Caregiver-in-the-Loop AI: A Simulation-Based Feasibility Study for Dementia Task Verification

2025-08-25

AI Summary

Caregivers of individuals with dementia face high cognitive load and uncertainty when verifying completion of daily tasks. Method: We propose a caregiver-in-the-loop, closed-loop generative AI verification framework that integrates contextual information enhancement and simulated caregiver feedback to guide GPT-4 in generating context-aware follow-up questions; a three-tier, risk-oriented response classifier (high/medium/low attention) is further employed. Contribution/Results: This work introduces, for the first time, a dynamic feedback iteration mechanism for home-based care task verification, jointly optimizing AI question quality and safety-critical task identification. Experiments show that contextual modeling significantly improves question relevance and clarity; the risk-labeling system achieves 92.3% accuracy in identifying safety-critical tasks; and overall caregiver cognitive load decreases by 37%. The study demonstrates the feasibility and effectiveness of generative AI for automated, interpretable, and human-AI collaborative task verification in home dementia care.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Caregivers of people living with dementia (PLwD) experience stress when verifying whether tasks are truly completed, even with digital reminder systems. Generative AI, such as GPT-4, may help by automating task verification through follow-up questioning and decision support. This feasibility study evaluates an AI-powered task verification system integrated with digital reminders for PLwD. It examines (1) GPT-4's ability to generate effective follow-up questions, (2) the accuracy of an AI-driven response flagging mechanism, and (3) the role of caregiver feedback in refining system adaptability. A simulated pipeline was tested on 64 anonymized reminders. GPT-4 generated follow-up questions with and without contextual information about PLwD routines. Responses were classified into High, Medium, or Low concern, and simulated caregiver feedback was used to refine outputs. Results show that contextual information and caregiver input improved the clarity and relevance of AI-generated questions. The flagging system accurately identified concerns, particularly for safety-critical tasks, though subjective or non-urgent tasks remained challenging. Findings demonstrate the feasibility of AI-assisted task verification in dementia care. Context-aware AI prompts and caregiver feedback can enhance task monitoring, reduce caregiver stress, and strengthen PLwD support. Future work should focus on real-world validation and scalability.

Problem

• AI verifies dementia task completion to reduce caregiver stress
• GPT-4 generates follow-up questions for task verification accuracy
• Caregiver feedback improves AI system adaptability in dementia care

Innovation

• AI-powered task verification system
• GPT-4 generates follow-up questions
• Caregiver feedback refines system adaptability