An intention-oriented multi-agent dialogue system for patient-centered decision-making.

Area of Science:

• Artificial Intelligence in Healthcare
• Human-Computer Interaction
• Clinical Decision Support Systems

Background:

• Current healthcare conversational agents primarily serve clinicians, neglecting patient-centered decision-making.
• Limited patient comprehension of medical information can decrease treatment adherence and create doctor-patient friction.
• Existing dialogue systems often lack interpretability, scalability, or have rigid architectures.

Purpose of the Study:

• To develop an interpretable, patient-centered medical dialogue system.
• To enhance patient understanding and support shared decision-making in healthcare.
• To improve communication efficacy between patients and healthcare providers.

Main Methods:

• An intention-oriented multi-agent system was designed, comprising intention understanding, argument reasoning, and interaction agents.
• A medical knowledge ontology and knowledge base were built using hybrid expert-data-driven methods and recent literature.
• System performance was evaluated using objective metrics and clinician user studies on a Graves' eye disease platform.

Main Results:

• The system achieved 85% F1 score for intent recognition and over 95% response accuracy, as rated by physicians.
• User experience evaluations indicated performance comparable to physician-led online systems and superior to traditional medical Q&A systems.
• Comparative analysis showed the system significantly outperformed ChatGPT in personalized reasoning and interpretability.

Conclusions:

• The developed multi-agent dialogue system effectively enhances patient-centered communication through interpretable, personalized decision support.
• The system shows significant potential for improving healthcare dialogue platforms and aiding clinical decision-making.
• This approach addresses limitations in current systems, offering a more patient-centric solution for medical communication.