WHAT COMMON GROUND EXISTS FOR DESCRIPTIVE, PRESCRIPTIVE, AND NORMATIVE UTILITY THEORIES

Descriptive and normative modeling of decision making under risk and u ncertainty have grown apart over the past decade. Psychological models attempt to accommodate the numerous violations of rationality axioms, including independence and transitivity. Meanwhile, normatively orien ted decision analysts continue to insist on the applied usefulness of the subjective expected utility (SEU) model. As this gap has widened, two facts have remained largely unobserved. First, most people in real situations attempt to behave in accord with the most basic rationalit y principles, even though they are likely to fail in more complex situ ations. Second, the SEU model is likely to provide consistent and rati onal answers to decision problems within a given problem structure, bu t may not be invariant across structures. Thus, people may be more rat ional than the psychological literature gives them credit for, and app lications of the SEU model may be susceptible to some violations of in variance principles. This paper attempts to search out the common grou nd between the normative, descriptive, and prescriptive modeling by ex ploring three types of axioms concerning structural rationality, prefe rence rationality, and quasi-rationality. Normatively the first two ar e mandatory and the last, suspect. Descriptively, all have been questi oned, but often the inferences involved have confounded preference and structural rationality. We propose a prescriptive view thal entails f ull compliance with preference rationality, modifications of structura l rationality, and acceptance of quasi-rationality to the extent of gr anting a primary role to the status quo and the decomposition of decis ion problems into gains and losses.