Machine Learning Culture
Machine learning inherits part of its culture from artificial intelligence: build systems that perform tasks humans can do, then evaluate whether the output is good.
That outcome orientation is powerful. It helped make image search, speech recognition, translation, recommendation, and modern generative systems feel almost magical.
But outcome orientation can be limiting in domains where the process matters as much as the prediction. In public infrastructure, a model may need to explain itself to engineers, regulators, residents, and field crews. The output cannot be separated from the decision it informs.
This is not an argument against machine learning. It is an argument for a wider modeling culture that includes: