I was talking with a friend about this, and we concurred that the problem with next-gen is it's getting too close to real. With current-gen, graphics look like computer generated graphics. Your brain knows as much and accepts it. With next gen they're getting closer to real. Your brain tries to interpret it as real but finds it can't, because of slight discrepencies between real-world and CG. The end result is an image that can't be accepted as CG as it looks realistic, but can't be accepted as real because it has faults - producing an image that grates, instead of impresses.
Human brains can be very picky when it comes to realism. Even a photo realistic image with a couple of flaws like a dodgy shadow here or there will not rest easy on the eyes. Perhaps this will represent a difficulty for art appreciation next gen? Whereas non-photo like renders will be lapped up as the brain can accept them as artworks.