Search papers, labs, and topics across Lattice.
This paper introduces Contrastive Concept Importance (CCI), a novel method for attributing the contribution of concepts to model decisions in a contrastive manner, specifically addressing the question of why a model predicts class P over class Q. By leveraging automatically extracted visual concept representations, CCI provides signed scores that indicate whether concepts support the target class or the foil class, allowing for a nuanced understanding of concept influence on model outputs. The evaluation on ImageNet demonstrates that CCI uncovers class-pair-specific behaviors that traditional concept importance methods fail to capture, revealing insights into fine-grained distinctions in model predictions.
Contrastive Concept Importance reveals how specific concepts drive model decisions between class pairs, uncovering insights that traditional methods miss.
Concept-based explanations are a prevalent way to explain the decisions of complex black-box methods through semantically meaningful, human-interpretable concepts. To attribute the contribution of such concepts to a model's decisions, feature attribution methods are used to quantify how strongly each concept contributes to a model output. These attributions are typically computed for a single output class and therefore answer a non-contrastive"why P?"question. In many situations, however, such as cases of misclassification, class confusion, and low-margin predictions, the more natural question to ask is"why P rather than Q?". We introduce contrastive concept importance (CCI), which attributes the logit margin between a target class and a contrast, or foil, class to concepts in an automatically extracted visual concept basis. The resulting scores are signed, indicating whether a concept supports the target over the foil or the foil over the target, and can be decomposed into target-logit and foil-logit effects. This makes it possible to distinguish globally important concepts from concepts that specifically influence a class-pair distinction, including whether their effect is shared, one-sided, or directly contrastive. We evaluate the method on ImageNet class pairs using CRAFT-style concept bases, insertion and deletion curves, logit-wise decomposition analysis, and semantic class hierarchy. The results show that contrastive concept importance reveals class-pair-specific model behavior that is not captured by ordinary concept importance alone, and that highly contrastive concepts can be evaluated against semantic superclass structure to assess whether they affect fine-grained distinctions rather than broad category evidence.
