Students of game and decision theory uniformly seem to agree that the representation of feasible plans for future actions in extensive or tree form is equivalent to the representation of such plans in normal or strategic form--at least, insofar as the problem is to evaluate feasible plans for admissibility. In this project, Professor Levi will undertake a critical examination of the assumptions about feasibility which underlie this widely shared presupposition and will point out how these assumptions are required to sustain some widely advocated arguments purporting to support conditionalization and maximization of expected utility--the twin pillars of Bayesianism. Professor Levi believes that this critical examination will thus undermine the Bayesian's assumptions and arguments alike. He will then explore the ramifications of questioning these assumptions about feasibility for decision theory, economics, statistics, epistemology and the topic of freedom of the will. He will embed the results of this study of planning future choices in a larger project aimed at providing a comprehensive critique of Bayesianism and an exploration of the merits of rival approaches. This larger project will be done in collaboration with Dr. T. Seidenfeld of Carnegie Mellon University.
