Between Normative and Descriptive: The Role of (Bounded) Rationality


Stanford University


Prof. Tomaso A. Poggio
Abstract: Classical frameworks for rationality — Bayesian decision theory prominent among them — enjoy elegant theoretical simplicity as well as widespread use, from behavioral sciences to engineering to philosophy. It has long been appreciated that both descriptive and normative applications might benefit from less idealization, e.g., by incorporating computational constraints. Equally canonical frameworks, however, seem not to have emerged, leading some to conclude, “There is no unified theory of bounded rationality, and probably never will be” (R. Aumann). The aim of the talk is to revisit this set of issues, focusing especially on one illustrative question: how should we understand the status of randomized decisions in intelligent behavior?