An Axiomatic Approach to Model-Agnostic Concept Explanations
This work addresses the need for interpretable AI tools that work across different models, though it appears incremental by building on existing concept explanation methods.
The paper tackles the problem of model-specific concept explanations by proposing a model-agnostic approach based on three axioms: linearity, recursivity, and similarity, and demonstrates its utility in scenarios like model selection and zero-shot vision language model improvement.
Concept explanation is a popular approach for examining how human-interpretable concepts impact the predictions of a model. However, most existing methods for concept explanations are tailored to specific models. To address this issue, this paper focuses on model-agnostic measures. Specifically, we propose an approach to concept explanations that satisfy three natural axioms: linearity, recursivity, and similarity. We then establish connections with previous concept explanation methods, offering insight into their varying semantic meanings. Experimentally, we demonstrate the utility of the new method by applying it in different scenarios: for model selection, optimizer selection, and model improvement using a kind of prompt editing for zero-shot vision language models.