Towards a computational model of human decisional autonomy and a formalization of constraints on machine autonomy