Agent theory · November 15, 2026
Learner Agency in AI Classroom Agents
This essay examines how artificial intelligence agents in education can preserve learner agency, drawing on UNESCO guidance and self-determination theory to inform design choices for teachers and researchers.
Artificial intelligence agents are increasingly present in classrooms as tutors, teaching assistants, and coordinators of student work. This essay is for teachers and researchers who want to understand how the concept of learner agency shapes the effectiveness of these tools.
When an AI agent enters a classroom, it does not merely deliver information. It alters the dynamics of decision-making between the teacher, the student, and the curriculum. The central question is whether the technology expands or contracts the student’s capacity to act intentionally in their own learning. Learner agency—the ability of a student to make meaningful choices, direct their own efforts, and influence their educational environment—is the psychological mechanism that determines whether an automated system supports genuine development or simply enforces compliance.

The Pedagogical Stakes of Automation
The introduction of automated instructional agents raises immediate questions about control. A system that answers every question instantly, sequences every task without consultation, and evaluates every response against a fixed rubric may appear efficient. Yet efficiency in delivery does not guarantee engagement in learning. The UNESCO report Artificial intelligence in education: A review highlights this tension directly. The report discusses the ethical and pedagogical implications of deploying AI in classrooms, emphasizing the importance of preserving learner agency and motivation when using automated instructional agents. If a system removes all friction from the learning process, it also removes the opportunities for students to exercise judgment, tolerate uncertainty, and develop persistence.
For teachers, this means evaluating an AI tool not only by its accuracy but by its architecture of choice. Does the agent allow a student to select a topic, choose a format for demonstrating understanding, or decide the pace of progression? Or does it lock the student into a single linear path? The UNESCO report warns that without deliberate attention to these design features, schools risk deploying technologies that undermine the very autonomy they claim to foster. When an AI coordinates group work, for instance, it might assign roles based on algorithmic optimization rather than allowing students to negotiate responsibilities. That negotiation is where social and cognitive skills develop. Bypassing it saves time but costs agency.

Psychological Foundations of Agency
Learner agency is not a vague educational ideal; it is grounded in established psychological science. Self-determination theory, as defined by Deci and Ryan in Self-determination theory: A macrotheory of human motivation, development, and health, identifies three basic psychological needs that must be supported for individuals to sustain intrinsic motivation: autonomy, competence, and relatedness. Autonomy refers to the experience of volition and willingness. Competence involves feeling effective in one’s interactions with the environment. Relatedness is the sense of connection to others. An AI agent that ignores these needs will struggle to maintain student engagement over time.
Consider how an AI tutor interacts with a student working through a difficult mathematics problem. If the agent immediately provides the correct formula whenever the student hesitates, it may protect the student from frustration, but it undermines both autonomy and competence. The student did not choose to receive help, and they were denied the opportunity to experience the efficacy of solving the problem independently. Conversely, if the agent asks guiding questions, offers optional hints, and allows the student to decide when to use them, it supports autonomy while building competence. The student remains the author of their own progress.
Relatedness presents a distinct challenge for AI agents. While a machine cannot replicate human connection, it can either facilitate or obstruct the relationships that matter in a classroom. An agent that isolates a student behind a screen for extended periods diminishes relatedness. An agent that prompts collaborative discussion, shares student insights with the class (with permission), or helps a teacher identify who needs personal support can enhance the relational fabric of the room.
Designing Digital Environments for Volition
Translating these psychological principles into software design requires careful attention to how digital environments structure interaction. The article Motivation and cognitive load in educational technology: A self-determination theory perspective examines how digital learning environments can be designed using self-determination theory principles to foster autonomous motivation while managing learner engagement. The authors note that technology often inadvertently shifts a student’s motivational orientation from intrinsic interest to external compliance. Points, badges, and rigid progress bars can crowd out the internal satisfaction of mastering a subject.
For researchers studying AI in education, this literature suggests specific variables to measure. Rather than only tracking test score improvements, researchers should assess whether students feel a greater or lesser sense of ownership over their learning after interacting with an agent. Do students perceive the AI as a tool they wield, or as an authority they obey? The distinction matters because agency is fragile. Once a student learns that an AI system will do the thinking for them, the habit of intellectual passivity is difficult to reverse.
Teachers can apply these insights practically by setting boundaries around AI use. An agent might be excellent at generating practice problems or providing immediate feedback on grammar, but the teacher retains the role of framing the purpose of the lesson and inviting student input on goals. When an AI assists a teacher by organizing administrative tasks or summarizing student progress, it frees the teacher to engage in the high-agency work of mentoring and facilitating discussion. The technology serves the human relationship, not the other way around.
Ultimately, learner agency is the filter through which all educational technology must pass. An AI agent that respects autonomy, builds competence, and protects relatedness aligns with the core findings of self-determination theory and the ethical cautions outlined by UNESCO. For teachers and researchers, the goal is not to reject automation but to demand that it operate within the psychological realities of human motivation. When an AI agent leaves the student more capable of directing their own learning than it found them, it has succeeded.