Incentive compatibility in sender-receiver stopping games.

Aditya Aradhye, János Flesch, Mathias Staudigl, Dries VermeulenGames 2020 · DOI · OpenReview

TL;DR

This paper introduces a novel sender-receiver game where a sender with private information advises a receiver on whether to stop or continue a game with uncertain future states. They demonstrate that under certain conditions, a unique equilibrium emerges where the receiver blindly follows the sender's advice, allowing the sender to unilaterally maximize their payoff. This finding is significant as it reveals a scenario where delegation of decision-making to an informed party can be perfectly aligned with the delegator's interests.

What Problem Does This Paper Address?

The paper addresses the challenge of strategic information transmission in dynamic settings, specifically when one player (the sender) has private information and influences the decisions of another player (the receiver) who ultimately controls the game's duration. Existing models often focus on static interactions or games with fixed horizons. This work introduces a "stopping game" element, where the receiver can choose to end the game at any point, adding a new layer of strategic complexity. The core question is how information is transmitted and how incentives align (or misalign) when the receiver's decision to continue or stop is influenced by the sender's message, and when the receiver can unilaterally end the game.

Methods

The authors employ the concept of Perfect Bayesian Equilibrium (PBE) to analyze the game. A key methodological contribution is the characterization of a "regular strategy profile," which is a specific type of PBE where the receiver is "obedient" (always follows the sender's recommendation) and the sender is "sincere" (sends a message that truthfully reflects their preference given the state and the receiver's obedience). They prove the existence and uniqueness of such responsive PBEs under conditions of sufficient player patience (high discount factors). The analysis involves backward induction for finite horizons and relies on identifying threshold strategies for the sender.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • The "Obedient Receiver" Equilibrium: The most striking insight is the existence of an equilibrium where the receiver perfectly trusts and obeys the sender's advice. This is surprising because the receiver has their own payoff function and could potentially deviate.
  • Sender's Unilateral Power: In this "regular" equilibrium, the sender effectively dictates the game's outcome. The receiver's decision to stop or continue is solely based on the sender's message, not on their own assessment of the state.
  • Threshold Strategies: The sender's optimal strategy in this equilibrium is a simple threshold strategy: "continue" below a certain state threshold and "stop" above it. This simplicity makes the equilibrium computable.
  • Patience is Key: The existence and uniqueness of these favorable equilibria are contingent on the players being sufficiently "patient" (i.e., having high discount factors), meaning they value future payoffs significantly.
  • Breakdown of Communication: For certain parameter choices (e.g., low discount factors or undiscounted infinite horizons), responsive PBEs may not exist, indicating a potential breakdown in effective communication.

Key Results

  • Finite Horizon: For finite horizon games with sufficiently high discount factors (or undiscounted payoffs), the regular strategy profile is the unique responsive PBE. However, for small discount factors, no responsive PBE may exist.
  • Infinite Horizon: For discounted infinite horizon games with high discount factors, the regular strategy profile is the unique responsive PBE among "essentially Markov" strategy profiles. For period-independent payoffs, no such PBE exists.
  • Computability: The regular strategy profile is shown to be entirely computable, relying solely on the sender's payoff function.

Limitations and Future Work

The authors acknowledge several limitations and suggest avenues for future research:

  • Arbitrary State Distributions: The current analysis assumes states are drawn from a uniform distribution. Extending the model to arbitrary state distributions is a natural next step.
  • Markov Chain States: The paper assumes i.i.d. state draws. Investigating games where states follow a Markov chain would add significant realism.
  • Multiple Senders: The current model features a single sender. Exploring scenarios with multiple senders, where the receiver can cross-reference information, is a promising direction.
  • Non-Responsive PBEs: While the paper focuses on responsive PBEs, the authors note that non-responsive (babbling) PBEs are generally easy to construct by adding babbling periods.
  • Discount Factor Sensitivity: The existence and nature of equilibria are highly sensitive to the discount factor. Further exploration of the phase transitions between different equilibrium structures would be valuable.

Summary updated

Abstract

We introduce a model of sender-receiver stopping games, where the state of the world follows an iid--process throughout the game. At each period, the sender observes the current state, and sends a message to the receiver, suggesting either to stop or to continue. The receiver, only seeing the message but not the state, decides either to stop the game, or to continue which takes the game to the next period. The payoff to each player is a function of the state when the receiver quits, with higher states leading to better payoffs. The horizon of the game can be finite or infinite. We prove existence and uniqueness of responsive (i.e. non-babbling) Perfect Bayesian Equilibrium (PBE) under mild conditions on the game primitives in the case where the players are sufficiently patient. The responsive PBE has a remarkably simple structure, which builds on the identification of an easy-to-implement and compute class of threshold strategies for the sender. With the help of these threshold strategies, we derive simple expressions describing this PBE. It turns out that in this PBE the receiver obediently follows the recommendations of the sender. Hence, surprisingly, the sender alone plays the decisive role, and regardless of the payoff function of the receiver the sender always obtains the best possible payoff for himself.
BibTeX
@inproceedings{aradhye2020incentive,
  author = {Aditya Aradhye and János Flesch and Mathias Staudigl and Dries Vermeulen},
  title = {{Incentive compatibility in sender-receiver stopping games.}},
  booktitle = {Games},
  year = {2020},
  doi = {10.1016/j.geb.2023.06.008},
}
Figure