Beyond Nash Equilibrium: Bounded Rationality of LLMs and humans in Strategic Decision-making

2025-06-11

AI Summary

This study investigates whether large language models (LLMs) exhibit human-like bounded rationality in strategic games (e.g., Rock-Paper-Scissors, Prisoner’s Dilemma). Adopting a behavioral game-theoretic paradigm, we conduct multi-round human–model and model–model contests, integrating behavioral trajectory analysis with reasoning-path tracing to systematically compare decision dynamics. Our key findings are threefold: First, LLMs partially replicate human heuristics—such as outcome-driven strategy switching and enhanced cooperation under repeated interaction—but apply them rigidly and with markedly weaker environmental adaptability. Second, LLMs exhibit fundamental limitations in opponent modeling and contextual awareness, leading to mechanistic rather than adaptive strategy adjustments. Third, they fail to stably converge to effective strategies in dynamic, interactive settings. Collectively, results indicate that current LLMs possess only superficial bounded rationality, lacking genuine adaptive rationality essential for robust strategic reasoning.

Technology Category

• Game Theory and Economic Paradigms: Behavioral Game Theory
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

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
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large language models are increasingly used in strategic decision-making settings, yet evidence shows that, like humans, they often deviate from full rationality. In this study, we compare LLMs and humans using experimental paradigms directly adapted from behavioral game-theory research. We focus on two well-studied strategic games, Rock-Paper-Scissors and the Prisoner's Dilemma, which are well known for revealing systematic departures from rational play in human subjects. By placing LLMs in identical experimental conditions, we evaluate whether their behaviors exhibit the bounded rationality characteristic of humans. Our findings show that LLMs reproduce familiar human heuristics, such as outcome-based strategy switching and increased cooperation when future interaction is possible, but they apply these rules more rigidly and demonstrate weaker sensitivity to the dynamic changes in the game environment. Model-level analyses reveal distinctive architectural signatures in strategic behavior, and even reasoning models sometimes struggle to find effective strategies in adaptive situations. These results indicate that current LLMs capture only a partial form of human-like bounded rationality and highlight the need for training methods that encourage flexible opponent modeling and stronger context awareness.

Problem

• Compare LLMs and humans in strategic decision-making deviations
• Assess bounded rationality in LLMs using game-theory paradigms
• Identify gaps in LLMs' human-like flexible strategic reasoning

Innovation

• Compare LLMs and humans in strategic games
• Evaluate bounded rationality in LLMs
• Highlight need for flexible opponent modeling