the difference is between having a "model" and a "theory". A theory tries to explain the "why" based on some givens, and a model tell you the "how". For engineering, we want why and not how. ie, for bugs, we want to root-cause, and fix - not fix by trial-and-error.

the hard sciences have theories. and soft sciences have models.

computer science is built on theory (turing machine/lambda calc/logic).

AI models are well "models" - we dont know why it works but it seems to - thats how models are.