Constructing Concept-based Models to Mitigate Spurious Correlations with Minimal Human Effort
This addresses the issue of spurious correlations in interpretable models for AI practitioners, offering a more efficient alternative to high-cost human annotation, though it is incremental as it builds on existing CBM frameworks.
The paper tackles the problem of spurious correlations in models by constructing Concept Bottleneck Models (CBMs) with minimal human effort, using foundation models to annotate concepts and reduce biases, and demonstrates effectiveness in reducing reliance on spurious correlations across multiple datasets.
Enhancing model interpretability can address spurious correlations by revealing how models draw their predictions. Concept Bottleneck Models (CBMs) can provide a principled way of disclosing and guiding model behaviors through human-understandable concepts, albeit at a high cost of human efforts in data annotation. In this paper, we leverage a synergy of multiple foundation models to construct CBMs with nearly no human effort. We discover undesirable biases in CBMs built on pre-trained models and propose a novel framework designed to exploit pre-trained models while being immune to these biases, thereby reducing vulnerability to spurious correlations. Specifically, our method offers a seamless pipeline that adopts foundation models for assessing potential spurious correlations in datasets, annotating concepts for images, and refining the annotations for improved robustness. We evaluate the proposed method on multiple datasets, and the results demonstrate its effectiveness in reducing model reliance on spurious correlations while preserving its interpretability.