INTRODUCTION

Why are normative theories so prevalent in the study of judgment and choice, yet virtually absent in other branches of science? For example, imagine that atoms and molecules failed to follow the laws supposed to describe their behavior. Few would call such behavior irrational or suboptimal. However, if people violate expected utility axioms or do not revise probabilities in accord with Bayes's theorem, such behavior is considered suboptimal and perhaps irrational. What is the difference, if any, between the two situations? In the latter we implicitly assume that behavior is purposive and goal-directed while this is less (if at all) obvious in the former. (It is problematic how one might treat plant and animal behavior according to a descriptive–normative dichotomy.) Therefore, if one grants that behavior is goal-directed, it seems reasonable to assume that some ways of getting to the goal are better, in the sense of taking less time, making fewer errors, and so on, than others. Indeed, much of decision research concerns evaluating and developing ways for improving behavior, thereby reflecting a strong engineering orientation (Edwards, 1977; Hammond, Mumpower and Smith, 1977; Keeney and Raiffa, 1976). Moreover, comparison of actual behavior with normative models has been important in focusing attention on the discrepancies between them, and this in turn has raised important questions about the causes of such discrepancies.