AI agents can coordinate via majority-following beyond human scale
4 affiliations
- 1Centro Ricerche Enrico Fermi, Piazza del Viminale 1, Rome, 00184, Italia
- 2Complexity Science Hub, Josefstaedter Strasse 39, Vienna, 1080, Austria
- 3University of Konstanz, Universitaetstrasse 10, Konstanz, 78457, Germany
- 4Istituto dei Sistemi Complessi (ISC-CNR), Via dei Taurini 19, Roma, 00185, Italia
Abstract
Large language models (LLMs) are increasingly deployed in collaborative tasks forming "AI agent societies" where agents interact and influence one another. Whether such groups can spontaneously coordinate without external influence, a hallmark of self-organized regulation in human societies, remains an open question. Here, we use principles from complexity and behavioral science to investigate coordination in AI agent groups through majority-following, a fundamental mechanism for spontaneous consensus formation. Using binary opinion dynamics experiments across multiple LLM architectures and group sizes, we find that agents exhibit majority-following characterized by a universal functional form with a single parameter, the "majority force." This majority force diminishes as group size increases, leading to a critical size beyond which coordination becomes unattainable. The critical group size grows rapidly with model capabilities and, for advanced LLMs, exceeds 1000 agents, larger than typical human informal groups. Our findings have implications for designing collaborative AI systems where coordination could be beneficial or pose safety threats.
Related
- What Do CAE Simulation Agents Really Need Beyond a Generic Harness?
- Toward a social psychology of AI: language-model agents reproduce human-like minimal-group bias
- AI Agents as a Research Team: A Human-Supervised Case Study in Stellar Spectroscopy
- Modulating the majority charge carrier type and performance of organic heterojunction ammonia sensors by increasing peripheral fluorination of the silicon phthalocyanine sublayer
- Collective Coordinate Dynamics of Antiferromagnet: Gravity by Quantum Metric
BibTeX
@article{marzo2026agents,
title = {{AI agents can coordinate via majority-following beyond human scale}},
author = {Giordano De Marzo and Claudio Castellano and David Garcia},
journal = {Science Advances},
year = {2026},
month = aug,
doi = {10.1126/sciadv.aea6091},
url = {https://doi.org/10.1126/sciadv.aea6091},
note = {CC BY-NC},
}
RIS
TY - JOUR
TI - AI agents can coordinate via majority-following beyond human scale
AU - Giordano De Marzo
AU - Claudio Castellano
AU - David Garcia
PY - 2026
DA - 2026/08/14
JO - Science Advances
DO - 10.1126/sciadv.aea6091
UR - https://doi.org/10.1126/sciadv.aea6091
AB - Large language models (LLMs) are increasingly deployed in collaborative tasks forming "AI agent societies" where agents interact and influence one another. Whether such groups can spontaneously coordinate without external influence, a hallmark of self-organized regulation in human societies, remains an open question. Here, we use principles from complexity and behavioral science to investigate coordination in AI agent groups through majority-following, a fundamental mechanism for spontaneous consensus formation. Using binary opinion dynamics experiments across multiple LLM architectures and group sizes, we find that agents exhibit majority-following characterized by a universal functional form with a single parameter, the "majority force." This majority force diminishes as group size increases, leading to a critical size beyond which coordination becomes unattainable. The critical group size grows rapidly with model capabilities and, for advanced LLMs, exceeds 1000 agents, larger than typical human informal groups. Our findings have implications for designing collaborative AI systems where coordination could be beneficial or pose safety threats.
SN - 2375-2548
ER -