Theories are always wrong
but that doesn’t mean we can’t use them. It does mean, however, that we need to be careful. And some theories truly are completely wrong!
There’s the idea of an effective theory - a theory that works. It’s a lovely idea. Think back to your physics education. You’ve learned Newton and gravity, only to be told that all this stuff is kind of not true since we discovered relativity.
But that’s not really a good way of phrasing that. Relativity might be wrong, too! Newtonian gravity is an effective theory in the sense that it works! It will help you calculate almost anything unless you want to zoom in on quarks.
I learned that concept by listening to Lisa Randall, a theoretical physicist. However, the idea of effective theories extends into the realms of data science and machine learning. You must know the limits of your theories and models, but if you do, they can work and thus are effective.