Infringements of AI on Epistemic Autonomy: A Graded Approach

This chapter examines the tension between two epistemic goods: having instant access to accurate information through AI, and thinking for ourselves. If you can get any answer immediately from AI, is that copying rather than genuine understanding? We begin with Kant’s call to “use your own reason” and ask what this means—and whether we should even want it. After rejecting both extreme and weak epistemic egoism, we develop a graded, non-egoistic account of autonomy. We suggest that epistemic autonomy consists of stereotypical features that come in degrees: figuring things out yourself, trusting others with good reason, forming beliefs without bypassing your cognitive competences, and believing conscientiously. Using Alvarado’s framework of AI as epistemic technology, we then examine concrete cases. When nine-year-old Agnes uses Grammarly, her autonomy is minimal; when professional editor Agnes uses it, her autonomy remains high. The same AI system can enhance or diminish autonomy depending on the user’s competences and AI literacy. We identify “autonomy self-deception”—maintaining illusions of competence while developing genuine dependency. The opacity of AI systems poses particular challenges, requiring users to develop calibrated trust to preserve autonomy while benefiting from AI assistance.