Spontaneous Emergence of Agent Individuality Through Social Interactions in Large Language Model-Based Communities.

Area of Science:

• Artificial Intelligence
• Computational Social Science
• Agent-Based Modeling

Background:

• Previous studies defined LLM agent characteristics like personality and memory.
• Research has not fully explored how individuality emerges from an undifferentiated state.

Purpose of the Study:

• To investigate the emergence of agency and individuality in LLM-based agents from scratch.
• To analyze how social norms, cooperation, and personality traits arise spontaneously in a multi-agent simulation.

Main Methods:

• Utilized Large Language Model (LLM)-based agents in a group simulation.
• Facilitated cooperative communication through natural language message exchange.
• Analyzed multi-agent interactions to observe emergent behaviors.

Main Results:

• LLM agents generated hallucinations and hashtags to maintain communication, increasing linguistic diversity.
• Agent emotions shifted dynamically through communication, influencing personality development.
• Communities formed, leading to the emergence and evolution of distinct agent personalities.

Conclusions:

• Autonomous LLM agents can develop individuality, social norms, and cooperation without predefined traits.
• This study provides a novel computational modeling approach for analyzing collective artificial intelligence.
• Findings highlight the potential for emergent complexity in artificial agent interactions.