The point about not having models is a very important point. Psychology is so sure that the mind exists and has certain structural properties, but to what extent is it empirical?
A model is useful if its behavior matches observed behavior reasonably well. So long as the abstraction doesn't leak, it doesn't need to represent the mechanisms correctly. Even if a "mind" doesn't exist, our mind-model may still be useful.
Physicists are so sure that the theory of gravity is true, but to what extent has this been verified? Have they seen the gravity particle yet?
Isn't the most useful model just the formla for the force of gravity? Sure there is lots of interest in understanding an underlying mechanism. But a useful model abstracts that away so you can predict behaviors.