Abstract
<p>Digital environments increasingly steer behavior through nudging, which alters choices without necessarily foreclosing any option. Digital nudges demand little effort from users but have been criticized for their opacity, limited adaptability, and potential to undermine autonomy, and their effects vary considerably across users and contexts. Self-nudging has emerged as an alternative emphasizing user agency, by which we mean the extent to which users themselves define the set of environmental changes available to them and select among those changes. We propose a framework that arranges (self-)nudging interventions along a continuum of user agency, with traditional nudging and standard self-nudging at its end points. Along it we distinguish four forms, namely other-directed, guided, recommended, and self-directed choice, differing in who defines the set of options for changing the environment and who selects among them. We argue that the appropriate level of agency depends on the heterogeneity of an intervention's treatment effects, on what researchers know about that heterogeneity, and on what users know about their intervention responses. The framework makes explicit the trade-offs between effectiveness, autonomy, and long-term maintenance of behavior change, guides the design and evaluation of digital interventions, and extends beyond digital settings.</p>