CARING for xAI

Over the last few years, Artificial Intelligence (AI) has pervaded our lives. As a result, automated tools that “reason” on different scenarios have become more and more common. As this trend continues to grow, it has become necessary to ensure that newly developed tools and technologies can be safely adopted, as demonstrated by the numerous EU regulations. This is especially true when the concept of AI is intertwined with the field of medicine, where every decision may be critical. That is why, in this work, we decided to tackle the problem of automated interpretation of Computed Tomography (CT) scans using an explainable approach. In fact, while several methods based on Machine Learning (ML) are currently available, these are still outperformed by medical doctors and provide answers that cannot be traced back to a logical deduction. This paper presents CARING, a new methodology based on Answer Set Programming (ASP), which returns reliable, easy-to-program and explainable interpretations of CT scans. In particular, CARING makes use of transparent technologies in order to handle medical knowledge provided either by experts or by verified ontologies. This proof of concept shows that Logic Programming is a mature technology that can match the newest challenges in the xAI field.