Dynamic, Behavioral, and Multi-Agent Persuasion Mechanisms

Project: Research project

Project Details


This proposal studies persuasion mechanisms in dynamic models, with behavioral agents and in multi-agent settings. Persuasion mechanisms are used to study the use of information disclosure as an incentive instrument in principal/agent environments. The principal is assumed to be better informed than the agent and selective disclosure of information can alter the agent's private incentives to take actions. When there is a conflict of interest between the principal and agent such disclosure can benefit the principal. In a dynamic setting the principal is privately observing a stochastic process and sends messages to the agent over time to control the evolution of the agent's beliefs. In this novel framework I show a convenient and intuitive characterization of the optimal mechanism and apply it to several new examples. I discuss various extensions including a long-run/strategic agent, endogenous evolution of the state, and endogenous information. I also show how the model can be analyzed in both discrete and continuous time. Persuasion mechanisms add a new dimension to behavioral mechanism design. In previous work I and my co-authors Emir Kamenica and Alex Frankel developed a model of suspense and surprise using a dynamic behavioral persuasion model in which the agent's have direct preferences over the evolution of their beliefs. I also discuss the role of information disclosure in standard mechanism design applications when the agents involved care not just about outcomes but also the information revealed. Finally there has been little progress on persuasion mechanisms with many agents. The new complication here is to characterize not just the beliefs that can be implemented but also the higher order beliefs that are relevant for interacting agents. In joint work with Marcin Peski we are exploring this question.
Effective start/end date8/1/147/31/17


  • National Science Foundation (SES-1427200)


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.