Debugging Concept Bottleneck Models through Removal and Retraining
This work addresses interpretability and bias issues in CBMs for domain experts, but it is incremental as it builds on existing CBM frameworks with a novel retraining method.
The paper tackles the problem of systemic misalignment in Concept Bottleneck Models (CBMs) due to biased data, by introducing a debugging framework with a novel method called CBDebug that converts concept-level feedback into sample-level labels for bias mitigation, resulting in significant outperformance over prior retraining methods across multiple architectures and benchmarks.
Concept Bottleneck Models (CBMs) use a set of human-interpretable concepts to predict the final task label, enabling domain experts to not only validate the CBM's predictions, but also intervene on incorrect concepts at test time. However, these interventions fail to address systemic misalignment between the CBM and the expert's reasoning, such as when the model learns shortcuts from biased data. To address this, we present a general interpretable debugging framework for CBMs that follows a two-step process of Removal and Retraining. In the Removal step, experts use concept explanations to identify and remove any undesired concepts. In the Retraining step, we introduce CBDebug, a novel method that leverages the interpretability of CBMs as a bridge for converting concept-level user feedback into sample-level auxiliary labels. These labels are then used to apply supervised bias mitigation and targeted augmentation, reducing the model's reliance on undesired concepts. We evaluate our framework with both real and automated expert feedback, and find that CBDebug significantly outperforms prior retraining methods across multiple CBM architectures (PIP-Net, Post-hoc CBM) and benchmarks with known spurious correlations.