Event Detail

Decision theory is supposed to accommodate any preferences that a rational decision maker might have. I show that standard decision theory – expected utility theory – cannot accommodate common preferences that stem from the way apparently rational decision makers account for risk. I propose a generalization of expected utility theory that can accommodate these preferences, and show that it has all the theoretical power of the standard theory: in particular, it has a representation theorem that allows us to derive an agent’s beliefs, desires, and attitudes towards risk from his preferences. I then explore a classic strategy for responding to counterexamples of the type I present: re-characterizing the choice problem facing an agent, by individuating outcomes more finely. This introduces an important question in the methodology of decision theory: under what circumstances should we interpret the agent as facing a different choice problem rather than adopt a different theory to describe his behavior? I approach this question formally, and show that we cannot individuate the outcomes to make his behavior compatible with expected utility theory without also making his behavior compatible with many other opposing theories: thus, I suggest, we cannot save the standard theory.