Silence is Not Consensus: Disrupting Agreement Bias in Multi-Agent LLMs via Catfish Agent for Clinical Decision Making

2025-05-27

AI Summary

Multi-agent large language models (LLMs) in clinical decision-making often suffer from premature convergence due to “silent consensus,” leading to overlooked differential diagnoses. Method: We propose the role-specialized Catfish Agent, which actively injects structured skepticism via two synergistic mechanisms: complexity-aware dissent triggering and tone-calibrated critical generation. The framework integrates medical knowledge guidance, dynamic complexity assessment, and controllable-tone dissent generation. Contribution/Results: Evaluated on nine medical question-answering and three medical visual question-answering benchmarks, our approach consistently outperforms state-of-the-art single- and multi-agent baselines—including GPT-4o and DeepSeek-R1—achieving statistically significant gains in diagnostic accuracy. It markedly enhances the robustness and interpretability of clinical reasoning while mitigating groupthink tendencies.

Technology Category

• Multiagent Systems: Agreement, Argumentation & Negotiation
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Large language models (LLMs) have demonstrated strong potential in clinical question answering, with recent multi-agent frameworks further improving diagnostic accuracy via collaborative reasoning. However, we identify a recurring issue of Silent Agreement, where agents prematurely converge on diagnoses without sufficient critical analysis, particularly in complex or ambiguous cases. We present a new concept called Catfish Agent, a role-specialized LLM designed to inject structured dissent and counter silent agreement. Inspired by the ``catfish effect'' in organizational psychology, the Catfish Agent is designed to challenge emerging consensus to stimulate deeper reasoning. We formulate two mechanisms to encourage effective and context-aware interventions: (i) a complexity-aware intervention that modulates agent engagement based on case difficulty, and (ii) a tone-calibrated intervention articulated to balance critique and collaboration. Evaluations on nine medical Q&A and three medical VQA benchmarks show that our approach consistently outperforms both single- and multi-agent LLMs frameworks, including leading commercial models such as GPT-4o and DeepSeek-R1.

Problem

• Addressing premature consensus in multi-agent LLMs for clinical diagnosis
• Introducing Catfish Agent to counteract silent agreement bias
• Improving diagnostic accuracy via structured dissent in complex cases

Innovation

• Introduces Catfish Agent for structured dissent
• Uses complexity-aware intervention for case difficulty
• Employs tone-calibrated intervention to balance critique