Lead Service Line Replacement
Lead service line replacement is a useful example of machine learning in public infrastructure.
The practical task is not merely to predict whether a parcel has lead. The prediction informs excavation, inspection, replacement scheduling, resident communication, and public accountability.
The original draft used Flint as an example because the problem combines:
- incomplete historical records
- strong domain knowledge from plumbers and local experts
- expensive inspections
- high public-health stakes
- the need to update beliefs as pipe materials are revealed
A model may detect that houses from a certain era or area have elevated risk. A better modeling process asks why. Was it code? material prices? contractor behavior? missing records? inspection selection? The answer changes how much trust we place in the pattern.
Related: