Center for Social Information Sciences (CSIS) Seminar

Abstract: We study the design of persuasion games – settings where many senders commit to report statistical data about their unrealized private information to a receiver who seeks to implement some arbitrary allocation rule. To characterize the set of implementable equilibrium allocations, we develop a revelation principle for settings where sender reports are Blackwell experiments as opposed to type-dependent messages. This allows incentive constraints to be formulated as a convex program which depends jointly on the ``worst-case'' punishments prescribed at each posterior belief and the allocation rule. We next introduce an iterative two-step algorithm which simplifies the set of posteriors that pin down implementability. Finally, we apply our results to analyze the impact of ex-ante uncertainty on strategic behavior in several settings, including contest design, disclosure in teams, and matching with preference uncertainty.