Strategies of cooperation and defection in five large language models

2026-01-14

AI Summary

This study systematically evaluates the strategic choices of five leading large language models in the repeated Prisoner’s Dilemma without explicit prompting, examining whether they can spontaneously generate rational cooperation or defection in social decisions involving others’ welfare. Through a combination of game-theoretic experimental design, strategy classification, parameter perturbation, and model-versus-model tournaments—integrated with evolutionary game theory and human behavioral benchmarks—the work presents the first multidimensional analysis of model adaptability and consistency across varying game parameters and framing conditions. Results reveal that while most models perform adequately under certain settings, none maintain consistent strategies across all configurations, exposing a fundamental limitation in current large language models’ capacity for reciprocal cooperation.

Technology Category

• Game Theory and Economic Paradigms: Cooperative Game Theory
• Natural Language Processing: (Large) Language Models
• Multiagent Systems: Mechanism Design

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Large language models (LLMs) are increasingly deployed to support human decision-making. This use of LLMs has concerning implications, especially when their prescriptions affect the welfare of others. To gauge how LLMs make social decisions, we explore whether five leading models produce sensible strategies in the repeated prisoner's dilemma, which is the main metaphor of reciprocal cooperation. First, we measure the propensity of LLMs to cooperate in a neutral setting, without using language reminiscent of how this game is usually presented. We record to what extent LLMs implement Nash equilibria or other well-known strategy classes. Thereafter, we explore how LLMs adapt their strategies to changes in parameter values. We vary the game's continuation probability, the payoff values, and whether the total number of rounds is commonly known. We also study the effect of different framings. In each case, we test whether the adaptations of the LLMs are in line with basic intuition, theoretical predictions of evolutionary game theory, and experimental evidence from human participants. While all LLMs perform well in many of the tasks, none of them exhibit full consistency over all tasks. We also conduct tournaments between the inferred LLM strategies and study direct interaction between LLMs in games over ten rounds with a known or unknown last round. Our experiments shed light on how current LLMs instantiate reciprocal cooperation.

Problem

• large language models
• prisoner's dilemma
• cooperation
• social decision-making
• reciprocal cooperation

Innovation

• large language models
• repeated prisoner's dilemma
• reciprocal cooperation
• evolutionary game theory
• strategy adaptation