When Is Collective Intelligence a Lottery? Multi-Agent Scaling Laws for Memetic Drift in LLMs

Researchers introduce Quantized Simplex Gossip (QSG) model to explain how multi-agent LLM systems reach consensus through 'memetic drift' - where arbitrary choices compound into collective agreement. The study reveals scaling laws for when collective intelligence operates like a lottery versus amplifying weak biases, providing a framework for understanding AI system behavior in consequential decision-making.

• →Multi-agent LLM systems can reach consensus even when individual agents have no initial preferences, through a process called memetic drift.
• →The QSG model shows how one agent's arbitrary choice becomes evidence for other agents, compounding toward agreement through mutual in-context learning.
• →There's a crossover point from drift-dominated regimes (lottery-like outcomes) to selection regimes where weak biases are amplified.
• →Scaling laws predict polarization based on population size, communication bandwidth, adaptation rate, and agents' internal uncertainty.
• →This research provides a framework for understanding how AI systems form collective representations in high-stakes decision-making contexts.