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 mechanism

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.