Clinician perspectives on trust and adoption of AI in breast cancer diagnosis

Purpose: As artificial intelligence (AI) and explainable AI (XAI) technologies become increasingly integrated into clinical workflows, understanding how clinicians perceive and trust these systems is essential for their successful adoption. This study aimed to explore the factors that influence clinicians’ trust in AI-based clinical decision support systems (CDSS) in the context of breast cancer detection, with a particular focus on the role of explainability features. Method: A mixed-methods approach was employed, combining thematic analysis of semi-structured interviews with an interrupted time series experimental design. The study involved 28 clinicians in total for the quantitative component. Of these, 11 clinicians also took part in follow-up qualitative interviews. Results: The findings were largely complementary, revealing consistent patterns across both approaches. Eight primary themes emerged from qualitative interviews with radiologists and oncologists, highlighting the perceived benefits of AI, such as early cancer detection, clinical prioritization, and disease progression monitoring, alongside concerns about unclear definitions, professional role erosion, and overreliance. Quantitative findings showed that trust was significantly higher in AI-assisted conditions compared to unaided diagnosis, with perceived accuracy and information richness emerging as key predictors of trust. Notably, the addition of explainability features did not significantly enhance trust or performance. Conclusion: These results underscore the importance of user-centered design, clinical validation, and contextual integration, suggesting that trust in clinical AI systems is shaped more by practical reliability than by transparency alone.