"Like Having Family Nearby": An LLM-Powered Family-Resembling Avatar for offering Psychological Support and Training Guidance during Stroke Rehabilitation

Read full text
Authors: Peixuan Xiong, Bingchen Guo, Chao Liu, Zhenyu Wang, Xian Li, Nianci Zhao, Kaishun Wu, Xian Li, Gong Chen, Yuling Wang, Mingming Fan
Cited by: 1
References: 38

Abstract

Stroke rehabilitation is prolonged and emotionally demanding, yet family members cannot provide continuous companionship or coaching. We explore a family-resembling avatar that offers psychological support and training guidance when relatives cannot be present. We conducted an empirical user study with six stroke patients and their family members (N=12). Our qualitative findings suggest such companions may reduce loneliness and increase perceived warmth, while action-relevant guidance can complement caregivers’ limited rehabilitation knowledge. Participants also raised concerns about emotional dependence, privacy, and the need for stronger clinical alignment to avoid inappropriate guidance. We discuss implications for designing family-resembling companion-coach avatars with clear boundaries, privacy-preserving personalization, and supervised deployment in rehabilitation workflows.

Fields of Study & Subjects

Stroke Rehabilitation and Recovery Traumatic Brain Injury Research Occupational Therapy Practice and Research

Publication details

Journal / source
Proceedings of the Extended Abstracts of the 2026 CHI Conference on Human Factors in Computing Systems
Publisher
ACM
Pages
1–7
Publication date
13 April 2026
Language
EN