Glossary
Black Box Model
Also called: opaque model
A model whose internal decision process cannot be meaningfully inspected, which describes most high-performing deep learning systems.
A black box model produces outputs whose derivation cannot be traced in human-understandable terms. Deep networks with millions of parameters are black boxes in practice regardless of whether the weights are technically inspectable.
Where This Gets Misread
The tradeoff between performance and interpretability is real but frequently overstated as an excuse. For many clinical prediction tasks, simpler interpretable models perform comparably, and opacity was a design convenience rather than a necessity. Ask whether a simpler model was tried before accepting that the black box was required.