Enhancing Mental Health Counseling Support in Bangladesh using Culturally-Grounded Knowledge

arXiv cs.AI / 4/17/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The study tackles a key limitation of large language models in mental health counseling: responses can be culturally insensitive, insufficiently grounded in context, and not always clinically appropriate.
It proposes a systematic way to inject clinically validated, domain-specific knowledge into LLMs, focusing on support for para-counselors in Bangladesh.
The authors compare retrieval-augmented generation (RAG) with a knowledge-graph (KG) approach, where the KG is manually built and expert/clinically validated to encode causal relationships among stressors, interventions, and outcomes.
Experiments evaluate multiple LLMs using both automatic text-similarity metrics (BERTScore, SBERT) and human assessments across five metrics that aim to measure counseling effectiveness beyond surface-level similarity.
The results indicate that KG-based methods outperform RAG alone by improving contextual relevance, clinical appropriateness, and practical usability, highlighting the value of structured expert knowledge for counseling-oriented AI.

Abstract

Large language models (LLMs) show promise in generating supportive responses for mental health and counseling applications. However, their responses often lack cultural sensitivity, contextual grounding, and clinically appropriate guidance. This work addresses the gap of how to systematically incorporate domain-specific, clinically validated knowledge into LLMs to improve counseling quality. We utilize and compare two approaches, retrieval-augmented generation (RAG) and a knowledge graph (KG)-based method, designed to support para-counselors. Our KG is constructed manually and clinically validated, capturing causal relationships between stressors, interventions, and outcomes, with contributions from multidisciplinary people. We evaluated multiple LLMs in both settings using BERTScore F1 and SBERT cosine similarity, as well as human evaluation across five metrics, which is designed to directly measure the effectiveness of counseling beyond similarity at the surface level. The results show that KG-based approaches consistently improve contextual relevance, clinical appropriateness, and practical usability compared to RAG alone, demonstrating that structured, expert-validated knowledge plays a critical role in addressing LLMs limitations in counseling tasks.