Sequential Non-Bayesian Persuasion

2025-08-12

AI Summary

This paper studies sequential persuasion when the receiver employs “conservative Bayesian” belief updating—i.e., beliefs are convex combinations of the prior and the Bayesian posterior. Motivated by the classical result that sequential information is valueless under standard Bayesian updating, we develop a non-Bayesian belief-updating game model and conduct joint sequential information design and theoretical analysis. Our contributions are threefold: (1) We prove that sequential persuasion yields strictly higher sender utility—even in environments where sequential disclosure is useless under the optimal single-period information structure—provided the receiver is conservative Bayesian. (2) When both sender and receiver hold belief biases, the sender’s maximum expected utility equals that under unbiased beliefs. (3) We extend non-Bayesian persuasion theory by establishing the robust value of sequential mechanisms against cognitive biases in information design.

Technology Category

• Reasoning under Uncertainty: Sequential Decision Making
• Game Theory and Economic Paradigms: Mechanism Design
• Multiagent Systems: Mechanism Design

Application Category

• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Responsible Web: Consent frameworks and practices on the web

Abstract

We study a model of persuasion in which the receiver is a `conservative Bayesian' whose updated belief is a convex combination of the prior and the correct Bayesian posterior. While in the classic Bayesian case providing information sequentially is never valuable, we show that the sender gains from sequential persuasion in many of the environments considered in the literature on strategic information transmission. We also consider the case in which the sender and receiver are both biased and prove that the maximal expected payoff for the sender under sequential persuasion is the same as in the case where neither of them is biased.

Problem

• Study persuasion with conservative Bayesian receiver
• Show sender gains from sequential persuasion
• Compare sender payoff in biased vs unbiased cases

Innovation

• Sequential persuasion for conservative Bayesian receivers
• Convex combination of prior and Bayesian posterior
• Sender gains from sequential information transmission