Research questionHow can we measure whether computer-vision models are understandable to independent human evaluators?XAI has developed many tools for inspecting what models represent and compute. However, those tools do not by themselves show whether independent people can understand the models well enough to support trust or certification.