Мультимодальная большая языковая модель для разговорной терапии пожилых людей с ADRD (SVR)

Nonpharmacological approaches, such as Reminiscence Therapy (RT), have been a major investigation for ADRD in recent years, but lack of human facilitators, lack of training, technology adaptation, and lack of multi-sensory features are major implementation barriers. In addition, a good RT requires utilizing the personal life experiences of patients as topics of sharing, while showing concrete objects such as photographs, videos, and music that are memorable to the individuals. However, older adults may have a hard time collecting these materials for themselves.

To address these limitations, the researchers have developed a novel SVR that is delivered through a computer program that can be accessed using a personal computer or smart device, and includes new functions so that there is no need for a human therapist to facilitate the RT sessions, and so that older adults and care partners do not need to collect reminiscence materials in advance. SVR presents a virtual therapist (avatar) who interacts with older adults through verbal (speech) and non-verbal communication (eye gaze, body language, etc.), understands the user's speech, and guides the user through reminiscence sessions.

In this study, patient/caregiver dyads will be randomized to receive SVR intervention or an attention control intervention (listening to music). Participants randomized to SVR will either visit the study site (accompanied by the caregiver) to complete SVR sessions or complete the sessions at home, up to twice a week for 12 weeks. Participants randomized to the control arm will receive a curated playlist of calming music, accessible online, that they need to listen to twice a week for 12 weeks. Patient BPSD will be collected at baseline and monthly for 3 months.