physicians' attitudes on four topics: A) human-AI interaction in clinical decisionmaking, B) the impact on radiologists' diagnostic performance, C) explainable AI (XAI), and D) general attitudes towards using AI in their work.Understanding physicians' lived experiences with these systems is essential for regulation of AI in oncology. Methods:The study was conducted by performing searches on Scopus.The search combined keywords and synonyms that encompass words for lung cancer, AI, user experience, and clinical implementation.After title and abstract screening, followed by full-text assessment, only 12 out of 150 articles met the inclusion criteria focusing on physicians' experiences with AI in lung cancer care.A qualitative synthesis of doctors' experiences using AI in lung cancer treatment was made based on these 12 articles.Results: Physicians generally view AI as potentially useful for efficiency and workload reduction, but not reliable enough for autonomous use.AI is mainly experienced as a decision support or a second-reader tool.Trust in AI is fragile, case-specific, and easily undermined.Physicians prioritize patient safety and they override AI outputs when these appear clinically unsafe or implausible.Having to interpret AI results is seen as a cognitive burden.Physicians feel insufficiently trained to understand AI limitations and to interpret AI outputs.Real-world experience is shaped by workflow integration, costs, and IT constraints.As for XAI, clinicians feel left out of XAI model development, and implementation is poorly integrated into workflows.Current explanations are often too technical or too simplified.Conclusions: Physicians are cautiously optimistic about AI.They see AI as helpful, but not as autonomous decision-makers.Trust is fragile and case-specific.Inconsistencies and poorly aligned explanations can decrease trust.The study did not find strong evidence that AI systematically increased confidence levels in lung cancer care.
Klimatsidas et al. (Tue,) studied this question.