Using Generative-AI to Support Carer Telephone Support Services: Preliminary Findings

This experimental study explores the utilization of Generative-AI, specifically Large Language Models (LLMs), to support Carer Telephone Support Services in Hong Kong. The demand for these services, which operate 24/7, requires substantial manpower for tasks such as answering calls, making outbound calls, and preparing call summaries. This study, conducted in collaboration with the Hong Kong Federation of Women’s Center Carer Support Services Hotline, involved recording 71 calls with the consent of caregivers. These recordings were transcribed verbatim and analyzed by both an LLM known as Copilot and an experienced social worker. The findings reveal that Copilot can effectively identify challenges faced by caregivers. However, it struggles to contextualize information within the dialog content, often over-generalizing, over-simplifying, or creating new content, likely influenced by the design of the prompts used. These initial results highlight the potential of LLM in reducing the manpower demands of such services, though they also underscore the need for future research focused on enhancing the comprehension abilities of LLM and improving prompt design to ensure more accurate and contextually appropriate responses.