Retrieval Augmented Generation System for Mental Health Information.

Area of Science:

• Artificial Intelligence
• Mental Health Informatics
• Computational Linguistics

Background:

• Large Language Models (LLMs) can hallucinate, limiting their use in knowledge-intensive, high-accuracy applications.
• Retrieval Augmented Generation (RAG) enhances LLMs by integrating external knowledge, improving factual accuracy.
• Existing RAG applications in health exist, but a specialized system for mental health using high-reputation evidence is lacking.

Purpose of the Study:

• To propose and validate an accurate Retrieval Augmented Generation (RAG) system specifically tailored for mental health.
• To ensure the reliability and accuracy of mental health information generated by LLMs.
• To explore the utility of RAG for supporting mental health policy and counselling.

Main Methods:

• Development of a novel RAG system focused on the mental health domain.
• Integration of a curated knowledge base of reputable scientific evidence.
• Validation of the system's accuracy in retrieving and generating mental health information.

Main Results:

• The proposed RAG system demonstrates enhanced accuracy in mental health information retrieval.
• The system effectively mitigates LLM hallucinations in the mental health context.
• The study validates the practical utility of RAG for mental health applications.

Conclusions:

• RAG systems are valuable tools for improving the accuracy and reliability of mental health information.
• The developed RAG system can aid policymakers in designing mental health interventions.
• Further expansion of the knowledge base can extend the system's utility for mental health counselling and support.