Decades of research in political behavior show that citizens struggle with information, rely on cognitive shortcuts, and face barriers to participation. They are increasingly using AI tools to make sense of political information and express views, while interacting with platform curation systems that shape how views circulate through democratic politics. Although these tools appear suited to address the limits of democratic citizenship, they do more than transmit content. They help interpret information, articulate views, and affect what is prioritized, seen, and circulated. In doing so, these systems shape the formation of political judgments and citizens’ capacity to translate them into action. As a result, they can reduce barriers to participation while weakening the independence of political judgment, or they can support reflective preference formation while constraining citizens’ ability to act on those preferences effectively. We call this tension the autonomy paradox. To make sense of it, we focus on how citizens engage with AI systems in ways that affect belief formation and political action. We develop a framework and research agenda for identifying when AI supports autonomous citizenship, when it substitutes for citizens’ judgment or agency, and why the same systems produce different outcomes across political tasks depending on how citizens use them.
Anderson et al. (2026) studied this question.