Ethically-Constrained Mental Health Chatbot for University Students: A Context-Aware NLP System using LLMs

Abstract

Limited access to mental health services, cultural stigma, and societal barriers continue to hinder help-seeking behaviors among Indonesian youth. While chatbots offer promising digital support, ethical, clinical, and safety concerns remain underexplored. This study investigates the foundational development of mental health chatbots for Indonesian university students by identifying key ethical considerations. Using reflective and exploratory design methodology focused on mapping ethical risks and developing a conceptual framework, data were gathered through expert interviews and focus group discussions involving AI developers, psychologists, and bioethics specialists. Results highlight four central challenges: ethically sourcing training data, defining nonclinical chatbot roles, ensuring cultural-linguistic sensitivity, and integrating a responsive safety keyword protocol. The chatbot’s role was refined to act as a peer-like first responder offering empathetic support and referral, not diagnosis. Interdisciplinary collaboration was critical to maintaining clinical boundaries and user protection. Ultimately, chatbot systems must be developed as ethically grounded, culturally responsive, and professionally supervised tools that complement, not replace, human mental health care and protect vulnerable users.

Ähnliche Arbeiten