Nearly everything is "arbitrary" in the sense that you are using it. Wherever a machine learning classifier draws the line between categories is going to be arb

Nearly everything is "arbitrary" in the sense that you are using it. Wherever a machine learning classifier draws the line between categories is going to be arbitrary, but that doesn't mean it's not useful. Having some policy for ethical behavior is better than nothing.