Domain Knowledge in Modeling
Domain knowledge is information about the world that is not naturally available as rows and columns.
In infrastructure contexts, this may include what field crews know about construction eras, pipe materials, contractor behavior, maintenance history, soil conditions, inspection practices, or regulatory incentives.
A useful model should make room for this knowledge without treating it as folklore. Sometimes domain knowledge becomes a feature. Sometimes it becomes a prior. Sometimes it becomes a constraint, a causal graph, a data quality warning, or an interpretation check.
The important move is to distinguish what we know from what we are trying to infer. That distinction is central to interpretable statistical models and probabilistic programming.
Related: