CoopEval: Benchmarking Cooperation-Sustaining Mechanisms and LLM Agents in Social Dilemmas
For AI safety researchers, it provides the first comparative evaluation of cooperation mechanisms for LLM agents, addressing the critical problem of LLMs' tendency to defect in mixed-motive games.
This paper benchmarks game-theoretic mechanisms (repetition, reputation, mediation, contracts) for sustaining cooperation among LLM agents in social dilemmas, finding that contracting and mediation are most effective, while repetition fails with varying co-players. Evolutionary pressure enhances mechanism effectiveness.
It is increasingly important that LLM agents interact effectively and safely with other goal-pursuing agents, yet, recent works report the opposite trend: LLMs with stronger reasoning capabilities behave _less_ cooperatively in mixed-motive games such as the prisoner's dilemma and public goods settings. Indeed, our experiments show that recent models -- with or without reasoning enabled -- consistently defect in single-shot social dilemmas. To tackle this safety concern, we present the first comparative study of game-theoretic mechanisms that are designed to enable cooperative outcomes between rational agents _in equilibrium_. Across four social dilemmas testing distinct components of robust cooperation, we evaluate the following mechanisms: (1) repeating the game for many rounds, (2) reputation systems, (3) third-party mediators to delegate decision making to, and (4) contract agreements for outcome-conditional payments between players. Among our findings, we establish that contracting and mediation are most effective in achieving cooperative outcomes between capable LLM models, and that repetition-induced cooperation deteriorates drastically when co-players vary. Moreover, we demonstrate that these cooperation mechanisms become _more effective_ under evolutionary pressures to maximize individual payoffs.