Towards Human-Understandable Multi-Dimensional Concept Discovery

Titelangaben

Grobrügge, Arne ; Kühl, Niklas ; Satzger, Gerhard ; Spritzer, Philipp:
Towards Human-Understandable Multi-Dimensional Concept Discovery.
In: Proceedings of the 37th IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR). - Nashville, USA , 2025

Volltext

Link zum Volltext (externe URL):

Abstract

Concept-based eXplainable AI (C-XAI) aims to overcome the limitations of traditional saliency maps by converting pixels into human-understandable concepts that are consis tent across an entire dataset. A crucial aspect of C-XAI is completeness, which measures how well a set of con cepts explains a model’s decisions. Among C-XAI meth ods, Multi-Dimensional Concept Discovery (MCD) effec tively improves completeness by breaking down the CNN la tent space into distinct and interpretable concept subspaces. However, MCD’s explanations can be difficult for humans to understand, raising concerns about their practical util ity. To address this, we propose Human-Understandable Multi-dimensional Concept Discovery (HU-MCD). HU MCDuses the Segment Anything Model for concept identi f ication and implements a CNN-specific input masking tech nique to reduce noise introduced by traditional masking methods. These changes to MCD, paired with the com pleteness relation, enable HU-MCD to enhance concept un derstandability while maintaining explanation faithfulness. Our experiments, including human subject studies, show that HU-MCD provides more precise and reliable explana tions than existing C-XAI methods. The code is available at https://github.com/grobruegge/hu-mcd.