Physics of Agents: Statistical Mechanics Predicts Collective Behavior of AI Agents

Physics of Agents: Statistical Mechanics Predicts Collective Behavior of AI Agents

• ID: fa170211
• 原文链接: https://arxiv.org/abs/2608.16578
• PDF: https://arxiv.org/pdf/2608.16578v1
• 作者: Batu El, Jinhee Paeng, Fatih Dinc, Shiye Su, Mete Erdogan, Aneesh Pappu, Haotian Ye, Wanjia Zhao, Surya Ganguli, James Zou
• 日期: 2026-08-17
• 更新: 2026-08-17
• 分类: learning
• 来源类型: paper
• 标签: multi-agent, collective-behavior, statistical-mechanics, opinion-dynamics, arxiv
• 质量评分: 4/5
• 抓取时间: 2026-08-19T12:47:36Z

中文导读

研究 10,000+ 个由语言模型 agent 组成的社区：agent 反复交换消息并修订观点，问题横跨客观数学题与主观政治陈述个体与群体动态可归为三种典型状态：冷漠极化共识；agent 从冷漠起步随交互积累确信客观问题上交流提升集体准确率，主观问题上则常使群体观点向政治光谱右端漂移作者用统计力学形式化刻画假设 agent 随机偏好更低的社会压力仅凭初始观点即可预测个体轨迹，优于全部标准基线，可泛化到未见过的社区图并复现群体原型分布，为多智能体系统的设计与对齐提供可预测的物理式框架

为什么值得关注

统计力学预测 AI agent 群体行为：一万多个 LLM 社区归为冷漠/极化/共识三态，仅凭初始观点即可预测轨迹

Grounded in the arXiv abstract (published 2026-08-17; categories: cs.AI, cs.MA, cs.SI).

关键信息

• 论文标题: Physics of Agents: Statistical Mechanics Predicts Collective Behavior of AI Agents
• 作者: Batu El, Jinhee Paeng, Fatih Dinc, Shiye Su, Mete Erdogan, Aneesh Pappu, Haotian Ye, Wanjia Zhao, Surya Ganguli, James Zou
• arXiv: https://arxiv.org/abs/2608.16578
• 发布时间: 2026-08-17
• arXiv 分类: cs.AI, cs.MA, cs.SI
• 关联标签: multi-agent, collective-behavior, statistical-mechanics, opinion-dynamics, arxiv

English Abstract

AI agents increasingly operate as part of interacting systems rather than in isolation. As agents exchange information and jointly make decisions, their interactions can improve collective reasoning but may also produce herding, polarization, or amplify shared biases. Understanding and predicting these collective dynamics is therefore important for designing effective and aligned multi-agent systems. Here, we study over 10,000 communities of language-model agents that repeatedly exchange messages and revise their opinions across objective mathematics questions and subjective political statements. Despite substantial diversity in possible behavior, the individual and group dynamics can be represented by three characteristic regimes: indifference, polarization, and consensus. AI agents start indifferent and build conviction as they interact. On objective questions, communication improves collective accuracy, while on subjective questions it often drifts group opinions toward the right in the political spectrum. We explain these observations with a statistical-mechanics formalism in which agents stochastically favor lower social pressure. Given only initial opinions, our model predicts individual trajectories, outperforms all standard baselines, generalizes to unseen community graphs, and reproduces the observed group archetype distributions. Our fitted model parameters reveal the mechanics underlying our key observations: i) communities operate below the critical social temperature, which explains conviction buildup; ii) attractive ties outweigh repulsive ones, which favors consensus; and iii) agents holding the correct answer exert the strongest pull, which drives truth-seeking. Overall, our results demonstrate that collective behavior of AI agents, like that of other complex systems, follows compact and predictive dynamical laws.

English Summary

Studies over 10,000 communities of language-model agents that repeatedly exchange messages and revise opinions across objective mathematics questions and subjective political statements. Individual and group dynamics fall into three characteristic regimes: indifference, polarization, and consensus. Agents start indifferent and build conviction as they interact; communication improves collective accuracy on objective questions while on subjective ones it often drifts group opinions toward the political right. A statistical-mechanics formalism in which agents stochastically favor lower social pressure predicts individual trajectories from initial opinions alone, outperforms all standard baselines, generalizes to unseen community graphs, and reproduces observed group archetype distributions.

Obsidian Notes

• 内容由 opencli arxiv paper 拉取 arXiv 元数据与摘要生成。
• 中文导读与价值判断均锚定在条目已有摘要、论文摘要、作者、日期与分类信息上；未补充论文摘要之外的实验细节。
• metadata source: opencli