Talk: Patricia Rich (Uni Bayreuth)

Title:

The Ecological Rationality of Ambiguous Beliefs

Joint work with Paolo Galeazzi

Abstract:

People regularly face decisions in which there is ambiguity; that is, the true probabilities of decision-relevant events are not given. In such cases, there are different possible ways to respond. The agent could form and use a specific, subjective probability function, for example maximizing their expected utility. Alternatively, they could retain a less specific belief, represented for example by a set of probability functions. For this latter case, various decision rules have been studied, such as maxmin, regret minimization, and the Hurwicz criterion. Ambiguous beliefs and the associated decision rules have been thoroughly discussed and studied from the perspective of classical rationality, primarily by considering which axioms agents conform to or violate. We know very little about the ecological rationality of ambiguity, however. Ecological rationality pertains to the expected real-world outcomes, and especially how a decision rule's performance depends on the context in which it is used. I report on two simulation studies of the ecological rationality of ambiguity; both simulate the evolution of ambiguous and non-ambiguous agent types in a varied environment. The first study compares expected utility maximization to classic choice criteria such as the Hurwicz criterion. The key finding is that a moderate Hurwicz criterion is the best choice heuristic for small menus, while expected utility maximization is the best choice heuristic for large menus. The second study compares generic single agent decision problems, generic games, and coordination games. Ambiguous beliefs and decision rules are shown to be evolutionarily viable in the context of games, and especially ecologically rational in coordination games. I explain the mechanisms behind these results.