XAI Requirements Elicitation for Stroke Rehabilitation

A recent study introduces a video-driven scaffolding method aimed at gathering explainable AI (XAI) requirements from stroke survivors, tackling the issue of algorithmic transparency in rehabilitation brain-computer interfaces. This protocol was tested with three stroke survivors—two of whom had moderate-to-severe aphasia—and three caregivers. Facilitators employed four scaffolding strategies in conjunction with videos: analogical bridging, projective personas, binary forcing, and extension techniques. The findings underscore significant methodological shortcomings in conventional elicitation techniques for individuals with acquired communication disorders.