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

2026

Open publication workspace · Sign in to read the full PDF.

AI-generated summary

This paper introduces a statistical mechanics framework to predict and understand the collective behavior of AI agents, revealing predictable dynamical laws governing their interactions.

* Explores how over 10,000 language-model agents interact and revise opinions on objective and subjective questions.
* Develops a statistical mechanics model, analogous to the Ising model, to explain observed collective dynamics like polarization and consensus.
* Demonstrates the model's ability to predict individual and group trajectories, generalize to unseen scenarios, and interpret the underlying mechanics of agent interaction.

AI agents, collective behavior, statistical mechanics, opinion dynamics, multi-agent systems

Check the original publication for accuracy and context.