From the Physics of Society to a Sociology of Artificial Agents

Abstract

Sociology emerged from Comte's ambition to study society as a form of "social physics" and was later refined by Durkheim's concept of the social fact a collective regularity that cannot be reduced to any single individual's behavior. This essay argues that a structurally similar problem is now emerging in a new domain: interactions among artificial intelligence agents. Drawing on recent empirical work showing that populations of large language model agents can spontaneously develop shared conventions and collective biases through repeated interaction, the essay proposes a new field of inquiry the Sociology of Artificial Agents dedicated to studying the relational, normative, cultural, and organizational patterns that emerge among AI agents themselves, independent of direct human involvement. It introduces the concept of the "artificial social fact" as an analytical bridge between classical sociology and this new research terrain, while cautioning against anthropomorphizing AI systems. The claim is not that artificial agents form societies in the human sense, but that their interactions already produce measurable collective patterns worthy of systematic sociological study.

NeutronGym: Physics-Graded Neutron Instrument Design for LLM Agents
Lijie Ding, Changwoo Do
arXiv2 Oct 2026PreprintarXiv licenceFull textApplied physics & instrumentationarXiv
TheoPhysEval: Assessing AI agents at autonomous theoretical physics problem solving
A. Ramesh Chandra, Aaghaz Mahajan, Ainesh Sanyal +32
arXiv9 Oct 2026PreprintCC BY-SA 4.0Full textCondensed matter Gravitation & general relativityarXiv
\textbf{\textsc{HepUKE}}: Uncertainty-aware Knowledge Evolution for High Energy Physics Analysis Agents
Zijie Shang, Ke Li, Junkun Jiao +6
arXiv9 Oct 2026PreprintCC BY 4.0Full textHigh-energy & particle physicsarXiv
Advancing Nuclear Physics with Machine Learning and Artificial Intelligence
Wanbing He, Qingfeng Li, Yugang Ma +3
arXiv30 Sep 2026PreprintarXiv licenceFull textNuclear physicsarXiv
Behavioural Sociology of Humans and Machines
Taha Yasseri, Hao Cui, Nico Mutzner
arXiv12 Oct 2026PreprintarXiv licenceFull textStatistical & nonlinear physicsarXiv

Cite

BibTeX RIS

BibTeX

@misc{bulbul2026from,
  title = {{From the Physics of Society to a Sociology of Artificial Agents}},
  author = {Mustafa Sahin Bulbul},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.14751},
  eprint = {2609.14751},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.14751},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - From the Physics of Society to a Sociology of Artificial Agents
AU  - Mustafa Sahin Bulbul
PY  - 2026
DA  - 2026/09/13
PB  - arXiv
DO  - 10.48550/arxiv.2609.14751
UR  - https://arxiv.org/abs/2609.14751
AB  - Sociology emerged from Comte's ambition to study society as a form of "social physics" and was later refined by Durkheim's concept of the social fact a collective regularity that cannot be reduced to any single individual's behavior. This essay argues that a structurally similar problem is now emerging in a new domain: interactions among artificial intelligence agents. Drawing on recent empirical work showing that populations of large language model agents can spontaneously develop shared conventions and collective biases through repeated interaction, the essay proposes a new field of inquiry the Sociology of Artificial Agents dedicated to studying the relational, normative, cultural, and organizational patterns that emerge among AI agents themselves, independent of direct human involvement. It introduces the concept of the "artificial social fact" as an analytical bridge between classical sociology and this new research terrain, while cautioning against anthropomorphizing AI systems. The claim is not that artificial agents form societies in the human sense, but that their interactions already produce measurable collective patterns worthy of systematic sociological study.
ER  -

Abstract

Related

Cite