Incentive Stackelberg Mean-payoff Games

TL;DR

This paper introduces incentive equilibria in multi-player mean-payoff games, allowing a leader to influence followers through payoffs, which can improve the leader's payoff and has implications for game analysis complexity.

Contribution

It defines incentive equilibria, proves their existence, analyzes their computational complexity, and provides algorithms with experimental validation.

Findings

Incentive equilibria always exist in mean-payoff games.

Constructing incentive equilibria is NP-complete.

When players are fixed, complexity matches two-player mean-payoff games.

Abstract

We introduce and study incentive equilibria for multi-player meanpayoff games. Incentive equilibria generalise well-studied solution concepts such as Nash equilibria and leader equilibria (also known as Stackelberg equilibria). Recall that a strategy profile is a Nash equilibrium if no player can improve his payoff by changing his strategy unilaterally. In the setting of incentive and leader equilibria, there is a distinguished player called the leader who can assign strategies to all other players, referred to as her followers. A strategy profile is a leader strategy profile if no player, except for the leader, can improve his payoff by changing his strategy unilaterally, and a leader equilibrium is a leader strategy profile with a maximal return for the leader. In the proposed case of incentive equilibria, the leader can additionally influence the behaviour of her followers by transferring parts of her payoff to her followers. The ability to incentivise her followers provides the leader with more freedom in selecting strategy profiles, and we show that this can indeed improve the payoff for the leader in such games. The key fundamental result of the paper is the existence of incentive equilibria in mean-payoff games. We further show that the decision problem related to constructing incentive equilibria is NP-complete. On a positive note, we show that, when the number of players is fixed, the complexity of the problem falls in the same class as two-player mean-payoff games. We also present an implementation of the proposed algorithms, and discuss experimental results that demonstrate the feasibility of the analysis of medium sized games.