From Attribution Maps to Human-Understandable Explanations through Concept Relevance Propagation
This addresses the transparency gap in deep learning models for users needing comprehensive explanations, though it appears incremental as it builds on existing local and global XAI methods.
The paper tackles the problem of partial insights in explainable AI by introducing Concept Relevance Propagation (CRP), which combines local attribution maps and global concept visualizations to answer both 'where' and 'what' questions for individual predictions, resulting in more human-interpretable explanations.
The field of eXplainable Artificial Intelligence (XAI) aims to bring transparency to today's powerful but opaque deep learning models. While local XAI methods explain individual predictions in form of attribution maps, thereby identifying where important features occur (but not providing information about what they represent), global explanation techniques visualize what concepts a model has generally learned to encode. Both types of methods thus only provide partial insights and leave the burden of interpreting the model's reasoning to the user. In this work we introduce the Concept Relevance Propagation (CRP) approach, which combines the local and global perspectives and thus allows answering both the "where" and "what" questions for individual predictions. We demonstrate the capability of our method in various settings, showcasing that CRP leads to more human interpretable explanations and provides deep insights into the model's representation and reasoning through concept atlases, concept composition analyses, and quantitative investigations of concept subspaces and their role in fine-grained decision making.