The Delegation Continuum: A Conceptual Framework of Cognitive Offloading and Epistemic Reliance in AI-Mediated Higher Education

Generative artificial intelligence (AI) differs from earlier educational technologies in a way that matters for learning: it can produce the reasoning itself, and not only support the learner’s reasoning. Analogies to calculators and search engines therefore understate both the benefits and the risks of AI-mediated study. This conceptual paper proposes the Cognitive Delegation Continuum (CDC), a five-level framework that locates any episode of AI use by where the reasoning is done and whether the learner keeps evaluative oversight of it: algorithmic verification, scaffolding and ideation, co-regulation and drafting, epistemic surrender, and harmful offloading and automation. The framework integrates distributed cognition, cognitive load theory and work on epistemic agency, and it separates productive cognitive extension from reliance that displaces the learning an activity was designed to produce. Five reusable tools accompany it: a comparison of educational tools with generative AI, the CDC matrix, diagnostic indicators of epistemic surrender, pedagogical counter-measures, and an empirical research agenda. The paper argues that delegation should be judged against learning goals and not against tool type, and that the strongest current evidence is short-term, correlational or self-reported. The CDC is a proposal for testing and not a validated model.