A framework for addressing fairness in consequential machine learning