Empirical studies on software evolution: should we (try to) claim causation?