Adaptive preferences: An evolutionary model of non-expected utility and ambiguity aversion

Abstract

We enrich an evolutionary model with common and idiosyncratic uncertainty as in Robson (1996) by allowing for hidden actions (or phenotypic flexibility). In contexts where common uncertainty is ambiguous and idiosyncratic uncertainty is risky, the model generates both ambiguity aversion and non-expected-utility preferences for risk, thereby providing a link between the two types of behavior. While the general evolutionarily optimal objective function does not have an obvious similarity to functional forms studied in the literature, our main results show that it can be recast as a combination of familiar models of ambiguity aversion and non-expected utility. We show that particular classes of hidden actions generate the special cases of rank-dependent or divergence risk preferences embedded within a model of ambiguity aversion.