Decision-making with partial information

Abstract

In this paper, we study choice under uncertainty with belief functions. Belief functions can capture partial information by describing what is objectively known about the probabilities of events. State-contingent acts together with a belief function over states induce belief functions over outcomes. We assume that decision makers have preferences over belief functions that reflect both their valuation of outcomes and the information available about the likelihood of outcomes. We provide axioms characterizing a preference representation for belief functions that captures what is (objectively) known about the likelihood of outcomes and combines it with subjective beliefs according to the “principle of insufficient reason” whenever the likelihood of events is unknown. This treatment of partial information yields a natural distinction between ambiguity and ambiguity attitudes. The approach is novel in its treatment of partial information and in its axiomatization of the uniform distribution in case of ignorance.