Damla Övek Baydar emphasizes explainability in medical AI

Damla Övek Baydar, an Assistant Professor in the Department of Computer Engineering at Biruni University, has highlighted that high accuracy rates in healthcare artificial intelligence are insufficient for ensuring true reliability. While AI applications assist in areas ranging from medical imaging to genomic data analysis, Baydar warns that a correct prediction is not always a reliable one.

Many deep learning models function as ‘black boxes,’ meaning their decision-making processes are often opaque. This lack of transparency can lead to models basing conclusions on random patterns or clinically irrelevant data features rather than biological realities. Baydar stresses that for AI to be effectively integrated into clinical decision support systems, it must be explainable, allowing professionals to understand which specific variables or data regions influenced a result.

Furthermore, Baydar notes that while explainability methods can help visualize what a model is focusing on, they do not automatically reveal the underlying biological mechanism. Therefore, AI outputs must be supported by clinical knowledge, biological evidence, and independent validation to ensure safety across diverse patient groups.