From "Where" to "What": Towards Human-Understandable Explanations through Concept Relevance Propagation

The emerging 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. Only few contemporary techniques aim at combining the principles behind both local and global XAI for obtaining more informative explanations. Those methods, however, are often limited to specific model architectures or impose additional requirements on training regimes or data and label availability, which renders the post-hoc application to arbitrarily pre-trained models practically impossible. In this work we introduce the Concept Relevance Propagation (CRP) approach, which combines the local and global perspectives of XAI and thus allows answering both the "where" and "what" questions for individual predictions, without additional constraints imposed. We further introduce the principle of Relevance Maximization for finding representative examples of encoded concepts based on their usefulness to the model. We thereby lift the dependency on the common practice of Activation Maximization and its limitations. We demonstrate the capabilities of our methods in various settings, showcasing that Concept Relevance Propagation and Relevance Maximization lead to more human interpretable explanations and provide deep insights into the model's representations and reasoning through concept atlases, concept composition analyses, and quantitative investigations of concept subspaces and their role in fine-grained decision making.Comment: 79 pages (40 pages manuscript, 10 pages references, 29 pages appendix) 51 figures (26 in manuscript, 25 in appendix) 1 table (in appendix