Abstract
Artificial intelligence (AI) tools can suggest criteria, evaluate learners’ work, and recommend or carry out revisions, yet it is often unclear how learners direct these decisions when they study with AI. In this conceptual paper, we introduce metacognitive ownership: learners’ governance of standards, evaluative judgments, and actions within a learning episode in which AI or another external source contributes to at least one of these functions. Here, governance refers to how learners direct the use of that contribution. Drawing on self-regulated learning, human–AI regulation, agency, and evaluative judgment, we define three components: standards appropriation, monitoring-judgment governance, and control authorization. These components address the criterion a learner uses, how the learner reaches an evaluation, and what guides the next action. Control authorization also requires a realistic opportunity to act differently. A hypothetical example illustrates how a learner can evaluate AI advice and reach an incorrect conclusion, or accept correct advice without examining it. After distinguishing metacognitive ownership from related constructs, we set out four principles for interpreting evidence about each component. We then offer two tentative propositions about later advice decisions and continued learner regulation when support is reduced. The research agenda asks whether these interpretations are reproducible, add to existing explanations, and withstand evidence that challenges them. The framework provides a basis for examining learner-governed assistance; its relationships with learning require empirical evaluation.