Can computational models of human decisional autonomy help us formalize constraints on robot autonomy?