Improving Health Professionals' Onboarding with AI and XAI for Trustworthy Human-AI Collaborative Decision Making

Onboarding matters for AI adoption in healthcare: Health professionals need more than just accuracy statistics to understand and trust AI systems. The study found that participants wanted benchmark information (comparisons to existing standards), practical benefits, and hands-on trials—not just technical performance metrics—to properly contextualize how AI would work in their actual practice.

Interactive experience is essential for non-technical users: Simply explaining AI explanations wasn't enough; participants needed to interact with the AI system directly during onboarding. This trial-and-error approach helped them refine their understanding of what the AI could and couldn't do, suggesting that passive learning materials are insufficient for building appropriate trust.

Context-specific onboarding improves human-AI collaboration: The research used a real-world example (stroke rehabilitation assessment) rather than abstract AI concepts, and this grounded approach resonated with health professionals. This suggests that effective AI onboarding should be tailored to specific clinical workflows rather than using generic AI training materials.

Current XAI presentations may not serve non-technical audiences well: The study identified a gap between how explainable AI is typically presented and what health professionals actually need to understand it. This implies that many existing XAI tools designed for technical audiences may need redesign before deployment in clinical settings with diverse staff backgrounds.