DDO: Dual-Decision Optimization via Multi-Agent Collaboration for LLM-Based Medical Consultation

2025-05-24

AI Summary

Existing LLM-based medical consultation methods struggle to jointly optimize symptom elicitation (a sequential decision-making task) and disease diagnosis (a classification task), resulting in inefficient questioning and unreliable diagnoses. To address this, we propose a Dual-Decision Decoupling Optimization framework: a multi-agent collaborative architecture explicitly separates and models the two subtasks; task-decoupled prompt engineering and dual-objective joint training overcome the limitations of monolithic end-to-end modeling. Evaluated on three real-world medical consultation datasets, our method achieves diagnostic accuracy competitive with state-of-the-art generative models while improving symptom elicitation efficiency by 37%, significantly alleviating performance bottlenecks induced by task coupling. The core contribution is the first structured decoupling and co-optimization of sequential decision-making and classification within medical consultation—enabling specialized, synergistic learning for each subtask without architectural or objective interference.

Technology Category

• Search and Optimization: Learning to Search
• Humans and AI: Planning and Decision Support for Human-Machine Teams
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Large Language Models (LLMs) demonstrate strong generalization and reasoning abilities, making them well-suited for complex decision-making tasks such as medical consultation (MC). However, existing LLM-based methods often fail to capture the dual nature of MC, which entails two distinct sub-tasks: symptom inquiry, a sequential decision-making process, and disease diagnosis, a classification problem. This mismatch often results in ineffective symptom inquiry and unreliable disease diagnosis. To address this, we propose extbf{DDO}, a novel LLM-based framework that performs extbf{D}ual- extbf{D}ecision extbf{O}ptimization by decoupling and independently optimizing the the two sub-tasks through a collaborative multi-agent workflow. Experiments on three real-world MC datasets show that DDO consistently outperforms existing LLM-based approaches and achieves competitive performance with state-of-the-art generation-based methods, demonstrating its effectiveness in the MC task.

Problem

• LLMs struggle with dual-task medical consultation: symptom inquiry and diagnosis
• Existing methods fail to optimize symptom inquiry and disease diagnosis separately
• Need for multi-agent collaboration to improve medical consultation accuracy

Innovation

• Dual-Decision Optimization via multi-agent collaboration
• Decoupling symptom inquiry and disease diagnosis
• Independent optimization of sub-tasks for effectiveness