blackbox

A black box is a system or device whose internal structure or workings are not known or not observable to the user. The external behavior—inputs, outputs, and responses—can be studied without access to the internal components. The term derives from the metaphor of a sealed container whose interior is opaque to the observer.

In engineering, a black-box model or component is analyzed solely by its input-output behavior. Engineers may substitute a black box with a known interface to test how a system responds to certain stimuli. Black-box testing in software similarly validates functionality without examining internal implementation, focusing on requirements and expected outcomes.

In machine learning and artificial intelligence, a black-box model refers to algorithms whose internal decision-making processes are difficult to interpret, such as deep neural networks. This opacity raises concerns about transparency, fairness, and accountability. Methods of explainable AI seek to illuminate why a model produces particular results, or to simplify models while preserving accuracy.

In aviation and transport, the term black box commonly refers to flight data recorders and cockpit voice recorders, which capture critical information after accidents. In practice these devices are bright orange for visibility, and they are designed to withstand extreme conditions. The term can also describe any data-logging device used to reconstruct events.

The concept appears in philosophy and systems theory as epistemic opacity, describing limits to what can be inferred about a system from its observable behavior alone.