Having an algorithm determine risk seems less like a method for correctly identifying risk and more like a method for offloading responsibility for poor decisions into a faceless, nameless entity which cannot be held accountable.