Explainable Artificial Intelligence (XAI) methods have the potential to increase confidence in deep models classifying skin pigment lesions in the context of melanoma diagnosis. This project takes advantage of SHAP, LIME, GradCAM and other XAI-based approaches to derive the most important features that contribute to the melanoma diagnosis and align the models’ results with clinical knowledge/practice. The reproducibility and computational efficiency of the explainers are also a goal of the project. Three public datasets are examined (MelanomaML, PH2, ISIC) and the Interactive Atlas of Dermatoscopy by Argenziano. The research is ongoing, but the first results are beneficial in helping to better understand the operation of models and their optimization. In this work, we present first results using the SHAP gradient explainer applied to dermoscopic images of pigmented skin lesions that have been clinically confirmed by a dermatologist.