Background independence and the causation of observations