaching an objective - with a negative judgment passed against the very long shot gamble for high stakes. The traditional arguments have been 41 about the relation between the objectively calculated (or mathematical ) probabilities and values as compared with the subjective estimations of the rational agent; about the most useful definition of rationality that can be devised for understanding the logic of choice; and about the adequacy of the match between this definition and actual behavior. Most of the path-finding work on risk, determining the way it is now discussed, has been put in hand before the recent grave concern about risks from technology. It started with the publication in 1944 of The Theory of Games by Von Neumann and Morgenstern and emerged complete as to its main axioms and theorems in a brief five years between 1948 and 1953. Later developments have been more in the nature of fine-tuning, criticisms , and improvements of different parts of a formidably rigorous conceptual apparatus. Subsequently, decision theory has been applied to practical questions of military strategy. There has also been some effort to apply these methods to the risks of industrial technology. Utility theory has been applied, with perplexing results, to assessing the compensation for a life or the claims of unborn generations. There is a real question as to how well the whole theoretical system may be adapted to producing answers to public policy questions about risks arising from nuclear power or toxic industrial wastes. Yet these are the kinds of questions the new subdiscipline has been formed to answer. Every choice we make is beset with uncertainty. That is the basic condition of human knowledge. A great deal of risk analysis is concerned with tlying to turn uncertainties into probabilities. What seems to be in each case a purely technical exercise quickly becomes one that rests directly upon the philosophical foundations of inference. Isaac Levi's "Brief Sermon on Assessing Accident Risks in U.S. Commercial Nuclear Power Plants," is a close examination of the statistical procedures used by the Nuclear Regulatory Commission's Reactor Safety Study (1975). The appendices to that report discuss the difficulties involved in making evaluations of the probabilities and the methods employed in meeting them. Little could the engineers who wrote the report have expected to find an epistemologist question their choices and particularly their reasons for them. The authors of the report, however, did seem anxious to ground judgments of credal probability on judgments as 42 to which of rival statistical hypotheses concerning the objective chance distribution of failure rates is correct. In the face of insufficient data, the authors did not conclude that they should suspend judgment between the rival statistical hypotheses and look to their credal state for the various seriously possible rivals for help in determining via direct inference which credal state to adopt. That approach would have led to an indeterminate state of credal judgment concerning failure rates. [po441 J Levi then goes on to relate their methodological...