Beliefs and Dynamic Consistency,

In this chapter, we adopt the decision theoretic approach to the representation and updating of beliefs. We take up this issue and propose a reconsideration of Hammond's argument. After reviewing the argument more formally, we propose a weaker notion of dynamic consistency. We observe that this notion does not imply the full fledged sure thing principle thus leaving some room for models that are not based on expected utility maximization. However, these models still do not account for ''imprecision averse" behavior such as the one exhibited in Ellsberg experiment and that is captured by non-Bayesian models such as the multiple prior model. We therefore go on with the argument and establish that such non-Bayesian models possess the weak form of dynamic consistency when the information considered consists of a reduction in imprecision (in the Ellsberg example, some information about the proportion of Black and Yellow balls)