All models are wrong, but some are useful (Box, 1979)

@XAI

Mittelstadt et al.'s explanation of this maxim is that each model is an approximation and, in a technical sense, an "explanation" of a complex system, that suggests reliable knowledge about the behavior of that system.

Similarly, in explainable AI (XAI) the 'explanations' that are produced are often simpler models making local approximations around a decision, of a general complex system.