Regret-minimizing Bayesian persuasion

We revisit the celebrated Bayesian persuasion setting and examine how well Sender can perform when ignorant of Receiver's utility. Taking an adversarial approach, we measure signaling scheme performance via (additive) regret over a single instance. focus on Receiver with two actions: adoption rejection , assuming that aims to persuade adopt. show while Sender's total ignorance utility is extremely harmful her, just knows ordinal preferences states nature upon suffices guarantee surprisingly low even number tends infinity. Moreover, exactly pin down minimum in this case. further such positive result impossible under alternative multiplicative approximation ratio .