Behavioral economics has shown that individuals sometimes make decisions that are not in their best interests. This insight has prompted calls for behaviorally informed policy interventions popularized under the notion of "libertarian paternalism." This type of "soft" paternalism aims at helping individuals without reducing their freedom of choice. We highlight three problems of libertarian paternalism: the difficulty of detecting what is in the best interest of an individual, the focus on freedom of choice at the expense of a focus on autonomy, and the neglect of the dynamic effects of libertarian-paternalistic policy interventions. We present a form of soft paternalism called "autonomy-enhancing paternalism" that seeks to constructively remedy these problems. Autonomy-enhancing paternalism suggests using insights from subjective well-being research in order to determine what makes individuals better off. It imposes an additional constraint on the set of permissible interventions highlighting the importance of autonomy in the sense of the capability to make critically reflected (i.e., autonomous) decisions. Finally, it acknowledges that behavioral interventions can change the strength of individual decision-making anomalies over time as well as influence individual preference learning. We illustrate the differences between libertarian paternalism and autonomy-enhancing paternalism in a simple formal model in the context of optimal sin nudges.