I've personally accepted that it's basically predictable/deterministic, but due to how complicated and unknowable the system is there's no practical way for an outside observer to get all the information.

I'm guessing the lower resolution imaging methods might still allow more or less accurate prediction, though? We don't need to know the details on every air molecule to do fairly accurate weather forecasting, so maybe the same approach can work to predict mindweather. Maybe it's possible to know a person's brain well enough and accurately adjust predictions very fast after random encounters/events influencing them – like the people they meet, the things they see, and a myriad of other things – and in that way get something more and more capable of predicting behavior?

I don't really know much about either field, though.