Algorithmic Cheap Talk

Yakov Babichenko,Inbal Talgam-Cohen,Haifeng Xu,Konstantin Zabarnyi

2024-09-10

Abstract:The literature on strategic communication originated with the influential cheap talk model, which precedes the Bayesian persuasion model by three decades. This model describes an interaction between two agents: sender and receiver. The sender knows some state of the world which the receiver does not know, and tries to influence the receiver's action by communicating a cheap talk message to the receiver. This paper initiates the systematic algorithmic study of cheap talk in a finite environment (i.e., a finite number of states and receiver's possible actions). We first prove that approximating the sender-optimal or the welfare-maximizing cheap talk equilibrium up to a certain additive constant or multiplicative factor is NP-hard. We further prove that deciding whether there exists an equilibrium in which the receiver gets utility higher than the trivial utility he can guarantee is NP-hard. Fortunately, we identify two naturally-restricted cases that admit efficient algorithms for finding a sender-optimal equilibrium - a constant number of states or a receiver having only two actions.

Computer Science and Game Theory

What problem does this paper attempt to address?

The problem that this paper attempts to solve is about the algorithmic complexity research of the "cheap talk" model. Specifically, the authors focus on how to systematically study the algorithmic aspects of cheap talk in a finite environment (that is, a finite number of states and the number of actions that receivers may take).

### Core Problems of the Paper

1. **Computational Complexity**:
- The authors prove that approximately computing the optimal utility of the sender or the cheap - talk equilibrium that maximizes social welfare is NP - hard within certain additive constants or multiplicative factors.
- Deciding whether there exists an equilibrium where the receiver's utility is higher than the utility it can guarantee in the uninformative case is also NP - hard.

2. **Efficient Algorithms in Special Cases**:
- Although the problem in general cases is NP - hard, the authors identify two naturally restricted cases where the sender - optimal equilibrium can be found using efficient algorithms:
- When the number of states is a constant.
- When the receiver has only two actions.

### Main Contributions of the Paper

- **Theoretical Contribution**: By introducing new technical means, a variant of the 3SAT problem is reduced to the cheap talk model, thus proving the NP - hardness of computing the sender - optimal cheap talk equilibrium.
- **Algorithmic Contribution**: For specific cases (such as when the number of states is a constant or the receiver has only two actions), polynomial - time algorithms are proposed to compute the sender - optimal equilibrium.

### Formula Representation

Some of the key formulas mentioned in the paper include:

- **Bayesian Update Formula**:
\[
p_\sigma(\omega)=\frac{\pi(\sigma\mid\omega)\cdot\mu(\omega)}{\sum_{\omega'\in\Omega}\pi(\sigma\mid\omega')\cdot\mu(\omega')}
\]
where \(p_\sigma(\omega)\) is the posterior probability of state \(\omega\) after receiving signal \(\sigma\).

- **Receiver's Expected Utility**:
\[
E_{\omega\sim p_\sigma}[u_R(\omega, a)]=\sum_{\omega\in\Omega}p_\sigma(\omega)u_R(\omega, a)
\]

- **Sender's Best - Response Signal Set**:
\[
BS(\omega, s)=\left\{\sigma\in\text{supp}(\pi):E_{a\sim s[\sigma]}u_S(\omega, a)\geq E_{a\sim s[\sigma']}u_S(\omega, a),\forall\sigma'\in\text{supp}(\pi)\right\}
\]

These formulas help in understanding the strategy selection and utility calculation in the cheap talk model.

### Summary

This paper mainly explores the algorithmic complexity of the cheap talk model and points out that in general cases, computing the sender - optimal equilibrium is NP - hard. However, for some problems under special conditions, there are still efficient algorithmic solutions.