More at Stake: How Payoff and Language Shape LLM Agent Strategies in Cooperation Dilemmas

2026-01-27

AI Summary

This study investigates how large language models (LLMs) adjust their cooperative strategies in the iterated prisoner’s dilemma under the influence of payoff incentives and linguistic context. By designing a multilingual game-theoretic experiment with scaled payoffs and employing a supervised classifier to map LLM behaviors to canonical repeated-game strategies, the work systematically analyzes strategic patterns across models and languages. It reveals, for the first time, that the framing effect of language can rival or even surpass architectural differences in shaping LLMs’ propensity to cooperate. The authors propose a unified auditing framework grounded in classical game theory and demonstrate that LLMs consistently exhibit incentive-sensitive conditional cooperation, with language significantly modulating their strategic choices—providing empirical foundations for safe multi-agent systems and AI governance.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Game Theory and Economic Paradigms: Cooperative Game Theory

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• 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

Abstract

As LLMs increasingly act as autonomous agents in interactive and multi-agent settings, understanding their strategic behavior is critical for safety, coordination, and AI-driven social and economic systems. We investigate how payoff magnitude and linguistic context shape LLM strategies in repeated social dilemmas, using a payoff-scaled Prisoner's Dilemma to isolate sensitivity to incentive strength. Across models and languages, we observe consistent behavioral patterns, including incentive-sensitive conditional strategies and cross-linguistic divergence. To interpret these dynamics, we train supervised classifiers on canonical repeated-game strategies and apply them to LLM decisions, revealing systematic, model- and language-dependent behavioral intentions, with linguistic framing sometimes matching or exceeding architectural effects. Our results provide a unified framework for auditing LLMs as strategic agents and highlight cooperation biases with direct implications for AI governance and multi-agent system design.

Problem

• LLM agents
• cooperation dilemmas
• payoff magnitude
• linguistic context
• strategic behavior

Innovation

• payoff scaling
• linguistic framing
• strategic behavior
• LLM agents
• cooperation dilemmas