Multi-Agent Systems are Mixtures of Experts: Who Becomes an Influencer?
2026-05-25 • Multiagent Systems
Multiagent SystemsMachine Learning
AI summaryⓘ
The authors explore how groups of AI agents make decisions together, focusing not just on their individual answers but on how they share and influence each other’s opinions. They use a model called Friedkin-Johnsen opinion dynamics to understand stubbornness and influence during group discussions. The study finds that the way agents influence each other changes depending on the situation, making the group act like a mix of different experts. They also look at how agents decide who to listen to, using signs like how confident agents feel or how much their opinions already agree. This helps explain how groups can sometimes be better at solving problems than any one agent alone.
Multi-agent systemsLarge language modelsFriedkin-Johnsen modelOpinion dynamicsDeliberationInfluenceStubbornnessMixture of expertsAgent confidenceCollaborative decision making
Authors
Franka Bause, Jonas Niederle, Martin Pawelczyk, Rebekka Burkholz
Abstract
The effectiveness of multi-agent LLM deliberation depends not only on the agents' individual predictions, but also on how they communicate and collaborate. We study this mechanism through the lens of Friedkin-Johnsen (FJ) opinion dynamics, a tractable model for analyzing stubbornness, influence, and opinion change in multi-agent systems that captures empirically observed deliberation patterns. We show that the FJ parameters are input-dependent, turning multi-agent deliberation into a mixture of experts. This perspective implies that multi-agent systems can outperform single agents and static ensembles when routing reflects agent competence. Since competence is latent in practice, we analyze how influence is established through observable proxies: agents' self-assessed confidence, their perceived confidence, and initial alignment with other agents' views.