Do Agent Societies Develop Intellectual Elites? The Hidden Power Laws of Collective Cognition in LLM Multi-Agent Systems

Researchers conducted the first large-scale study of coordination dynamics in LLM multi-agent systems, analyzing over 1.5 million interactions to discover three fundamental laws governing collective AI cognition. The study found that coordination follows heavy-tailed cascades, concentrates into 'intellectual elites,' and produces more extreme events as systems scale, leading to the development of Deficit-Triggered Integration (DTI) to improve performance.

• →Multi-agent AI systems develop coordination patterns that follow predictable mathematical laws, including heavy-tailed cascades and preferential attachment.
• →Large AI agent societies naturally form 'intellectual elites' where certain agents become coordination hubs, similar to social network dynamics.
• →Scaling multi-agent systems produces increasingly frequent extreme events due to integration bottlenecks in reasoning processes.
• →The new Deficit-Triggered Integration (DTI) method improves multi-agent performance by selectively increasing integration during coordination imbalances.
• →The research establishes quantitative laws for collective AI cognition, providing a framework for understanding and improving scalable multi-agent intelligence.