Explainable AI isn’t the silver bullet: minimum requirements for evaluating human-AI systems

Resilience & Proactive Safety Initiative webinar from Mar 2026 Dr. Dane Morey, Research Scientist What does it take to safely and effectively deploy AI-infused technologies in safety-critical settings? Although AI algorithms demonstrate impressive capabilities, they still make frequent errors and mistakes. “Explainable AI” has been proposed as a solution to help people appropriately use or reject AI recommendations. However, research has suggested that as little as 5% of explainable AI is evaluated with end users. Do these techniques even work? And, if so, how can we evaluate their effectiveness? This webinar will review our recent findings published in Nature’s Digital Medicine journal to discuss (1) a general method for evaluating the effectiveness explainable AI and other AI-infused technologies and (2) promising techniques and research opportunities towards designing resilient human-AI systems. Dane Morey, PhD is a Research Scientist and part time Lecturer in the Cognitive Systems Engineering Laboratory within the Department of Integrated Systems Engineering at The Ohio State University. His research has focused on design and evaluation strategies that facilitate resilient human-machine performance, especially for AI-infused technologies. His experience designing and evaluating human-machine joint activity spans healthcare, intelligence analysis, aviation, space operations, and public health.