Artificial intelligence (AI) is increasingly integrated into clinical diagnostic workflows, shifting evaluation priorities from standalone algorithm performance toward collaborative human-AI systems. Although many AI models demonstrate high diagnostic accuracy under controlled conditions, understanding how AI affects clinician productivity, diagnostic accuracy, and decision confidence in real-world settings remains a critical challenge. This paper examines the emerging paradigm of human-AI collaborative diagnostics, focusing on three core performance dimensions: productivity gains through workflow augmentation, accuracy improvements derived from complementary human-machine reasoning, and the impact of AI assistance on clinician decision confidence. Evidence from multi-reader studies, randomized workflow evaluations, and early clinical deployments is synthesized to identify when collaborative systems deliver measurable benefits. The article further proposes a structured evaluation framework for healthcare organizations to quantify the operational and clinical value of AI-assisted diagnostics, emphasizing outcome-based metrics such as time-todiagnosis, diagnostic concordance, workload distribution, and clinical decision stability. By positioning diagnostic AI as a collaborative capability that augments rather than replaces clinicians, this work supports the development of human-centered evaluation approaches for responsible and measurable adoption of intelligent diagnostic systems.
Hugo Raposo (Wed,) studied this question.