Supporting mental health of older adults through adaptive generative AI

Conversational agents (CAs), or chatbots, are increasingly valuable for supporting the mental health of older adults. They offer accessible tools for emotional monitoring, remote consultations, psychoeducation, and self-help. For those facing barriers to in-person care—like mobility issues, social isolation, or stigma—CAs provide consistent, approachable support. However, many current systems are rule-based and lack the flexibility to respond to older users’ diverse communication styles and emotional needs. Advances in AI, especially natural language processing and generative technologies, have led to more adaptive CAs. Yet most still rely solely on text input and analysis, limiting their understanding of emotional nuance.
Our proposed CA goes further by incorporating multimodal data—specifically, facial expressions and eye movements—captured via built-in cameras on laptops or mobile devices. These signals help infer users’ motional and cognitive states in real time. By leveraging neurophysiological data (e.g., pupil dilation, gaze), the system can detect subtle signs of stress or disengagement. It then generates a personalized mental health summary and offers targeted video-based therapy. This individualized, moment-by-moment adaptation creates a more responsive and empathetic interaction. The project aims to transform mental health support for older adults by reducing loneliness, promoting emotional well-being, and supporting healthy aging.