AI-Care: A Conversational Agentic System for Task Coordination in Alzheimer's Disease Care
Researchers have introduced AI-Care, a conversational artificial intelligence system designed to assist individuals with Alzheimer's disease (AD) and related dementias in managing daily tasks. Recognizing that complex digital interfaces pose significant barriers for these users, AI-Care employs a voice-first chatbot interface to reduce cognitive load during activities like setting calendar reminders and organizing to-do lists. Built on a LangGraph-based stateful orchestration framework, the system processes requests through rigorous steps including sanitization, intent classification, and safety checks. Crucially, it avoids autonomous medical decisions, grounding safety-critical information such as medication details in caregiver-verified records. The system handles ambiguous inputs through multi-turn clarification rather than guessing, supporting both typed and spoken interactions with ElevenLabs text-to-speech output. A preliminary pilot study involving four participants with mild-to-moderate AD demonstrated that users found the system trustworthy, competent, and likable. They successfully completed coordination tasks through natural language conversation. This formative evaluation highlights the potential of agentic AI to enhance independence and quality of life for dementia patients by simplifying digital tool interaction while maintaining strict safety protocols.
Wire timeline
AI-Care: A Conversational Agentic System for Task Coordination in Alzheimer's Disease Care
Researchers have introduced AI-Care, a conversational artificial intelligence system designed to assist individuals with Alzheimer's disease (AD) and related dementias in managing daily tasks. Recognizing that complex digital interfaces pose significant barriers for these users, AI-Care employs a voice-first chatbot interface to reduce cognitive load during activities like setting calendar reminders and organizing to-do lists. Built on a LangGraph-based stateful orchestration framework, the system processes requests through rigorous steps including sanitization, intent classification, and safety checks. Crucially, it avoids autonomous medical decisions, grounding safety-critical information such as medication details in caregiver-verified records. The system handles ambiguous inputs through multi-turn clarification rather than guessing, supporting both typed and spoken interactions with ElevenLabs text-to-speech output. A preliminary pilot study involving four participants with mild-to-moderate AD demonstrated that users found the system trustworthy, competent, and likable. They successfully completed coordination tasks through natural language conversation. This formative evaluation highlights the potential of agentic AI to enhance independence and quality of life for dementia patients by simplifying digital tool interaction while maintaining strict safety protocols.
cs.AI updates on arXiv.org