Abstract
109. AI CHATBOT BUILT ON LARGE LANGUAGE MODELS FOR DELIVERY OF REMINISCENCE THERAPY TO SOCIALLY ISOLATED OLDER ADULTS: MEET REMI
The American journal of geriatric psychiatry, v 33(10), S80
Oct 2025
Featured in Collection : Drexel's Newest Publications
Abstract
Social isolation is a prevalent issue impacting older adults in the United States, increasing their risk of dementia, stroke, and various health complications. With the growing tech literacy among older populations, AI-based tools including Large Language Models (LLMs) present a promising avenue for delivering care to low-access communities. We developed Remi, an AI-powered chatbot, to provide conversational reminiscence therapy. By engaging older adults in their native languages and encouraging the recollection of personal memories, Remi aims to alleviate social isolation and foster connections with family and friends. This study utilizes a participatory design approach, incorporating feedback from a multidisciplinary clinical research team and older adults, to create a patient-centered, equitable health technology designed to reduce social isolation in this demographic. In this poster, we will highlight a procedure for introducing participants to this new technology and engaging older adults in LLM driven therapy.
The study employs participatory design sessions with socially isolated older adults, their family members, and a multidisciplinary team comprising neuropsychologists, psychiatrists, social workers, and computer scientists. Reminiscence Therapy serves as the therapeutic framework, with large language models (LLMs) supporting Remi's ability to summarize and reflect user narratives, leveraging Retrieval Augmented Generation and memory graphs for effective goal navigation and conversation management. We are aiming to recruit 20 participants for this initial testing period. Participants engage in a 20-minute interaction with Remi, followed by a 40-minute discussion session. These discussions evaluate the chatbot's concept, user experience design, prototype feedback, and safety features. Outcome data will include satisfaction surveys and feedback data for further fine tuning of the chatbot.
Remi is currently undergoing user feasibility testing with non-clinical participants. Initial impressions are positive, particularly amongst users who have prior experience engaging with AI chatbots. Further testing is ongoing. Available survey data and feedback will be included in the final poster.
Remi presents a scalable and innovative solution to combat social isolation among older adults through AI-driven reminiscence therapy. The study underscores the importance of participatory design in developing culturally sensitive and trustworthy AI healthcare interventions. Remi not only contributes to academic advancements in therapeutic applications of LLMs but also offers practical impacts by enhancing social connectivity and mental health resources for marginalized populations. This work sets the foundation for future design guidelines applicable to a broad range of healthcare chatbot interventions.
Metrics
8 Record Views
Details
- Title
- 109. AI CHATBOT BUILT ON LARGE LANGUAGE MODELS FOR DELIVERY OF REMINISCENCE THERAPY TO SOCIALLY ISOLATED OLDER ADULTS: MEET REMI
- Creators
- Alexander Rasgon - The University of Texas at AustinEunhye Ko - The University of Texas at AustinLily Boddy - The University of Texas at AustinJunyuan Hong - The University of Texas at AustinJinhao Duan - Drexel UniversityKaidi Xu - Drexel UniversityErica Garcia-Pittman - The University of Texas at AustinYing Ding - The University of Texas at Austin
- Publication Details
- The American journal of geriatric psychiatry, v 33(10), S80
- Publisher
- Elsevier Inc
- Resource Type
- Abstract
- Language
- English
- Academic Unit
- Computer Science (Computing)
- Other Identifier
- 991022064898504721