Abstract: The explosion of empirical work on how and when human beings depart from perfect rationality has led to a wholesale rethinking of paternalism and its limits. Over the last decades, three camps have emerged: (1) coercive paternalists, who urge that behavioral findings undermine John Stuart Mill's Harm Principle and greatly strengthen arguments for mandates and bans; (2) libertarian paternalists, who urge that behavioral findings justify a host of freedom-preserving interventions or "nudges," such as warnings, reminders, labels, and automatic enrollment; and (3) antipaternalists, who urge that behavioral findings justify only, or at most, efforts to strengthen or "boost" people's competences, or their capacities to make good choices. On welfare grounds, it is possible to identify the assumptions under which one or another approach would be best. There are certainly domains where antipaternalism (alongside boosting) on the one hand or coercive paternalism on the other hand is best, but libertarian paternalism often has significant advantages, whether our focus is on welfare or autonomy.