It's a distinction in machine learning, too: discriminative approaches vs generative approaches, calculating P(Y | X) "learning to assign a classification to a given thing" vs P(X | Y) "learning common features of things with a given classification".