Framework for Automatic Interpretability in Machine Learning – FAbLe

Recent technological advances rely on accurate decision support systems that have been constructed as black boxes. That is, the system's internal logic is not available to the user, e.g., due to the complexity of system. This lack of explanation can lead to technical, ethical, and legal issues. For this reason, multiple research approaches provide comprehensible explanations for the decisions of traditionally accurate but black-box-like machine learning algorithms such as neural networks. All these approaches rely on explanations on simpler models such as linear functions, rules, and decision trees. Nevertheless, data scientists do not count on a straightforward way to choose the most suitable explanation model for a given use case. Our proposal, called FAbLe (Framework for Automatic Interpretability in Machine Learning), aims at fully automating this process in order to provide the most faithful and comprehensible explanations.