Predictivity and Stability in Classifications: some Comments on Recent Studies