The Simplicity of Large Bayesian Persuasion Strategies
(joint with Yakov Babichenko and Rann Smorodinsky)
Abstract: We consider a Bayesian persuasion problem with a one-dimensional interval as the state space. We find a simple class of semi-interval policies such that every persuasion problem has an optimal solution in the class. Conversely, for any policy in the class, we show that there exists a persuasion problem that the policy is its unique maximiser. We discuss our results in the context of a partially informed sender.