Explainability of artificial intelligence (AI) tools in assisted reproductive technologies (ARTs): promise, clinical reality, and conceptual limits

Artificial intelligence (AI) is increasingly integrated into clinical medicine, offering new opportunities for decision support, prediction, and personalization of care. However, many high-performing AI systems, particularly those based on deep learning, operate as 'black boxes', raising concerns about transparency, trust, and professional accountability. Explainable artificial intelligence (XAI) has been proposed as a solution to these challenges, yet its clinical value remains contested. In this Perspective, we critically examine the role of explainability in medical AI using assisted reproductive technologies (ARTs) as a paradigmatic clinical context. ART is characterized by biological complexity, probabilistic decision-making, and growing reliance on AI tools for embryo assessment, gamete selection, and outcome prediction. Drawing on interdisciplinary expertise in medicine, AI, philosophy, and social sciences, we briefly review current AI applications in ART, clarify key conceptual distinctions between interpretability, explainability, and transparency, and analyze the potential benefits and limitations of XAI. We argue that explainability should not be regarded as an intrinsic requirement for all clinical AI systems. Instead, its value is context-dependent and must be weighed against other critical factors such as transparent reporting, rigorous validation, and demonstrated clinical impact. Moving beyond a binary opposition between black-box and explainable models, we advocate for a pragmatic, evidence-based approach to the integration of AI into digital medicine.