Dynamic Semi-Consistency Very Preliminary and Incomplete

The behavior of dynamically consistent agents who follow through with any ex ante optimal plan cannot be distinguished from the behavior of Bayesians. The notion of dynamic semi-consistency allows for models in which different ambiguity attitudes generate different predictions while at the same time keeping the normatively appealing tenet that agents follow through with ex ante optimal pure strategies. Semi-consistent agents do, however, not follow through with all ex ante optimal mixed strategies, as they do not update their preferences upon learning independent randomization outcomes. If players are dynamically semi-consistent, the equilibria of games with different ambiguity attitudes may differ substantially. If we complement the assumption of ambiguity aversion with the assumption of dynamic consistency, the equilibrium sets of games with ambiguity averse agents coincide with the equilibrium sets of Bayesian games. While the standard revelation principle for mechanism design applies in full force if agents are dynamically consistent a modification is required for semi-consistent