Aligned Agents, Biased Swarm: Measuring Bias Amplification in Multi-Agent Systems

TL;DR

This study investigates how multi-agent system structures and feedback loops can unintentionally amplify biases, revealing that increased complexity does not necessarily improve ethical robustness.

Contribution

Introduces Discrim-Eval-Open benchmark and provides empirical evidence that complex MAS architectures can exacerbate bias amplification rather than reduce it.

Findings

Bias can be amplified through structured workflows acting as echo chambers.

Architectural sophistication often worsens bias rather than mitigates it.

Objective context injection can accelerate polarization.

Abstract

While Multi-Agent Systems (MAS) are increasingly deployed for complex workflows, their emergent properties-particularly the accumulation of bias-remain poorly understood. Because real-world MAS are too complex to analyze entirely, evaluating their ethical robustness requires first isolating their foundational mechanics. In this work, we conduct a baseline empirical study investigating how basic MAS topologies and feedback loops influence prejudice. Contrary to the assumption that multi-agent collaboration naturally dilutes bias, we hypothesize that structured workflows act as echo chambers, amplifying minor stochastic biases into systemic polarization. To evaluate this, we introduce Discrim-Eval-Open, an open-ended benchmark that bypasses individual model neutrality through forced comparative judgments across demographic groups. Analyzing bias cascades across various structures reveals that architectural sophistication frequently exacerbates bias rather than mitigating it. We observe systemic amplification even when isolated agents operate neutrally, and identify a 'Trigger Vulnerability' where injecting purely objective context drastically accelerates polarization. By stripping away advanced swarm complexity to study foundational dynamics, we establish a crucial baseline: structural complexity does not guarantee ethical robustness. Our code is available at https://github.com/weizhihao1/MAS-Bias.