Concept-based Explainable AI (C-XAI): a paradigm shift in XAI

ABOUT THE SPEAKER
Dr. Gabriele Ciravegna

ABSTRACT
The field of eXplainable Artificial Intelligence (XAI) emerged in response to the growing need for more transparent and reliable models. However, using raw features to provide explanations has been discussed in several works lately, advocating for more user-understandable explanations. To address this issue, a wide range of papers proposed Concept-based XAI (C-XAI) methods have been published in recent years. Nevertheless, a unified categorization and precise field definition are still missing. In this talk, we will try to fill the gap through a review of C-XAI approaches. We will identify and define what is a concept and a concept-based explanation. We will then provide guidelines for selecting a suitable category based on the application context. Additionally, we will analyse three prominent proposals focusing on supervised concept-based model. We will see that embedded representations of the concepts overcome the limit in the representation capability of concept-bottleneck models without preventing to provide interpretable predictions.