arXiv cs.AI· Patrick Taillandier, Jean Daniel Zucker, Arnaud Grignard, Benoit Gaudou, Nghi Quang Huynh, Haojia Kong, Alexis Drogoul·· 6 天前AI 评分24
从流畅性谬误到微观-宏观效度鸿沟：LLM 在社会仿真中的机遇与陷阱
From the Fluency Fallacy to the Micro-to-Macro Validity Gap: Opportunities and Pitfalls of LLMs in Social Simulation
AI 导读
一篇批判性综述指出，LLM 作为社会仿真中智能体的认知或决策组件时，其幻觉、随机不一致、表征偏见与对齐效应等微观局限会传播、放大或收缩为智能体层面的失真（如人格趋同）和宏观风险（如人工共识与行为同质化）。
正文 · 原文
Abstract:The integration of Large Language Models (LLMs) into social simulation has generated enthusiasm but also raises methodological and epistemological challenges. This critical review examines LLMs as cognitive or decision-making components of simulated agents from a computational social science perspective. We argue that their behavioral and epistemic limitations can become systemic vulnerabilities when scaled to multi-agent societies. We map the evolving landscape of LLM-driven platforms, from narrative sandboxes to population-scale simulations, and develop a framework for analyzing the Micro-to-Macro Validity Gap: the propagation, amplification, or contraction of micro-level limitations-including hallucinations, stochastic inconsistency, representational biases, and alignment effects-into agent-level distortions (e.g., persona convergence) and macro-level risks (e.g., artificial consensus and behavioral homogenization). We also highlight evaluative vulnerabilities such as the Fluency Fallacy and automation bias. We identify contexts where LLM-based agents offer operational value, including serious games, participatory environments, and exploratory modeling, distinguishing these uses from confirmatory research with unconstrained models and precise social forecasting. Finally, we examine theory-driven hybrid architectures embedding LLMs in explicit, mechanistic, and reproducible Agent-Based Modeling (ABM) frameworks. While promising, these architectures do not guarantee epistemic control: validity requires multi-level evaluation of environmental dynamics, individual behavior, cross-level interactions, and aggregate outcomes. They also remain vulnerable to physics washing.
| Subjects: | Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2507.19364 [cs.AI] |
| (or arXiv:2507.19364v4 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2507.19364 arXiv-issued DOI via DataCite |
Submission history
From: Patrick Taillandier [view email]
[v1]
Fri, 25 Jul 2025 15:15:35 UTC (33 KB)
[v2]
Fri, 27 Feb 2026 17:03:14 UTC (54 KB)
[v3]
Mon, 7 Sep 2026 12:14:34 UTC (1,558 KB)
[v4]
Fri, 9 Oct 2026 10:22:50 UTC (65 KB)
来源：arXiv cs.AI · arxiv.org