Heard someone say this recently and it has stuck in my head. We use models as abstractions of reality to capture some aspect of it in order to understand it, tame it and predict it. We do this subconsciously, all the time, without even realizing it.

We rely on models to a fault, expecting them to be right every time. Models stop working because the underlying reality changes. They stop working because they never fully captured reality anyways and their initial success in predicting outcomes was a happy coincidence.

All this and yet modeling is a very useful exercise. Trying to impose structure on observed reality makes us better thinkers. That in itself is a huge win.