AI and Agency: Benefiting Users by Understanding and Serving Explicit Long-Term Aspirations

This paper introduces a framework for understanding how generative AI affects user agency, drawing on psychological research distinguishing between expressed long-term preferences and immediate, often-regretted, desires. We review research on how AI can increase users’ self-reported capacity for action and the benefits of providing users with a sense of control. Two studies with a nationally representative sample of American generative AI users examine a) users’ aspirational preferences for AI usage and b) where they perceive desired capability (outcome control) gains from AI (Study 1), along with c) users’ perceived control over AI experiences (process control) and d) the relationship between perceived control over AI experiences and reported capability gains (Study 2). Results from Study 1 indicate that users primarily value and report benefit from AI in instrumental domains (e.g. task completion, information gathering, and learning). Benefits in social and affective domains were found to be less common and less salient for users. In Study 2, users who felt more empowered to shape LLM outputs perceived greater control over their AI experiences, and also reported greater capability gains. We argue that centering user agency, encompassing control over product experience and product outcomes, should be a primary design principle.