Attributes for Risk Evaluation. Karen Jenni and Baruch Fischhoff, Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, PA 15213
Risk policy makers are increasingly recognizing the need to consider factors other than simply expected fatalities in their concept of "risk." This paper addresses the question of what other risk-related factors should be included in evaluating, comparing, and ranking risks. We propose that to provide a solid basis for differential risk policies, a risk attribute or dimension should (a) reflect what people care about, (b) be specified in a way that reflects the normative basis of those concerns, and (c) be measurable. Claims about whether a variable reflects people’s concerns can be either normative or empirical. We consider normative arguments for incorporating each of the most commonly used risk variables in risk-based decisions, and point to the empirical evidence for their importance from the psychometric literature. In addition, we report on two original empirical studies of risk perceptions that look at the relationship between the qualitative risk dimensions and the judged acceptability of hazards using the psychometric approach, and also explicitly address the prescriptive uses of those dimensions. The relative importance of various risk dimensions for risk-based decision-making is assessed implicitly through a set of psychometric tasks and policy-capturing analysis, and directly through a set of important questions. Results are used to recommend a set of variables to be used in the Carnegie Mellon risk-ranking exercise for school-related risks. The results have implications for risk policy generally, as an approach for determining what risk variables people value and believe should be used for establishing differential risk policies.
Work supported by an NSF Graduate Research Fellowship, by NSF grant SRB-9512023, by a grant from CMA, and by EPRI grant WO2955-12.