Multi-modal Large Language Model-Empowered Talk Therapy for Older Adults With ADRD

SVR Therapy is delivered through a computer program that can be accessed using a personal computer or smart device. It uses virtual reality, automated speech recognition, and large language models to achieve virtual therapist-patient interaction. The SVR therapy platform also automatically sources multi-modal reminiscence materials (e.g., images, text, etc.) through search engines and databases to facilitate the reminiscence therapy sessions. Each participant will participate in two 25-60 min (depending on content) sessions/week across 12 weeks. Each session will include multiple trials of reminiscence, such as music, sports, arts, etc. Each trial takes about 10 minutes to finish. The exact number of trials finished will depend on each patient's speed, preference, and how long they engage in the interaction with the SVR system. The participants can take breaks as needed.

Participants in the attention control group will receive a curated playlist of calming music that they need to listen to twice a week for 12 weeks. The music can be conveniently accessed online (e.g., through YouTube).