Abstract
<p>Self-direction is foundational to adult and lifelong learning, yet generative AI can participate in identifying needs, setting goals, interpreting information, solving problems, producing outputs, and evaluating performance. This conceptual paper revisits Knowles’s six andragogical assumptions under AI-mediated cognitive delegation. Integrating adult-learning theory with cognitive offloading, extended and distributed cognition, human–automation function allocation, agentic information-system delegation, and appropriate reliance, it argues that behavioral independence is insufficient evidence of meaningful self-direction. Five analytical dimensions are derived: need and goal ownership, delegation control, epistemic calibration, cognitive recoverability and transfer, and motivational autonomy. Rather than treating technological dependence as the opposite of autonomy, the framework asks whether learners retain meaningful authority over the purposes, allocation, evaluation, and consequences of delegated cognition. The paper concludes with testable propositions, three applicability gates, and implications for adult education, assessment, AI literacy, system design, and lifelong-learning policy.</p>