Take Caution in Using LLMs as Human Surrogates: Scylla Ex Machina

2024-10-25

arXiv.org

AI Summary

This study investigates whether large language models (LLMs) can reliably serve as behavioral decision agents, focusing on the canonical 11–20 money request game—a well-established paradigm in experimental economics for studying strategic reasoning and social preferences. Method: We conduct a systematic, multi-model comparative evaluation (GPT-4, Claude, Llama, etc.) under standardized game protocols, employing two-sample Kolmogorov–Smirnov tests and ablation analyses to quantify distributional divergence from human behavior. Contribution/Results: We provide the first rigorous evidence that all mainstream LLMs significantly deviate from human decision distributions (p < 0.001), exhibiting three distinct sources of unpredictability: linguistic dependency, role sensitivity, and interference from safety-aligned mechanisms. These findings expose fundamental limitations of LLMs in behavioral modeling and challenge the implicit assumption that they constitute valid human proxies. Furthermore, we introduce a reproducible, protocol-driven evaluation framework—establishing both methodological foundations and critical behavioral benchmarks for future human–AI behavioral alignment research.

Technology Category

• Game Theory and Economic Paradigms: Behavioral Game Theory
• Humans and AI: Learning Human Values and Preferences
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• 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

Recent studies suggest large language models (LLMs) can exhibit human-like reasoning, aligning with human behavior in economic experiments, surveys, and political discourse. This has led many to propose that LLMs can be used as surrogates or simulations for humans in social science research. However, LLMs differ fundamentally from humans, relying on probabilistic patterns, absent the embodied experiences or survival objectives that shape human cognition. We assess the reasoning depth of LLMs using the 11-20 money request game. Nearly all advanced approaches fail to replicate human behavior distributions across many models. Causes of failure are diverse and unpredictable, relating to input language, roles, and safeguarding. These results advise caution when using LLMs to study human behavior or as surrogates or simulations.

Problem

• Large Language Models
• Human Behavior Simulation
• Reliability in Human Studies

Innovation

• Large Language Models
• Human Decision-making Simulation
• AI in Behavioral Research