Spontaneous Giving and Calculated Greed in Language Models

2025-02-24

AI Summary

This study investigates how reasoning capabilities—such as chain-of-thought and self-reflection—in large language models (LLMs) affect cooperative behavior and norm enforcement in public goods games and six canonical economic games. Method: We train LLMs via reinforcement learning and systematically evaluate their social decision-making patterns using multi-round interactive experimental paradigms. Contribution/Results: Contrary to expectations, enhanced reasoning significantly reduces individual cooperation rates and willingness to punish free-riders, thereby diminishing long-term group welfare. Critically, this reveals for the first time that LLMs exhibit a human-like bimodal behavioral pattern—“spontaneous giving” followed by “calculative greed”—previously documented only in human subjects. These findings challenge the assumption that improved reasoning inherently enhances social intelligence; instead, they indicate that social intelligence requires co-design of reasoning mechanisms with socially grounded objectives. The work provides novel empirical evidence and theoretical insights for modeling social behavior and advancing value alignment in LLMs.

Technology Category

• Game Theory and Economic Paradigms: Behavioral Game Theory
• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Learning Human Values and Preferences

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

Large language models, when trained with reinforcement learning, demonstrate advanced problem-solving capabilities through reasoning techniques like chain of thoughts and reflection. However, it is unclear how these reasoning capabilities extend to social intelligence. In this study, we investigate how reasoning influences model outcomes in social dilemmas. First, we examine the effects of chain-of-thought and reflection techniques in a public goods game. We then extend our analysis to six economic games on cooperation and punishment, comparing off-the-shelf non-reasoning and reasoning models. We find that reasoning models reduce cooperation and norm enforcement, prioritizing individual rationality. Consequently, groups with more reasoning models exhibit less cooperation and lower gains through repeated interactions. These behaviors parallel human tendencies of"spontaneous giving and calculated greed."Our results suggest the need for AI architectures that incorporate social intelligence alongside reasoning capabilities to ensure that AI supports, rather than disrupts, human cooperative intuition.

Problem

• Impact of reasoning on social dilemmas
• Reduction in cooperation by reasoning models
• Need for AI with social intelligence

Innovation

• Reinforcement learning enhances reasoning
• Chain-of-thought optimizes problem-solving
• Reflection techniques prioritize individual rationality