The promise and the price of autonomy

Why guidance, not premature independence, is the real foundation of freedom

Autonomy has a peculiar glamour in educational discourse. It sounds enlightened, progressive, humane, even virtuous. Give children independence and they will flourish. Loosen the reins and curiosity will bloom. Let them choose, explore, discover, and somehow the experience will alchemise into understanding. Supervision, by contrast, carries an air of mistrust. It feels fussy and faintly Victorian, as though the very act of watching might somehow diminish the child.

Autonomy flatters adults every bit as much as children. Professionals hear the word and imagine themselves elevated: trusted, liberated, recognised as masters of their craft. Autonomy promises dignity. It speaks to our desire to be left alone to get on with the work we believe we understand. Supervision, by contrast, feels like an intrusion. To be observed is to risk judgement. To be guided is to risk being corrected. No wonder autonomy has become such an easy banner to rally behind. But do we dress it up as principle when its real function is to provide cover for our insecurities about scrutiny?

A new study by Li Zheng and Yu Xiao, Supervised or autonomous? Exploring the effect of instructional strategies on elementary school students’ mathematics and science achievement, provides compelling evidence on the costs and benefits of autonomy vs supervision.

Before we analyse the paper, I want to get into the etymology of these two words.

The origins of autonomy

The romance of autonomy is potent because it gestures toward ideals we hold dear: freedom, self direction, trust. But if we are to take the concept seriously, we should start with what the word actually means. In a recent blog, Autonomy, authority and anarchy, Adam Boxer reminded me of the origins of autonomy. As is so often the case, the word comes from the Greek autonomia: autos, meaning self, and nomos, meaning law. To be autonomous is not to be free in the loose modern sense of doing as one pleases but the discipline of obeying a law one has made one’s own. The idea carries an echo of ancient cities declaring independence from their overlords, setting their own rules, establishing their own order. It is a story not of liberation from structure but of replacing someone else’s structure with one you are ready to take responsibility for upholding.

In the classical world autonomia was a political condition long before it became a psychological aspiration. Herodotus uses it to mark the dignity of a polis that governs itself rather than submitting to Persian rule. The loss of autonomia meant surrendering one’s customs, laws and identity. Thucydides sharpens the point with his typical bleakness. Athens grants its allies “autonomy”, knowing full well that without the power to defend it the word becomes a fiction. A city might keep its council and its name, but if it cannot resist its patron, its law has no teeth. Aristotle, treats autonomia as the sign of a city that has reached institutional adulthood: able to sustain its own order, able to stand without leaning on another’s will.

Across these writers the meaning is clear. Autonomia is never a synonym for free floating choice. It presupposes discipline, responsibility and readiness. It is a status that can be claimed only by those who have the resources to uphold it. A polis that declared autonomy before it had the institutions to maintain it was courting ruin. The Greeks knew this, which is why the term always carried an undertone of sobriety.

Somewhere along the way our imagination shed this original connotation. Autonomy has come to mean license, preference and unbounded self expression. Restoring even a hint of its classical weight changes how we view the concept. Autonomy was never supposed to be granted to the unprepared. It was something to be won, something that emerged only when knowledge, judgement and responsibility had settled firmly into place. Freedom without structure is not autonomy. It is drift. True autonomy arrives only when a person can bear the weight of their own decisions, when their inner law is sturdy enough to stand without someone else holding it up.

Seen through this older lens, our contemporary debates take on a new guise. The original meaning of autonomy has more meaning, more resonance. It reminds us that oversight is a precondition of freedom. Only when the scaffolding is no longer needed does it make sense to extend autonomy.

The etymology of supervision

To supervise comes from the Latin supervidere: super meaning above, and videre meaning to see. To supervise, at root, is simply to look over, to keep something in view. The earliest uses in medieval Latin carry no hint of micromanagement. They describe a guardian who watches over an estate, a steward who oversees a task, a bishop who keeps an eye on the conduct of his flock. The emphasis is on care, vigilance, responsibility. For instance, a bishop might be described as supervidens clerum, watching over the clergy in his diocese, not prying into their every movement but ensuring that the wider order of the church was protected. Monastic rules also use the term for practical stewardship. A cellarer could be tasked supervidere officinas, to oversee the workshops, meaning to keep an eye on stores, tools and labour so the community functioned smoothly.

There is nothing furtive or officious in the original word. It is not policing but watchfulness. The supervisor sees further because she has a better vantage. She watches not to catch someone out but to keep their flock from drifting into danger.

Only later did English attach the sour aftertaste of interference. By the seventeenth century the word acquires the shadow of managerial authority. For example, in early colonial records, officials are appointed to supervise the building of bridges and highways, meaning not only to watch over but to direct, instruct and, when necessary, correct the labourers. Similarly, in the minutes of trading companies such as the East India Company, merchants refer to officers set to supervise the factors abroad, a role that includes auditing accounts, enforcing procedures and reporting failures. The watching is no longer pastoral. but administrative, even disciplinary. However, seen from its root, supervision is not the opposite of trust. It is a form of care made possible by perspective. It says: I will keep an eye on potential hazards while you get on with your work.

The meanings we attach to these words stretch far beyond what our cognitive and moral realities can support. In practice, what adults call autonomy often collapses into little more than preference. I want to do it my way. I want to choose. I want to be unencumbered by oversight. It feels noble until we ask the awkward question: what if my way is wrong? What if my autonomy, exercised uncritically, becomes a source of harm to the people who must rely on me? Why should anyone have the autonomy to choose badly when others pay the cost?

This is why earned autonomy matters. It is not a rejection of freedom but a recognition of its conditions. Autonomy is worth having only when it rests on competence and only when it is bounded by shared knowledge and shared responsibility. That is as true for adults as it is for children.

Teachers are swept up in this moral script as strongly as anyone. Autonomy has been cast as the hallmark of professionalism, a signal that one has earned the right to be left alone. But it is a curious sort of freedom. In almost every other profession, autonomy is something granted only after sustained demonstration of expertise and always within a framework of standards. Surgeons cannot improvise on a whim and pilots cannot shrug off checklists. The freedoms that matter in expert practice are disciplined freedoms. The very idea that teachers should enjoy complete instructional latitude is seductive only because we rarely ask what that freedom permits. No one should have the autonomy to make bad decisions. A school that allows harmful practice to persist in the name of trust is not a sanctuary of professional respect, it is a directionless institution unwilling to confront its responsibilities.

This idea of earned autonomy creates a more honest picture both of professional practice and of students’ learning. Freedom growsfrom rather than precedes competence. Children cannot think mathematically until someone has shown them what mathematical thinking feels like. Teachers cannot make wise instructional choices without a secure grasp of how novices learn. Autonomy, when granted prematurely, becomes a kind of burden, a responsibility thrust upon someone who has not yet learnt to use the tools to exercise good judgment.

Anyway, after this long, rambling preamble, back to Zheng and Xiao’s recent TIMSS study on supervised versus autonomous approaches in primary mathematics.¹ This paper brings the argument into sharp relief. When lessons lean heavily on independent activity, achievement declines. When teaching is structured, monitored and deliberately guided, children learn more. The study’s dry statistical tables simply confirm what cognitive science has been saying for decades.

Beginners are not miniature experts; their schemas are rudimentary and easily overwhelmed by element interactivity, the sheer number of unfamiliar ideas that must be coordinated for understanding to take hold. In these conditions, autonomy is paralysing. The child who is left to puzzle through an inadequately scaffolded task is not experiencing freedom but cognitive overload. Supervision here is truly pastoral, stripping away the extraneous so that attention can cling to the right things.

Seen in this light, the moral intuitions that favour autonomy look increasingly fragile. We assume stepping back is a form of respect and stepping in is a form of control. But the direction of travel is the other way round. To guide a child is to acknowledge their dependence, to take responsibility for the complexity they cannot yet manage. To guide a teacher, likewise, is to protect the integrity of the curriculum and the wellbeing of the children who must rely on it. Autonomy offered too early is an abandonment of moral responsibility.

There is a deeper irony here. Advocates of independence often appeal to ideals of dignity and agency, yet genuine agency grows out of practice, imitation and the slow internalisation of structure. Kant spoke of freedom as obedience to a law one gives oneself. In Groundwork of the Metaphysics of Morals he says,

Freedom is independence from the constraint of another’s will, yet in such a way that one’s own will gives itself the law. … The will is not merely subject to the law, but subject in such a way that it must be regarded as giving the law to itself.

Or, as Bob Dylan put it in ‘Absolutely Sweet Marie,’ “To live outside the law you must be honest.” In education, that honesty emerges only after repeated exposure to good models. Until then, someone else must lend both clarity and direction.

The allure of autonomy will always be strong. It flatters the adult and reassures the child. But talthough independence is precious, it is something to be grown into. Extending premature autonomy may feel compassionate but it is functionally neglectful. Supervision, in contrast, becomes an act of care. It says, simply: I will stay close until you can stand on your own. Only then does freedom acquire the depth and resonance we want it to have.

1. TIMSS is the Trends in International Mathematics and Science Study, a large scale international assessment of students in Years 5 and 9 (Grades 4 and 8). It collects both achievement data and detailed questionnaires about teaching practices, classroom conditions and school contexts. The paper in question uses the TIMSS teacher questionnaire to classify instructional approaches into two broad clusters: supervised (structured tasks, teacher guidance, close monitoring) and autonomous (independent work, self directed problem solving, reduced teacher intervention). The authors analyse data across participating countries using multilevel modelling, which allows them to account for variation at the level of students, classes and schools. They then correlate these instructional profiles with mathematics achievement. Because the data are observational, the study cannot establish causation, but its design does allow for robust comparison of patterns across large, diverse samples.