There is a game studied in Philosophy of Science and Probably Approximately Correct (machine) learning. It's a cousin to the Looney Labs game "Zendo", but less fun to play with your friends. http://en.wikipedia.org/wiki/Zendo_(game) (By the way, playing this kind of game is excellent practice at avoiding confirmation bias.) The game has two players, who are asymmetric. One player plays Nature, and the other player plays Science. First Nature makes up a law, a specific Grand Unified Theory, and then Science tries to guess it. Nature provides some information about the law, and then Science can change their guess, if they want to. Science wins if it converges to the rule that Nature made up.