Black Box (AI)
Working knowledgeGovernance, Safety and Ethics
Also called: Black Box
A black box, applied to AI, is a system whose internal decision-making cannot be inspected or explained, even by the people who built it. You can see what goes in and what comes out, but not why. This is the core problem that explainability and interpretability work exists to solve.
In practice
When a bank cannot explain why an AI declined a loan application, that is the black box problem in practice, not a hypothetical one. Regulated decisions need a reason that survives a customer complaint and a regulator's question. Decide before deployment whether the use case can tolerate an answer nobody can account for.