Agentic AI in Education: Whose Agent? Whose Agency?
摘要
This article examines a linguistic shift with significant stakes for education. As the technology industry markets increasingly autonomous software as "agentic AI," it draws on a vocabulary (agent, agency, agentic) that has done centuries of work in moral and educational thought. We argue that agent and agency are different kinds of words. Agent is a role-word that extends easily to software, while agency is a rights-word whose moral weight was won through centuries of struggle to widen the circle of who counts as a person whose voice matters. When vendors sell "agentic AI" to schools, they borrow the moral weight of the second word to legitimize the extension of the first. Tracing this distinction through the history of agency as a hard-won right and through educational traditions that center learner and teacher voice, we show that so-called agentic learning platforms are structurally double agents, serving institutional principals under the vocabulary of serving the learner. The result is a slow erosion of the vocabulary educators need to defend student and teacher self-determination against the very systems now marketed under its name.