Learning (Not) To Yield: An Experimental Study of Evolving Ultimatum Game Behavior