Super-additive Cooperation in Language Model Agents

2025-08-21

AI Summary

This study investigates the evolution of cooperative behavior among large language model (LLM) agents in complex social settings characterized by intra-group cooperation and inter-group competition. We propose a two-level game-theoretic framework—“intra-group cooperation vs. inter-group competition”—and simulate multi-round repeated prisoner’s dilemmas via virtual tournaments to quantitatively assess agents’ cooperation propensity in both one-shot and long-term interactions. Results show that inter-group competition significantly increases initial one-shot cooperation rates (+23.6%), challenging the conventional assumption that competition inherently undermines cooperation; this effect is mediated by competition-induced group identity formation and reputation management mechanisms. To our knowledge, this is the first systematic empirical validation of the positive moderating role of inter-group competition on LLM agent cooperation. The findings provide theoretical grounding and a reproducible technical pathway for designing trustworthy, collaborative multi-agent systems. Code is publicly available.

Technology Category

• Multiagent Systems: Mechanism Design
• Game Theory and Economic Paradigms: Cooperative Game Theory
• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction

Application Category

• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

With the prospect of autonomous artificial intelligence (AI) agents, studying their tendency for cooperative behavior becomes an increasingly relevant topic. This study is inspired by the super-additive cooperation theory, where the combined effects of repeated interactions and inter-group rivalry have been argued to be the cause for cooperative tendencies found in humans. We devised a virtual tournament where language model agents, grouped into teams, face each other in a Prisoner's Dilemma game. By simulating both internal team dynamics and external competition, we discovered that this blend substantially boosts both overall and initial, one-shot cooperation levels (the tendency to cooperate in one-off interactions). This research provides a novel framework for large language models to strategize and act in complex social scenarios and offers evidence for how intergroup competition can, counter-intuitively, result in more cooperative behavior. These insights are crucial for designing future multi-agent AI systems that can effectively work together and better align with human values. Source code is available at https://github.com/pippot/Superadditive-cooperation-LLMs.

Problem

• Investigating cooperative behavior in AI agents using game theory
• Studying how intergroup competition affects cooperation in language models
• Designing multi-agent systems for better alignment with human values

Innovation

• Simulated team dynamics and external competition
• Used Prisoner's Dilemma game in virtual tournament
• Combined repeated interactions with inter-group rivalry