Global Counterfactual Directions
This work addresses the need for global interpretability in AI systems, particularly for visual classifiers, by extending counterfactual explanations beyond local techniques, though it is incremental in building upon existing diffusion models.
The authors tackled the problem of generating global counterfactual explanations for visual classifiers by discovering global directions in the latent space of Diffusion Autoencoders, enabling decision-flipping on entire datasets and increasing explanation diversity.
Despite increasing progress in development of methods for generating visual counterfactual explanations, especially with the recent rise of Denoising Diffusion Probabilistic Models, previous works consider them as an entirely local technique. In this work, we take the first step at globalizing them. Specifically, we discover that the latent space of Diffusion Autoencoders encodes the inference process of a given classifier in the form of global directions. We propose a novel proxy-based approach that discovers two types of these directions with the use of only single image in an entirely black-box manner. Precisely, g-directions allow for flipping the decision of a given classifier on an entire dataset of images, while h-directions further increase the diversity of explanations. We refer to them in general as Global Counterfactual Directions (GCDs). Moreover, we show that GCDs can be naturally combined with Latent Integrated Gradients resulting in a new black-box attribution method, while simultaneously enhancing the understanding of counterfactual explanations. We validate our approach on existing benchmarks and show that it generalizes to real-world use-cases.