Beneficially Imperfect Persuaders
When a decision maker’s (DM’s) choice depends on the information provided by persuaders, does the DM benefit from that information? I address this question in the context of a Bayesian persuasion game in which independent persuaders with no private information try to persuade a DM by gathering information using verifiable tests. All persuaders want the DM to switch her action from a default action to a new action, but whether the DM also finds it optimal to switch depends on the state of the world. The persuaders strategically design tests that may be biased towards the new action and that best respond to the test designs of the other persuaders. I show that although the DM never benefits when there is only one persuader, there always exist strict equilibria with high payoffs for the DM when there is more than one persuader, even if the persuaders have perfectly aligned incentives. Moreover, in these equilibria, persuaders choose noisy tests that sometimes misreport their desired state as the undesired one.