ExAID: A Multimodal Explanation Framework for Computer-Aided Diagnosis of Skin Lesions

Explainability is critical for clinical adoption: The paper identifies lack of transparent decision-making as a "principal impediment" to deploying AI diagnostic systems in real clinical workflows, highlighting that doctors need to understand why an AI makes recommendations before trusting them with patient care.

Expert-only explanations aren't good enough: Current explainable AI methods often produce explanations that are "convoluted and not readily comprehensible except by highly trained experts," meaning they fail to serve the broader medical team who would use these systems in practice.

Multi-modal explanations work better than single formats: ExAID combines text-based explanations with visual maps showing where the AI is "looking" in medical images, recognizing that doctors understand diagnostic reasoning better when it's presented in multiple complementary ways rather than numbers or heatmaps alone.

Medical AI needs to handle inherent ambiguity: The framework specifically addresses that skin lesion diagnosis is inherently ambiguous—even the underlying medical problem isn't perfectly clear-cut—so explanations must reflect this complexity rather than presenting false certainty.

Educational value extends beyond individual diagnoses: The system includes an "educational mode" for exploring dataset-level patterns and model behavior, suggesting responsible AI deployment should support medical research and training, not just individual patient decisions.