Axiomatic characterization for non-Bayesian receivers

Extend the axiomatic characterization of robust Bayesian persuasion representations to settings in which the receiver’s choice correspondence is not represented by a Bayesian information processor.

Background

The axiomatic characterization in the paper assumes that the receiver updates beliefs according to Bayes’ rule and chooses an act maximizing expected utility. The authors explicitly state that this Bayesian-receiver assumption is essential to their proof, particularly because a key lemma relies on the receiver’s choice correspondence admitting a Bayesian representation.

The paper therefore leaves unresolved whether an analogous characterization can be obtained for non-Bayesian receivers, such as receivers who process information in ways that depart from Bayesian updating or expected-utility maximization.