The Hidden Strength of Disagreement: Unraveling the Consensus-Diversity Tradeoff in Adaptive Multi-Agent Systems

TL;DR

This paper explores how implicit consensus through in-context learning in multi-agent systems maintains diversity, enhancing adaptability, exploration, and robustness in dynamic environments compared to explicit coordination methods.

Contribution

It formalizes the consensus-diversity tradeoff and demonstrates the effectiveness of implicit consensus in multi-agent systems through theoretical analysis and empirical experiments.

Findings

Implicit consensus improves exploration and robustness.

Partial diversity enhances adaptability in dynamic scenarios.

Emergent coordination via in-context learning is beneficial.

Abstract

Consensus formation is pivotal in multi-agent systems (MAS), balancing collective coherence with individual diversity. Conventional LLM-based MAS primarily rely on explicit coordination, e.g., prompts or voting, risking premature homogenization. We argue that implicit consensus, where agents exchange information yet independently form decisions via in-context learning, can be more effective in dynamic environments that require long-horizon adaptability. By retaining partial diversity, systems can better explore novel strategies and cope with external shocks. We formalize a consensus-diversity tradeoff, showing conditions where implicit methods outperform explicit ones. Experiments on three scenarios -- Dynamic Disaster Response, Information Spread and Manipulation, and Dynamic Public-Goods Provision -- confirm partial deviation from group norms boosts exploration, robustness, and performance. We highlight emergent coordination via in-context learning, underscoring the value of preserving diversity for resilient decision-making.