Visual Programming Support for the Explainable Artificial Intelligence

The field of Explainable artificial intelligence is showing promising growth in recent years, thus giving researchers the option of deeply exploring the benefits and drawbacks of many different proposed models for solving the enigma of interpretability and explainability regarding machine learning models and their predictions. There are currently a number of techniques that can help to assist researchers in understanding the logic behind decisions made by various models, but the focus of this paper will mostly be on discussing and comparing two strong options, LIME (Local Interpretable Model-Agnostic Explanations) and SHAP (Shapley Additive Explanations). The proposed pipeline for the comparison will be given in a form of an Orange Data Mining workflow. Secondly, the paper aims to give a proposal of how a custom widget encapsulating the functionality of the LIME library can be integrated into the graphical interface, making its usability more appropriate towards less experienced users.