Agency in self-assessment is a developmental achievement, not a starting point: why prior knowledge and expert structure are preconditions for student evaluative capacity

Calls to recentre self-assessment around student agency and identity have gathered momentum in higher education. We argue that the strongest versions of this call rest on an idealised assumption: that students can and should be the self-initiating leaders of their own evaluative activity. The evidence indicates otherwise. Accurate self-assessment depends on prior content knowledge, and students learning a subject for the first time do not yet possess the knowledge their accurate judgement requires. Agency in self-assessment is a developmental achievement, not a starting point. Students can and should be increasingly active in self-assessment, and with practice they can become genuinely agentic, but inaccurate agentic self-assessment harms learners more than expert-calibrated self-assessment does. We develop this position in four steps: distinguishing natural self-assessment from designed instructional practice; showing that self-regulated learning theory treats agency as an outcome of regulatory development, not as opposed to teacher expertise; establishing that prior knowledge is a precondition of accurate self-judgement, because minimal-guidance approaches overload learning; and setting out how to design for the development of evaluative agency. The framework treats agency, regulation, and expert structure as mutually constitutive, offering a basis for self-assessment design in a digital and AI-mediated era.