A Simple Experiment on Simple Bayesian Persuasion (joint with Pak Hung Au and King King Li)
Abstract : This experiment tests Bayesian persuasion (Kamenica and Gentzkow, 2011) in a simple setting. We adopt an experimental design where the Sender chooses a partition of the state space instead of an information structure. We find that 1) the Senders generally set the weaker signal to be fully revealing, but 2) their strategies are persistently suboptimal in the sense that the stronger signal is systematically set weaker than what the Receivers require to take the Senders' preferred action, resulting in persistently high rate of persuasion failure. However, 3) once we replace the Receivers with a robot who plays a known strategy, most Senders quickly learn to play the optimal strategy. This suggests that the key strategic element of Bayesian Persuasion is easy to understand for the Senders, although guessing what posterior probability a human Receiver would require to take the preferred action is a more difficult problem.
* 본 세미나는 VEAEBES(Virtual East Asia Experimental and Behavioral Economics Seminar series) 주최로 열리는 세미나입니다.
참여신청은 아래의 링크로 해주시길 바랍니다.
이후 이메일로 zoom링크가 발송됩니다.