Agency, in the Oxford Rubric, is the extent to which AI use preserves and strengthens the autonomy and professional judgement of teachers, and the developing independence of students. It is the most distinctively educational of the five criteria — and the one that most directly asks whether AI is serving the human purposes of schooling, or quietly undermining them.

Of all the criteria in the Oxford Rubric — which also covers Safety, Efficacy, Accountability, and Transparency — Agency is the one that tends to get the least attention in school AI conversations. It's harder to measure than Safety. It's more philosophical than Accountability. And it surfaces tensions that most AI vendor pitches are designed to sidestep.

But it may be the most important criterion of all. Because Agency is where the question of what education is actually for gets asked directly.

What Does the Oxford Rubric Mean by Agency?

The rubric defines Agency as the extent to which AI use preserves and strengthens the autonomy, judgement, and developmental agency of both teachers and pupils. Its rationale is stated plainly: "Education is fundamentally a human endeavour. AI should augment, not diminish, the professional agency of teachers or the developing independence of learners."

Two distinct forms of agency are at stake here. Teacher agency — the professional autonomy to make informed educational decisions based on expertise and judgement — and student agency — the developing capacity to think, decide, create, and learn independently.

Both can be eroded by AI, and the erosion doesn't always look like a problem. It often looks like efficiency. It often feels like support.

The Deskilling Risk for Teachers

The Oxford Rubric explicitly names over-automation as a risk to teacher professionalism. When AI takes over the parts of teaching that require judgement — designing feedback, differentiating tasks, identifying what a student needs next — teachers who stop exercising that judgement regularly become less capable of exercising it independently.

This is deskilling. It's well-documented in other industries where automation has replaced skilled tasks. And it's a genuine risk in education, where the AI tools with the highest adoption rates are often the ones that automate the most cognitively demanding aspects of teaching: assessment, feedback, and planning.

The rubric's warning is not that AI shouldn't be involved in these tasks. It's that AI use must preserve the teacher's capacity to perform them independently — and must not create conditions where teachers feel unable to deviate from what the AI recommends.

The Dependency Risk for Students

The parallel risk for students is perhaps more familiar, but no less serious. Secondary school is precisely the developmental period when students should be building intellectual independence — the capacity to form arguments, evaluate evidence, manage uncertainty, and produce original work.

AI tools that do too much of that work — generating first drafts, structuring arguments, providing answers rather than prompts — don't just risk academic dishonesty. They risk something more fundamental: students who reach the end of secondary education less capable of independent thought than they would have been without the tool.

The Oxford Rubric captures this precisely: "Over-automation risks deskilling educators and fostering passive dependency in pupils." The word "passive" is well-chosen. The concern isn't just that students are using AI — it's that they're using it in a way that makes them passive recipients of generated content rather than active, struggling, developing thinkers.

What Appropriate Agency Looks Like

The indicators of appropriate Agency practice in the Oxford Rubric are:

• Teachers can adapt, refuse, or modify AI-supported processes without institutional pressure to comply.
• Students understand AI as a tool, not an authority.
• Students are taught to question and critically engage with AI outputs.
• AI use is framed explicitly as support, not substitution.
• Human judgement is augmented, not replaced.
• AI use supports independence and metacognition — students' capacity to reflect on their own thinking.

The failure indicators are: teachers who are deskilled or over-reliant on automation; students who defer thinking or decision-making to AI; AI outputs treated as authoritative or unquestionable; creativity, voice, or autonomy constrained; and dependency on the system increasing over time.

The Questions Secondary Schools Should Be Asking

• Do our teachers feel genuinely free to ignore, override, or modify AI recommendations? Or is there institutional pressure to follow them?
• Are we actively teaching students to engage critically with AI outputs, or implicitly encouraging them to treat those outputs as correct?
• Is AI use in our school increasing students' capacity for independent thought — or reducing it?
• Are there tasks or skills that teachers used to perform independently that they now find difficult without AI support?
• Do our students know when they're interacting with AI, and do they understand what it can and can't do?
• Is dependency on our AI tools increasing over time? What does that tell us?

That last question is one of the most revealing. The Oxford Rubric explicitly identifies increasing dependency as an indicator of inappropriate use. A school that couldn't function without a particular AI tool — or whose teachers couldn't perform key functions without it — has moved from augmentation to dependency. That's a red rating on Agency, regardless of how the tool performs on the other four criteria.

Agency Is Where Educational Values Show

The other four Oxford Rubric criteria — Safety, Efficacy, Accountability, and Transparency — can be substantially addressed through policy, governance, and documentation. Agency is different. It's enacted in individual classrooms, in the specific ways teachers use AI in their practice, and in the culture a school builds around AI as a tool rather than an authority.

This makes Agency both the hardest criterion to systematise and the most important to get right. Because it's the criterion that most directly answers the question at the heart of secondary education: are we developing independent, capable, thoughtful human beings? Or are we outsourcing that work to an algorithm?

The Oxford Rubric's answer is clear. AI should serve that developmental purpose. It should never replace it.

Frequently Asked Questions

What does the Agency criterion mean in the Oxford Rubric?

The Agency criterion assesses whether AI use preserves and strengthens the autonomy of both teachers and students. For teachers, this means maintaining professional judgement and the freedom to override AI recommendations. For students, it means developing independent thinking rather than becoming passively reliant on AI outputs. A use case that increases dependency or deskills teachers or students does not meet the Agency criterion.

How does AI risk deskilling teachers?

AI can deskill teachers when it automates the tasks that require professional expertise — such as designing feedback, differentiating learning, or assessing student needs — and teachers stop regularly exercising those skills independently. Over time, this can reduce teachers' capacity to perform those functions without AI support. The Oxford Rubric's Agency criterion requires schools to actively guard against this risk.

Should secondary school students be allowed to use AI tools?

The Oxford Rubric doesn't prohibit AI use by students — but it requires that any such use is framed explicitly as support rather than substitution, and that students are taught to engage critically with AI outputs. Use that encourages passive reliance on AI-generated content, reduces students' independent thinking, or treats AI outputs as authoritative does not meet the Agency criterion and should be reconsidered.