svartbox

Svartbox is a term used to describe a system whose internal workings are opaque or not readily observable. The term is the Swedish form of the English "black box" and is used across disciplines to discuss processes where inputs and outputs are visible, but the internal decision logic or mechanisms are hidden from users or observers.

In aviation, the "black box" refers to flight data recorders (FDR) and cockpit voice recorders (CVR). The devices are conventionally called black boxes, though the equipment is physically bright orange to aid recovery after accidents. They record flight parameters and cockpit audio to help reconstruct events for safety investigations, with data governed by aviation authorities and typically recovered during or after incidents.

In technology, a black-box model is a system whose internal state and reasoning are not transparent. Complex AI and machine learning models, including deep neural networks, often operate as black boxes. Researchers and practitioners seek explainability using methods such as feature importance, surrogate models, or model-agnostic explanations. There is ongoing debate about trade-offs between accuracy and interpretability.

Beyond aviation and AI, "svartbox" can describe any opaque process in business, governance, or consumer devices, such as data-processing pipelines or proprietary algorithms in services. The term highlights the tension between useful performance and the ability to scrutinize internal decision-making, a consideration increasingly prominent in policy, ethics, and design.

Related terms include black box, explainable AI, and data recorder.