Network Formation and Dynamics among Multi-LLMs
Marios Papachristou111Department of Information Systems, W.P. Carey School of Business, Arizona State University, Tempe, AZ, USA and Department of Computer Science, Cornell University, Ithaca, NY, USA. Supported by a scholarship from the Onassis Foundation (Scholarship ID: F ZT 056-1/2023-2024), and in part by a Simons Investigator Award, a Vannevar Bush Faculty Fellowship, AFOSR grant FA9550-19-1-0183, a Simons Collaboration grant, and a grant from the MacArthur Foundation. Yuan Yuan222Graduate School of Management, University of California Davis, Davis, CA, USA.
Abstract
Social networks profoundly influence how humans form opinions, exchange information, and organize collectively. As large language models (LLMs) are increasingly embedded into social and professional environments, it is critical to understand whether their interactions approximate human-like network dynamics. We develop a framework to study the network formation behaviors of multiple LLM agents and benchmark them against human decisions. Across synthetic and real-world settings, including friendship, telecommunication, and employment networks, we find that LLMs consistently reproduce fundamental micro-level principles such as preferential attachment, triadic closure, and homophily, as well as macro-level properties including community structure and small-world effects. Importantly, the relative emphasis of these principles adapts to context: for example, LLMs favor homophily in friendship networks but heterophily in organizational settings, mirroring patterns of social mobility. A controlled human-subject survey confirms strong alignment between LLMs and human participants in link-formation decisions. These results establish that LLMs can serve as powerful tools for social simulatio
中文速览
大语言模型（LLMs）被越来越多地引入社交和职场场景，但它们在构建人际网络时的行为是否符合人类社会规律，此前几乎无人系统研究。研究团队搭建了一套多智能体仿真框架，让多个 LLM 代理在合成网络和真实社会网络（友谊网络、电信网络、企业协作网络）中独立做出"与谁建立连接"的决策，并与真实人类的选择进行对比。结果发现，LLMs 在微观层面稳定地再现了优先连接（preferential attachment）、三角闭合（triadic closure）和同质性（homophily）三大核心原则，在宏观层面也自发涌现出社群结构和小世界效应；更关键的是，这些倾向会随情境自适应调整——在友谊网络中偏好同质连接，在企业网络中则倾向于跨层级连接，与人类社会流动模式高度吻合，受控人类被试调查也验证了二者的强一致性。这项研究表明，LLMs 具备成为社会仿真与合成数据生成强力工具的潜力，同时也提醒研究者必须正视 AI 深度嵌入人类网络后可能带来的偏见与公平性风险。
原文 arXiv:2402.10659;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2402.10659v7