Institute of Information Science, Academia Sinica

This talk will provide a fine analysis of the classical theory of subjective expected utility of Leonard J. Savage. The talk has three main parts. First, I will comment on subjectivists’thesis of reducing a theory of probability into a normative theory of rational decision making, and I will outline the general framework devised by Savage. In the second part, I will discuss an unsought defect of Savage’s decision model due to fact the derived subjective probability in Savage’s system is merely finitely but not countably additive. It will be pointed out that Savage's argument for rejecting countable additivity is inconclusive due to an oversight of set-theoretic details. This will be followed, in the third part, by my proposed solution to the problem. It will be shown that Savage’s model can still be maintained as a coherent system once we replace his last postulate with a new continuity axiom.