Epistemic Debt: How Performed Competence Sustains AI Productivity Gains by Depreciating Human Value

While current discourse frames AI-assisted productivity as a straightforward augmentation of human capability, I argue that a significant portion of these gains is made possible not by the technology itself, but by the user’s capacity to bypass the cognitive and ethical friction that stems from performed competence, defined as the adoption of outputs one cannot evaluate, presented as though they reflect genuine understanding. Drawing on qualitative interview data and behavioral observation across professional domains, I label this phenomenon epistemic debt: the progressive accumulation of knowledge deficits incurred each time a user presents borrowed intellectual output as their own. My findings suggest that epistemic debt exists along two axial questions. The first axis is whether the user can identify an error in the AI’s output. The second axis is whether the user understands why a successful output succeeded. I further demonstrate that this phenomenon exists at scale because the interaction between user and model is private (i.e., using AI is a largely a solitary event), and because there exists no reliable social or institutional mechanism to detect the accumulation of epistemic debt, allowing it to propagate unchecked across roles, organizations, and industries. I conclude that productivity gains built on epistemic debt benefit companies by creating a spike in output produced by workers who, by persistently feigning competencies they have not developed, ultimately devalue themselves and become more economically vulnerable over time.