Listening to Patients: A Framework of Detecting and Mitigating Patient Misreport for Medical Dialogue Generation

2024-10-08

AI Summary

Misreporting of symptoms by patients in medical dialogues introduces diagnostic bias, undermining the reliability of AI-assisted clinical decision-making. Method: We propose the first integrated framework for misreport detection and mitigation. It constructs a dialogue entity graph to model symptom expression relationships and introduces graph entropy to quantify semantic inconsistency for interpretable misreport identification. Based on detected misreports, the framework automatically generates semantically precise clarification questions and performs end-to-end fine-tuning with GPT-4. Contribution/Results: Our approach uniquely unifies structured graph-based analysis with large language model (LLM) generation capabilities, enabling both explainable misreport detection and proactive intervention. Experiments demonstrate significant improvements in consultation accuracy and response quality, outperforming all baselines across F1 score, BLEU, and clinical plausibility metrics. The framework substantially enhances the robustness and reliability of LMs in real-world medical dialogues.

Technology Category

• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Medical Dialogue Systems aim to provide automated healthcare support through patient-agent conversations. Previous efforts typically regard patients as ideal users -- one who accurately and consistently reports their health conditions. However, in reality, patients often misreport their symptoms, leading to discrepancies between their reports and actual health conditions. Overlooking patient misreport will affect the quality of healthcare consultations provided by MDS. To address this issue, we argue that MDS should ''listen to patients'' and tackle two key challenges: how to detect and mitigate patient misreport effectively. In this work, we propose PaMis, a framework of detecting and mitigating Patient Misreport for medical dialogue generation. PaMis first constructs dialogue entity graphs, then detects patient misreport based on graph entropy, and mitigates patient misreport by formulating clarifying questions. Experiments indicate that PaMis effectively enhances medical response generation, enabling models like GPT-4 to detect and mitigate patient misreports, and provide high-quality healthcare assistance.

Problem

• Medical Chatbots
• Inaccurate Symptom Description
• Misleading Medical Advice

Innovation

• PaMis
• Error Correction
• Medical Advice Accuracy