termikos

Termikos is a neologism used in discussions of explainable artificial intelligence (XAI) to refer to a boundary of interpretability. It denotes the point at which explanations of a model’s behavior are considered sufficiently understandable by non-expert decision-makers to support trust and action, rather than merely reflecting technical fidelity. The concept emphasizes user-centered evaluation of explanations, including cognitive load, relevance to tasks, and the practical ability to anticipate the consequences of model outputs.

Termikos appears to be a blended neologism, combining "term" with the suffix "-ikos," a pattern seen in theoretical terminology to denote a property. Its exact origin is unclear, but the term has circulated in AI ethics and explainability discussions since the late 2010s, typically without formal definition in core literature.

In practice, termikos functions as a qualitative benchmark rather than a formal metric. Proponents say it helps align explanations with user understanding and task relevance. Critics argue that the lack of a precise definition makes the term subjective and context-dependent, risking confusion with trust or usability. It is often treated as a guiding principle alongside metrics such as fidelity, sparsity, and actionability.

See also: Explainable AI, model interpretability, human-centered AI.