Emergent Relational Order in LLM Agent Societies: From Collective Affect to Authority Stratification

ACL ARR 2026 January Submission10842 Authors

Keywords: LLM-based Multi-Agent Simulation，Differential Order Pattern，Emergent Authority，Fei Xiaotong，Division of Labor and Relational Reciprocity

Abstract:

Fei Xiaotong’s Differential Order Pattern characterizes rural society as egocentric and relationally graded, with cooperation attenuating over social distance, yet lacks mechanistic operationalization. Existing LLM-based social simulations primarily address short-term coordination, leaving the emergence of authority, affect, and division of labor underexplored.
We propose CAREB-MAS, a multi-agent framework modeling general social cognition rather than culture-specific rules. Agents reason through an emotion–ethics–belief chain, maintain dynamically evolving egocentric identities, and act under relationally graded utilities, while the macro environment specifies only individual production, preference-based allocation, and minimal interaction protocols.
Across long-horizon simulations, agents spontaneously reproduce five core Differential Order phenomena: stable labor specialization, guanxi-based reciprocity, relationally decaying cooperation, emergent relational authority
, and clan-based center–periphery stratification. These patterns systematically vary with production structure, bridging mechanical and organic modes of social integration.
Together, these results recast Differential Order as a structure-sensitive emergent outcome of general social mechanisms, demonstrating LLM-based multi-agent simulation as a generative testbed for studying social structure and change.

Paper Type: Long

Research Area: Computational Social Science, Cultural Analytics, and NLP for Social Good

Research Area Keywords: human behavior analysis

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Data analysis

Languages Studied: English

Submission Number: 10842