Modeling Ignorance without Bayesian beliefs

Talk based on a monograph: "Ignorance and Uncertainty" (joint with Andy Postlewaite) and recent applications (not in the book)

Summary of the talk: The book is an attempt to challenge the way we deal with ignorance in models. In this talk, I discuss some issues with standard Bayesian models, and review a number of examples in which we propose an alternative path. As we see it, one of the main modeling challenge is that a model combines two perspectives: that of the analyst that makes precise the model (preferences, signals, distributions etc...), and that of the agents, who we would like to be not too omniscient, but who are nevertheless assumed to behave as if they knew the model in every details. We advocate direct strategy restrictions as a way to limit the agents' ability to exploit details of the model that are not meant to be known by them, or at least not intended to significantly drive their decisions. I will discuss examples of (plausible?) strategy restrictions and what these restrictions can accomplish in different settings (All pay/Blotto, repeated games, information aggregation in networks).