"As soon as we look at the nature of inference at this many-moves-ahead level of perception, our attitude toward probability theory and the proper way to use it in science becomes almost diametrically opposite to that expounded in most current textbooks. We need have no fear of making shaky calculations on inadequate knowledge; for if our predictions are indeed wrong, then we shall have an opportunity to improve that knowledge, an opportunity that would have been lost had we been too timid to make the calculations.

Instead of fearing wrong predictions, we look eagerly for them; it is only when predictions based on our present knowledge fail that probability theory leads us to fundamental new knowledge."

E.T. Jaynes's "Bayesian Methods: General Background"