Retrieval-Augmented LLM-Based Empathetic Chatbot for Early Postpartum Depression Screening in Aceh

Abstract

Postpartum Depression (PPD) remains a significant maternal mental health concern, particularly in low-resource settings where access to professional psychological services is limited. Although digital mental health tools have emerged to address this gap, most existing chatbot-based systems rely on rule-based interactions, offer limited personalization, and lack integration of structured clinical screening mechanisms. This study addresses the lack of culturally adapted, LLM-based empathic chatbots for postpartum mental health screening in low-resource Indonesian settings. We design and implement an AI-driven conversational chatbot that integrates a Retrieval-Augmented Generation (RAG) architecture with a Large Language Model (LLM) to enable context-aware, knowledge-grounded response generation. The system incorporates a Patient Health Questionnaire-9 (PHQ-9)–based screening module to support early identification of depressive symptoms and adaptive conversational support. An early-stage usability evaluation was conducted through a seven-day user interaction study involving 30 postpartum mothers in Aceh, with 12 participants completing the System Usability Scale (SUS). The system achieved an average SUS score of 85.63, indicating excellent perceived usability. While the evaluation focuses on usability rather than clinical effectiveness, the findings suggest that the proposed system demonstrates feasibility as a culturally adapted, scalable digital support tool for early postpartum mental health screening. Further studies with larger samples and long-term evaluation are required to assess clinical impact and sustained user engagement.

GS Cited Analysis