Perancangan Arsitektur Kecerdasan Buatan "Bubu": Agen Pendamping PsikososialMahasiswa Berbasis Multimodal dan Kearifan Lokal

Abstract

The mental well-being of Indonesian university students is currently facing complex pressures, ranging from heavy academic demands to economic instability and the psychological impacts of social media interactions. Unfortunately, the availability of professional psychologists is not commensurate with the population in need, creating a significant service gap. This study aims to design "BUBU," a web-based artificial intelligence agent positioned as a first-line defense for psychosocial support. The system is developed using a Human-Centered Design approach, integrating multimodal inputs (text, voice, and visuals) to detect user emotions in real-time. BUBU is trained using a Large Language Model (LLM) with prompt engineering techniques that adopt student vernacular (slang/code-mixing) and Active Listening protocols. Test results show that the system is capable of providing contextual emotional validation, adjusting response tones based on facial expressions, and possesses a fail-safe mechanism to detect suicide risks and provide appropriate referrals. This prototype offers potential as an effective mental health triage tool in campus environments.