I'd go as far as saying that this question makes your interviewing process to be more "cheat-friendly" - the chances of an interviewee not knowing of the proble

I'd go as far as saying that this question makes your interviewing process to be more "cheat-friendly" - the chances of an interviewee not knowing of the problem and solving it on the spot are a lot lower than those of him having heard of it and pretending to solve it on the spot. The first time I heard of it I was surprised about how counter-intuitive it seemed, but it didn't baffle me completely because I have a strong information theory background. From an information-theoretical point of view, it's not completely counter-intuitive that new information can change probabilities a posteriori, particularly when there is mutual interdependence.