I've found that I have benefitted from "making up" probabilities in some circumstances. I have an unfortunate but probably normal tendency to think something along the lines of, "Just as I would have predicted," whenever something happens which confirms my general attitude about that thing. Of course, I don't notice when the converse happens, and it doesn't occur to me to be surprised, and thus I don't update. Standard Confirmations Bias really.

But if I explicitly think, "There's a 70% chance Bob will be late to our meeting," and then Bob is early, I've provided myself with the useful and inescapable information that I may be miscalibrated about Bob's punctuality and may be selectively remembering one time his lateness cost me something. If I never tried to make an internalized probability assessment, I would not have been surprised in the same way. This is probably because I would cover up my mistaken nonverbal expectation of Bob's lateness with some other fallacious thought like, "Well, nobody is late all the time."