Counterfactual Contrastive Analysis
The episode discusses a paper titled "Counterfactual Contrastive Analysis" by Yunlong He and Pietro Gori. The hosts examine how this method improves upon existing AI explainability methods by operating on data distributions rather than brittle classifier boundaries. They conclude that the technique offers reliable, high-fidelity explanations for complex systems, particularly in medical imaging.