Independent Reading · 2021

Markovian Persuasion

Ehud Lehrer and Dimitry Shaiderman · arXiv preprint

Bayesian PersuasionGame TheorySequential Decision-Making

Full text

Read the Paper

Use View PDF or Download PDF to read the full paper on this device.

Paper Information

Authors
Ehud Lehrer and Dimitry Shaiderman
Venue
arXiv preprint · 2021
Open official paper

Summary

This paper studies Bayesian persuasion when the state of nature evolves according to a Markov process.

Core Idea

An informed sender must balance immediate gains from information disclosure against the effect of current signals on future receiver beliefs and payoffs.

Method

The paper analyses optimal disclosure strategies under different discount factors and characterises conditions for attaining the maximal asymptotic value.

Key Takeaways

  • Information disclosure has intertemporal consequences in dynamic environments.
  • Belief evolution depends jointly on signalling and the underlying Markov transition law.
  • Optimal persuasion may require sacrificing short-term payoff to influence future beliefs.

Questions

  • How do optimal policies change when state transitions are only partially known?
  • What happens when the receiver also learns or adapts strategically over time?