Approaching fairness in machine learning

I have to say: I cannot wait to see what Moritz Hardt does here: "A third post will use our framework to interpret the recent controversy on COMPAS scores.”