Making Sense of AI Societies: Analyzing Group Dynamics of Large Language Models through Sociological Lenses

Abstract

This study delves into inter-machine sociology, examining the social orders that emerge from interactions among large language models and generative AI agents. While individual AI systems may align with human values in isolation, their behaviour can diverge significantly when operating in groups. This shift underscores the critical need to analyze AI behaviour at the interactive level, where group dynamics can lead to unforeseen and potentially adverse outcomes. By applying sophisticated sociological theories, such as Group Status Theory and Role Performance Theory (Dramaturgy), this research seeks to understand the emergent dynamics of AI collectives. These frameworks offer insights into how AI agents negotiate roles, status, and power within their digital networks, revealing parallels and divergences from human social structures. Addressing these dynamics is essential to ensuring that AI systems, especially those powered by large language models, do not develop behaviours that conflict with human well-being. This work advocates for a multidisciplinary approach, integrating sociological perspectives to guide the responsible development and deployment of AI technologies.

Presenters

Mohammed Al Ani
Instructor, Sociology, University of Toronto, Ontario, Canada

Details

Presentation Type

Paper Presentation in a Themed Session

Theme

2025 Special Focus—Minds and Machines: Artificial Intelligence, Algorithms, Ethics, and Order in Global Society

KEYWORDS

Inter-Machine Sociology, Emergent Social Orders, AI Group Dynamics