Development of a Mental Health Chatbot Using Large Language Models for Indonesian Undergraduates

Mental health issues among university students are an escalating global problem, especially in Indonesia. Young adults, especially college students, are a vulnerable demographic due to academic pressures, social transitions, and the challenges of independent living. This paper describes the development of a mental health chatbot specifically for Indonesian university students, utilizing large language models (LLMs). The development encompassed three essential phases: establishing a knowledge base, refining the LLM with culturally pertinent Indonesian data, and assessing the system. The findings demonstrated notable enhancements in the chatbot's precision, with training accuracy rising from 0.6 to 0.85 and validation accuracy from 0.5 to 0.79 due to fine-tuning and regularization. User acceptability testing including 58 students resulted in an acceptance percentage of 87.38, indicating substantial satisfaction with the chatbot's response and usefulness. This research highlights the efficacy of AI-driven interventions in addressing the mental health requirements of Indonesian university students. © 2024 IEEE.