CONCEPT
Epistemic Dependence (AI)
The AI-era extension of Pariser’s analysis—the atrophy of independent productive capacity as the cognitive functions externalized to AI weaken without use, producing a vulnerability that is invisible during normal operation and catastrophic when the tool is removed.
The thought experiment is simple and uncomfortable: take a builder who has worked with AI tools every day for three years, remove the tool, and ask her to build what she built yesterday using only the skills she possesses independently of it. The result, Pariser argues, reveals a dependence that daily productive life conceals. Epistemic dependence on AI is not the same as dependence on a calculator or a spell-checker, which handle narrow instrumental functions while leaving the broad cognitive architecture of the user intact. The capacities being externalized to AI are broad and foundational: planning complex work, synthesizing information from multiple sources, generating original approaches to novel problems, organizing thought into coherent structure, evaluating the quality of one’s own output against intuitive rather than explicit standards. These are precisely the capacities that
[YOU] on AI identifies as what remains after AI handles execution—the “twenty percent” of judgment, direction, and taste. Pariser’s analysis locates the danger: