Reaping the Informational Surplus in Bayesian Persuasion

Ronen Gradwohl, Niklas Hahn, Martin Hoefer, Rann SmorodinskyarXiv 2020 · OpenReview

Abstract

The Bayesian persuasion model studies communication between an informed sender and a receiver with a payoff-relevant action, emphasizing the ability of a sender to extract maximal surplus from his informational advantage. In this paper we study a setting with multiple senders, but in which the receiver interacts with only one sender of his choice: senders commit to signals and the receiver then chooses, at the interim stage, with which sender to interact. Our main result is that whenever senders are even slightly uncertain about each other's preferences, the receiver receives all the informational surplus in all equilibria of this game.

Introduction

In the foundational model of Bayesian persuasion (Kamenica and Gentzkow, 2011), a single informed sender can strategically reveal information to influence a receiver's actions. By carefully designing a signal, the sender can often extract the majority of the "informational surplus"—the value generated by the private data—leaving the receiver with minimal gains.

While subsequent research has explored scenarios with multiple senders, these models typically assume the receiver can process all available signals simultaneously. However, real-world decision-making is often characterized by "congestion in advice." A policymaker may only have time to consult one lobbyist; a patient might seek a single second opinion; or a judge might hear from only one expert witness. This paper investigates how this restriction—the necessity of choosing exactly one sender—fundamentally alters the balance of power between experts and decision-makers.

The Persuasion Game with Limited Attention

The authors model this interaction as a three-stage game involving $n$ senders and one receiver:

  1. Commitment: Senders simultaneously commit to a "signal" or reporting strategy. This is a rule determining how they will report information based on the underlying state of the world.
  2. Interim Choice: The receiver observes all committed signal rules but must choose exactly one sender to interact with.
  3. Realization: The state of the world is realized. The chosen sender provides a message according to their pre-committed rule, and the receiver takes an action based on that message.

To ensure the model reflects realistic environments, the authors rely on two mild assumptions. First, every player has a unique optimal action in every state (eliminating preference ties). Second, senders are not perfectly aligned; there is at least a "weak" misalignment where one sender might prefer the receiver's optimal action while another does not.

The War of Attrition for Attention

The core insight of the paper is that the receiver’s limited attention acts as a catalyst for intense competition. Because a sender receives a payoff only if they are selected by the receiver, they enter a competitive race to be the most "attractive" source of information.

To be chosen, a sender must offer a signal that provides the receiver with more expected utility than any signal offered by a competitor. This dynamic creates a "race to the bottom" regarding secrecy, or more accurately, a race to the top regarding transparency. While a monopoly sender uses commitment power to withhold information, competitive senders use that same commitment power to "out-bid" each other with clarity.

Interestingly, the receiver is actually better off being restricted to a single signal. In models where receivers can listen to everyone, senders can sometimes implicitly collude to withhold information. When the receiver is forced to pick just one, that collusion breaks down, as any sender can "steal" the receiver's attention by offering slightly more information.

Main Results: Full Informativeness

The paper’s primary contribution is Theorem 1, which establishes that every Nash Equilibrium of this game is fully informative. In any stable state of competition, the receiver learns everything necessary to take their personally optimal action.

This leads to a total transfer of the informational surplus. The extra value derived from the private information—which the sender would capture in a monopoly—is entirely reaped by the receiver. In illustrative examples, such as a regulator choosing between competing industry experts, the receiver’s utility reaches its theoretical maximum, while the senders' utilities drop significantly compared to the single-sender benchmark.

This result is remarkably robust, characterized by an "$\epsilon$ discontinuity." Even if there is only a negligible chance (e.g., 0.01%) that senders' interests are misaligned, the equilibrium still collapses into full disclosure. The sender-advantage only returns if the senders are perfectly, 100% aligned in their goals.

Limitations and Conclusion

While the result is powerful, the authors note specific boundary conditions. The model assumes a finite state space and relies on the assumption that senders are not perfectly aligned. Furthermore, while the result holds under various tie-breaking rules (how the receiver chooses when two senders offer identical utility), certain complex, "punishing" tie-breaking behaviors could theoretically disrupt the equilibrium.

Ultimately, the paper demonstrates that in a world of competing experts, a decision-maker's limited bandwidth is not a liability but a source of power. By forcing experts to compete for a single "slot" of attention, the receiver can compel full transparency and capture the entirety of the informational surplus.

Summary updated

Abstract

The Bayesian persuasion model studies communication between an informed sender and a receiver with a payoff-relevant action, emphasizing the ability of a sender to extract maximal surplus from his informational advantage. In this paper we study a setting with multiple senders, but in which the receiver interacts with only one sender of his choice: senders commit to signals and the receiver then chooses, at the interim stage, with which sender to interact. Our main result is that whenever senders are even slightly uncertain about each other's preferences, the receiver receives all the informational surplus in all equilibria of this game.
BibTeX
@misc{gradwohl2020reaping,
  author = {Ronen Gradwohl and Niklas Hahn and Martin Hoefer and Rann Smorodinsky},
  title = {{Reaping the Informational Surplus in Bayesian Persuasion}},
  year = {2020},
  eprint = {2006.02048},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2006.02048},
}
Figure