Using an evolutionary approach, we address the prominent tensions in the literature that updating of ambiguity and non-expected-utility preferences cannot, in general, be both dynamically consistent and consequentialist. Perhaps not surprisingly, evolutionary optimality requires dynamic consistency. The more subtle insight is the evolutionary optimality of systematic violations of consequentialism. We base our investigation on the model of adaptive preferences (Sadowski and Sarver (2024)), which generalizes the model of Robson (1996) and nests variants of many well known models in the literature.

We undertake a decision-theoretic analysis of a model of bilateral risk sharing, conceptualized in two stages: in the first stage agents choose risky endowments (Savage acts), and in the second stage they form a risk-sharing arrangement. Only the first-stage choices are observable to the analyst. We formulate axioms that put joint restrictions on best responses in the first stage and a representation result according to which agents behave as if they are risk averse expected-utility maximizers who anticipate the subsequent sharing arrangement. All the parameters of the model, including the sharing arrangement, can be identified from this first-stage choice data.

We axiomatize a model of preferences over menus of acts in which not only beliefs but also state-dependent utilities depend on the individual’s choice of information. Our most general model features both contemplation about the appropriate way to evaluate alternatives as well as acquisition of information about the payoff relevant state of the world, before a choice is made. We then focus on the special case where the value of alternatives depends directly and exclusively on the state of the world and on the choice of information about that state.