Learning the rules of a game: neural conditioning in human-robot interaction with delayed rewards