Artificial Intelligence, Education, and the Responsibility to Become
Greg Culos
Author’s note: The ideas, experiences, memories, judgments, writing style, and structural decisions in this essay are my own. Artificial intelligence served only in an editorial role, assisting with organisation, testing, refinement, and preparation. Final responsibility for the argument and its expression remains mine.
Abstract
This reflective essay examines artificial intelligence in education by placing it within a much older ecology of human augmentation. Learners have always extended themselves through teachers, families, peers, books, tools, travel, culture, specialist expertise, and experiences beyond the classroom; AI is distinctive in its speed, reach, availability, and ability to produce complete-looking work, but it is not a separate or holistic force standing above that field. Beginning with the AI-supported redevelopment of a school handbook, the essay considers the growing separation between the quality of a finished product and the independent capability of the person presenting it. It argues that education should preserve a human baseline of knowledge and judgment while allowing ambitious, resource-rich production. A classroom-studio rhythm is proposed: the classroom develops and reveals foundational capacity through regular, often unassisted practice, while the studio gives that foundation reach through projects drawing upon the full legitimate spectrum of augmentation. Brief dipstick checks and a three-part assessment framework – independent demonstrated capacity, augmented production, and ownership and defence – reconnect the product with the learner without assuming infinite teacher time. The essay distinguishes mechanical friction from cognitive friction, defends the importance of lived and embodied experience without treating personal memory as universal proof, and resists binary responses shaped by fashion, fear, or novelty. By connecting AI with the discovery of controllable fire, it places the present technological leap within an older human pattern: every expansion of capacity opens a consequential space between power and wisdom, and responsibility for that space remains ours. AI should enlarge human agency without becoming the centre of the educational story or allowing the learner to disappear behind the artefact.
Keywords: artificial intelligence; education; assessment; augmentation; human development; agency; experiential learning; responsibility
Introduction: Moving the Goalposts
The public discussion of artificial intelligence is easily drawn towards poles. AI will save education or destroy it. It is either progress or cheating, liberation or dependency, a zero or a one with little patience for everything in between. Human beings have a strong attraction to arrangements of this kind. Poles simplify the field, give us somewhere obvious to stand, and spare us the slower work of proportion, context, contradiction, and judgment. Fads and trends benefit from the same tendency. They arrive with a complete vocabulary, demand an immediate position, and encourage us to confuse movement with truth.
Education cannot afford that luxury. The technology is already present and will become more capable, but novelty does not relieve us of older and more durable responsibilities. Fads ask where the movement is going. Bedrock asks what must remain true wherever it goes. In education, that bedrock is the developing learner: knowledge, judgment, agency, experience, self-belief, and the capacity to participate responsibly in a world no one can predict with precision.
This is not an argument against artificial intelligence, and it is not a celebration of it. I have no interest in pretending that the world can be returned to an earlier technological condition, nor do I believe education should preserve every familiar practice simply because it once appeared dependable. The useful question begins elsewhere: what relationship should education construct with a tool capable of producing work that may exceed the visible capability of the person presenting it, when the purpose of education remains the development of that person?
A second distortion needs to be corrected before that question can be answered. AI is frequently treated as though it were an independent and holistic construct, standing above and apart from every other influence upon learning. It is not. Students have always extended themselves through teachers, parents, siblings, tutors, peers, books, libraries, tools, technologies, travel, conversation, culture, specialist knowledge, and experiences acquired beyond school. AI enters that existing ecology as an unusually powerful source of augmentation. Its scale and speed matter. Its ability to produce complete-looking work matters even more. But it is still part of a larger field, not the whole field itself.
Once the goalposts move, the educational question changes. We are no longer asking only how schools should control AI. We are asking how education should understand, guide, and evaluate the full range of resources through which learners extend what they can do, while remaining clear about what they independently know, understand, and can still call their own.
I approach that question from the only position available to me: my own. Like anyone else, I perceive reality through a collection of experiences. Some have been successful, others not. Some were chosen; others simply happened. They include the people I have known, the places I have lived, the work I have done, the mistakes I have made, the responsibilities I now carry, and the problems currently sitting in front of me.
That perspective is partial. It offers no view from above and makes no claim to universal correctness. It does give me a source point from which to examine what I have observed and remain responsible for what I conclude.
At present, my responsibilities include leading a growing school, preserving the trust placed in us by families, supporting teachers and students through rapid change, and trying to ensure that development does not become confused with appearance. A recent experience brought the issue into clearer focus. Over several weeks, I worked intensively on the redevelopment of our school handbook. The handbook is not the subject of this essay. It is the experience through which the possibilities, weaknesses, and proper place of artificial intelligence became unusually visible.
The answer cannot be prohibition. It cannot be surrender either. Somewhere between those two easy positions lies the work.
1. The Handbook as a Practical Encounter
The handbook has been developing, in one form or another, since May 10, 2023, before the school itself opened. It has followed the institution through its early formation. New roles emerged. Programmes developed. Procedures were tested. Assumptions changed. Things that seemed clear at one stage became less adequate at another. Earlier versions of the document reflected earlier versions of the school. That is what a developing institution should look like.
A handbook of this kind is not simply a collection of information. It is part operational guide, part policy framework, part expression of institutional identity, and part navigational system through which the people inside the school can understand how its many areas fit together. It cannot be the viewpoint of one person. Every division, role, and area of responsibility begins from its own source point.
The institutional tapestry includes teachers, who experience the school through daily relationships with learners; academic leaders, who see progression, curriculum, standards, and the relationships among areas of learning; student services teams, who encounter needs, patterns, vulnerabilities, and realities that may remain invisible elsewhere; and operations staff, who understand the physical, administrative, financial, and logistical systems upon which everything else depends. It also includes specialised teams responsible for security, transportation, extracurricular experiences, health, communication, admissions, and family support; families, who understand the school through trust, communication, expectation, and the experience of their children; and students, who inhabit a reality adults can shape and protect but can never completely experience on their behalf.
Each perspective is partial. That does not make it unimportant; it makes it necessary. I think tapestry is the right word here. A mosaic is formed from separate pieces. A tapestry depends upon threads remaining distinct while being woven into relationship. Each thread has a source, a direction, and a contribution. Alone, it can only be itself. Interwoven properly, it becomes part of something stronger and more meaningful than any single thread could become on its own. Leadership in that context is not the act of flattening difference until everything sounds consistent. It is not taking many voices and forcing them into one voice. It is the more difficult work of understanding where perspectives originate, what each is responsible for, what must remain distinct, and how the different threads can become coherent without losing their integrity. That was part of what the handbook had to accomplish.
The handbook was designed to explain, delineate, orient, and guide; to create enough draw and engagement that people would actually enter it and use it; to make complex information accessible without pretending that the institution is simple; to show how individual policies, programmes, procedures, and responsibilities belong within a larger whole; to direct users to the people, systems, and resources where complete specialised knowledge properly resides; and to remain close enough to the school’s developmental reality to be useful. Creating draw did not make the handbook promotional. Nor was it defensive. It was almost the opposite.
The document was not designed primarily to persuade people outside the school that the institution was impressive. Its meaning is internal. A policy matters because it affects a child. A schedule matters because a student experiences it. A procedure matters because someone must rely upon it. A promise matters because a family has trusted us to honour it. A clear and engaging presentation may create confidence beyond the school. That is not undesirable. But it is an effect, not the purpose.
The purpose is use.
The handbook should help people understand the school they are already participating in. It should allow them to find what they need, recognise how responsibilities are distributed, and know where fuller information properly resides. That final point is important.
Orientation rather than centralisation
The handbook is not intended to contain every detail of every specialised area. Student services, academics, security, extracurricular experiences, health, transportation, and operations each possess their own complete bodies of information, expertise, procedures, and responsibility.
Trying to absorb all of that into one central document would not create clarity. It would create something enormous, difficult to maintain, and quickly detached from the reality it claimed to represent. The handbook therefore needed to serve as a source of orientation rather than a replacement for every source.
It points. It shows where knowledge lives. It identifies the people, systems, and resources through which complete information can be obtained. It provides an intelligible point of entry while allowing specialised knowledge to remain connected to those responsible for it. In that sense, it is both an account of the whole and a guide to the integrity of its parts.
Why pace mattered
A developing school cannot wait indefinitely for its explanatory systems to catch up with what it has already become. When roles, programmes, expectations, and responsibilities evolve more quickly than the documents used to explain them, confusion opens between the lived institution and its representation. At some point, the handbook had to be taken apart. Not revised politely around the edges. Taken apart.
Its earlier iterations had been useful, but the accumulation of information had begun to hide the whole. The task was to return to the fundamentals, reconsider the logic, and rebuild the document through a new organisational and navigational structure. Artificial intelligence made that work possible at a pace we had not previously been able to achieve.
What AI contributed
AI contributed by gathering and reorganising large bodies of material, comparing alternative structures, making provisional ideas visible quickly enough to be judged, testing navigation and relationships, supporting repeated revision without the cost of rebuilding everything manually, and shortening the distance between an intention and something concrete enough to examine. It also helped me move between the source points of the school without pretending to replace them: academics, student services, security, extracurricular experience, operations, family support, and the other specialised areas where complete knowledge properly resides. That is no small advantage.
There are times when an idea remains trapped not because it lacks merit, but because the cost of externalising it is too high. The time required to build one version prevents the creation of five. The labour required to test one structure discourages the consideration of another. Limited resources turn provisional arrangements into permanent ones. AI changes that. It permits iteration at a scale and speed that can be genuinely transformative. But nothing within the handbook existed because AI originated it.
What existed first
What existed first were the educational commitments, the operational knowledge, the perspectives of teachers, leaders, families, staff, and students, the institutional memory, the successes and errors accumulated through experience, and the trust the document was meant to serve. AI did not create those things. It helped us work with them. That distinction became clearer the more powerful the tool appeared.
AI could reorganise the information, but it could not know what the school meant. It could produce a polished sentence that was entirely wrong for our reality. It could create a visually coherent page that concealed an important relationship. It could resolve one inconsistency while introducing another. It could confidently present something that did not honour the source from which the information arose. Those judgments remained human because the responsibility remained human. I had to know what I was looking at.
I had to communicate with the different parts of the school, understand their needs, recognise where their perspectives intersected, and decide whether the whole remained faithful to the promise we make to families. The usefulness of the tool did not reduce that responsibility. It increased it. The faster the production became, the more closely I had to observe. The more effectively AI amplifies contribution, the more important it becomes that the contribution being amplified is real. The handbook did not spring from the tool. It emerged from nearly three years of work, discussion, revision, practical consequence, and the distinct perspectives of the people who constitute the school. AI entered late in that story and helped us see and organise more of it at once.
2. AI Within an Ecology of Augmentation
One of the problems in the current discussion is that artificial intelligence is often treated as though it arrived from outside the history of learning. It is spoken about as a complete educational environment in itself, almost a rival institution: teacher, tutor, library, editor, researcher, and producer rolled into one. Its range encourages that perception, but the perception is misleading. AI did not invent assistance, unequal access, external influence, or the expansion of human capability through resources beyond the self.
Students have never entered a classroom as isolated units. They arrive carrying accumulated worlds. One has parents who read widely and ask difficult questions at dinner. Another has a sibling who has already taken the course. One has travelled, crossed languages, handled tools, cared for animals, or listened to adults discuss work. Another has access to a tutor, a quiet room, a library, specialist equipment, or the confidence that comes from knowing how schools operate. There are also peers, coaches, religious communities, cultural traditions, search engines, calculators, software, museums, workplaces, and the endless accidental teachers encountered in ordinary life.
The classroom has never controlled this ecology completely. It cannot, and probably should not try. Education has always had to assess students whose work carries invisible traces of other people and other places. AI changes the scale, speed, availability, and opacity of that augmentation. It can make elite forms of assistance widely accessible, which is no small democratic possibility. It can also produce a finished surface so complete that the learner’s contribution becomes difficult to see.
That is the shift. AI is not a new category above all others; it is a powerful new member of an old family. Treating it as the whole family distorts the discussion and, perhaps more importantly, keeps the technology at the centre when the learner should be there.
Fire and the void
In earlier writing, I used the discovery of controllable fire as perhaps the clearest example of the position AI now occupies in the escalation of human ability. Fire allowed us to reach beyond the limits of the body: warmth, protection, cooking, light, movement into harsher environments, craftsmanship, metallurgy, industry, and eventually much of the constructed world. It did not simply make an existing task easier. It changed the range of what human beings could become capable of doing.
The same fire could nourish or destroy. It could protect a community, clear land, consume a forest, or erase the community itself. Nothing within the flame supplied the moral direction. Every leap in capacity opens some version of a void between what we are newly able to do and what we are wise enough to do with it. At the far edge of that void is the possibility of destroying ourselves. AI is opening that space again, not as a separate intelligence hovering above the human story, but as another tool through which human ability has been abruptly enlarged.
It expands memory, speed, synthesis, production, reach, and access. It may also expand error, dependency, manipulation, concentration of power, and destruction. The temptation is to speak as though the technology owns the danger or as though the direction has already been decided by progress itself. Neither is true. The consequential decisions remain ours: what we develop, what we permit, what we refuse, what we delegate, and what capacities we preserve in ourselves while using the power we have created.
We own that void.
The stories I tell later in this essay matter particularly for this reason. They are not decorative memories offered as universal proof that every person learns in precisely the same way. They show the source point from which I understand development and expose the fiction that capability is ever formed in isolation. My father’s corrections, my grandfather’s tools, the rock walls, Italy, farm work, karate, mountaineering, employment, mistakes, and responsibilities all entered the classroom through me. They were forms of augmentation long before the word was attached to a machine.
The educational question, then, is not whether influence from beyond the learner should exist. It already does. The question is what that influence does to the person. Does it deepen judgment, enlarge experience, and create future independence? Does it simply improve the immediate product? Or does it become so complete that the person disappears behind it?
3. The Person Behind the Work
A student is not a handbook. The comparison has limits, and it should. But the experience raised a question that moved quickly beyond the document itself.
If AI can produce something polished, coherent, impressive, and apparently complete, what exactly does the final product tell us about the person behind it? Schools have traditionally relied upon a fairly simple relationship between work and capability. A strong essay suggested a strong writer. A sophisticated project suggested sophisticated understanding. A well-structured presentation suggested that the student had organised the knowledge represented within it. That relationship was never perfect. Students have always received unequal forms of help.
The wider ecology becomes easier to see when I look at my own education. My father was an English teacher. When I brought writing home, I had access to support another student may not have had. He could recognise a weak sentence, identify a problem in structure, or ask the kind of question that improved the work. But that was only one part of the advantage.
I grew up around books, language, argument, practical work, and adults who expected participation. My father taught me some Latin. At thirteen, I spent several months in Italy with family, participating in the grape harvest and living inside another language and social environment. We lived on a small farm. I helped care for animals, work in gardens, build things, and dig irrigation. I also collected rocks.
My father built dry rock walls. I would go out repeatedly and return with more material, increasing the size of the pile at his disposal. I did not always know how a particular stone would fit. That was his work. He could see possibilities I could not. He understood where a rough edge might hold, where weight had to settle, and how one irregular piece could support another. Those walls have now survived generations. None of these experiences appeared directly on a school assignment, yet all of them shaped the person producing it. I do not offer them as a laboratory proof of how all learning works. I offer them because they make visible how much of a learner is formed beyond the formal lesson and beyond the teacher’s line of sight. Education has never taken place only within schools. Student work has never emerged from equal conditions.
The democratising possibility
A student without a tutor can ask for an explanation at any hour. A learner working in a second or third language can receive help expressing an idea. A child who has never seen a particular kind of project can receive examples, models, and guidance. A learner can enter areas of knowledge that previously appeared inaccessible. We should not be so preoccupied with the risks of AI that we fail to recognise this democratising potential. But the scale of assistance has changed.
A student can now produce work that may bear almost no relationship to the student’s present understanding. The language may be mature, the reasoning well organised, and the presentation highly sophisticated while the learner remains unable to explain the vocabulary, reconstruct the argument, defend the conclusion, or reproduce even a small part independently. The product may be excellent. The learning may be almost absent. That is where the difficulty lies. The wrong response is to prohibit the tool. The equally wrong response is to decide that because the tool can perform the task, the underlying human capacity no longer matters.
Two questions that must remain separate
What can the learner create with the resources available? This reveals augmented reach, creativity, ambition, and the ability to use teachers, peers, family knowledge, books, technology, AI, specialist expertise, and contemporary tools intelligently. What does the learner independently know, understand, and remain capable of doing? This reveals the foundation from which responsible judgment and future growth remain possible. Both are valid questions. They are not interchangeable.
A student should be allowed to produce excellent work. There is value in seeing what is possible, in participating in ambitious creation, and in feeling pride in a result that may have been unreachable without assistance. We should not deliberately require inferior products merely so that authorship is easier to verify. But we cannot allow the product to replace the person. A child is not valuable because the work appears impressive. A learner is not simply a producer of school artefacts. There is something irreducible in the individual that education must preserve as its priority. I have used the word sanctity for this, and I mean it in the broadest sense. The person is not merely the means through which the artefact is produced.
The person is the purpose of the educational process.
4. Assistance, Dependence, and the Human Baseline
AI can teach. That point should be stated plainly. It can explain concepts patiently, generate examples, translate language, compare interpretations, identify weaknesses, and respond to questions that a student may be reluctant to ask publicly. It can help learners enter areas of knowledge that previously seemed inaccessible. It may become one of the most powerful educational supports ever created. But support is not automatically development. There is a difference between assistance that builds capability and assistance that conceals its absence. Assistance develops capacity when it changes what the learner can subsequently do, requires decisions rather than passive receipt, makes thinking visible and open to correction, gradually reduces the need for the same level of support, and increases future independence. A teacher models how to organise an argument, and eventually the student can organise one independently. A parent helps revise a paragraph, and the child begins to recognise the same weakness in later writing. A coach demonstrates a movement, watches the attempt, offers correction, and slowly withdraws support. The assistance succeeds when the learner becomes less dependent upon it. AI can work in the same way. It can ask questions rather than provide finished answers. It can challenge a student’s assumptions. It can help compare two structures, reveal an inconsistency, or offer feedback on a draft the learner has already produced. It can also bypass the process entirely. A student can enter an assignment prompt, receive a complete response, make a few superficial changes, and submit it. The work may receive a strong evaluation. The student may feel successful. Yet little may have changed in the person. This is often framed primarily as dishonesty. That is too narrow. The deeper problem is dependency. A learner may appear increasingly capable while becoming less able to function without the tool. The surface improves as the foundation thins.
Questions more useful than “Did the student use AI?”
What did the AI, tutor, parent, peer, specialist, or other resource do? What did the learner do? Which decisions remained with the learner? What did the learner understand? What can the learner now do that could not be done before? Can the learner identify an error in the AI’s response? Can the learner disagree with it? Can the learner revise or reject what was produced? Can the learner explain why the final work deserves to exist? Did the assistance increase future independence, or did it simply improve the immediate product? That final question may be the dividing line.
The baseline
Technological advancement does not mitigate what people should know as a baseline. This idea seems simple, but education has been moving away from it for some time. Whenever a tool becomes capable of performing a task, there is a temptation to conclude that the underlying human capacity is no longer necessary.
Calculators did not eliminate numerical reasoning. GPS did not eliminate orientation. Search engines did not eliminate knowledge. Spellcheck did not eliminate language. AI does not eliminate thought. A person does not need to memorise every fact. No one ever did. But without some internal structure of knowledge, judgment becomes impossible. You cannot recognise an anomaly if you have no sense of what is normal. You cannot identify a false premise if you have no understanding of the subject. You cannot determine whether a conclusion follows from evidence if you have never learned how reasoning works. You cannot delegate judgment responsibly if you are incapable of exercising judgment yourself. That last point matters.
Do not delegate a judgment you cannot exercise.
I do not mean that people must become experts in everything before using AI. That would be absurd. We rely constantly upon knowledge we do not personally possess. I cannot build every system I use, diagnose every mechanical problem, or reproduce the expertise of every professional whose judgment I trust. But trust is not the same as surrender. Some baseline must remain.
Enough knowledge to ask a meaningful question. Enough understanding to recognise when something does not fit. Enough judgment to know when another source is required. Enough humility to stop when we do not know. The Cadillac is useful here.
A person may buy a Cadillac without understanding every system that makes it function. The vehicle represents the ingenuity and expertise of countless people. The buyer trusts the brand, the engineering, the price, and the unseen machinery. Most of the time, that trust is justified. Then one small system fails, and the entire car is stranded. The driver may know that the car has failed but have no idea why. There is nothing shameful in that. Modern life depends upon specialisation. The educational problem begins when students are assessed as though they built the Cadillac because they arrived in one. A polished, heavily augmented product may reflect enormous sophistication. AI makes the question newly urgent because it can supply so much so quickly, but the question itself is older: whose sophistication is it?
If the learner cannot explain the systems, identify the decisions, or recognise the failure, then the quality of the vehicle tells us little about the capability of the driver. Education must preserve enough internal structure for students to remain responsible participants in what their tools produce.
The baseline includes reading comprehension and independent written expression, numerical and scientific reasoning, historical and cultural understanding, oral communication and ethical reflection, the ability to organise, compare, filter, and apply information, and physical and practical competence: observing, making, repairing, persisting, collaborating, and adapting. These are not nostalgic remnants from an earlier age. They are what make amplification possible. AI should extend capability. It should not conceal its absence. Foundation before amplification. That phrase may be too neat, but I think it holds.
5. Process, Failure, and Formative Resistance
One of the dangers of AI is not simply that it produces answers. It produces the appearance of arrival. The paragraph appears complete. The image appears finished. The code arrives in working form. The design looks as though it has already passed through the stages that ordinarily precede it. The journey disappears. Yet much of learning lives inside that journey. People develop through trying, failing, revising, testing, waiting, reconsidering, and beginning again.
Development through iteration
An idea becomes an attempt. The attempt meets reality. Reality exposes what the idea failed to anticipate. The person sees differently. The next version becomes possible.
The second version is not simply the first version corrected. It is evidence that the person has become capable of seeing what could not be seen before. That is why process matters. The final product does not exist despite the struggle. It exists because of it.
I have seen this repeatedly in the handbook project. A version is produced. Something is wrong. The error may be obvious or almost invisible. A page number shifts. A navigation link points to the wrong location. A section appears balanced but misrepresents the relationship among its parts. Each problem reveals something. Sometimes the error is local. Sometimes it exposes a weakness in the structure. The mistake is not always the most important part. What matters is what it teaches us to see.
What testing should reveal
The most useful question is not always, “Did it work?” Sometimes it is, “What did it reveal?” Education should understand this instinctively. A first draft is not a failed final draft. An incorrect hypothesis is not the failure of inquiry. A student struggling with a concept is not necessarily incapable. A developing school is not deficient because it does not yet resemble an institution that has existed for generations. We have become too willing to judge present reality against imagined perfection. That comparison is unfair because the imagined version has never had to survive reality. It has no costs, contradictions, compromises, or unintended consequences. It is perfect because it has not yet been required to exist. The present version, however imperfect, is real. It has weight. It can be examined. It gives us somewhere to stand.
The foothold
Mountaineering taught me something similar. Wherever I was standing, if the position was secure, I had time. I could look around. I could consider another route, wait for conditions to change, move sideways, or retreat. The summit did not make the foothold meaningless. The foothold made the next decision possible. This is how development should be understood.
Where is the learner now? Is the position secure? What has become possible? What is the next sensible move? What has the previous attempt revealed? What support is still necessary? What support can now be removed? The unfinished state is not a failed identity. It is where development happens. AI can either support that process or hide it. That depends upon how we use it.
Mechanical and cognitive friction
Modern life is increasingly organised around the removal of friction. Much of that is welcome. There is no educational virtue in forcing students to spend hours on a mechanical task that a tool can perform instantly and accurately. Manual formatting does not automatically deepen thought. Repeating a calculation already mastered may add nothing. Converting references from one citation style to another can consume time without improving the quality of the judgment behind the research. No medals are awarded for needless inconvenience.
Still, not all resistance is waste. Some of it is the very place where capacity develops. Wrestling an argument into shape, interpreting difficult language, identifying a logical flaw, deciding what evidence deserves trust, or confronting the failure of an idea all demand more than endurance. They require the learner to change.
The distinction I find useful is between mechanical friction and cognitive friction. Mechanical friction is the drag created by routine operations: transcription, repetitive formatting, surface correction, retrieval, and procedures whose underlying principle is already secure. AI can remove much of this without damaging learning and may free considerable time for work that matters more.
Cognitive friction is different. It appears when the learner must decide, interpret, structure, test, compare, doubt, revise, or defend. It is the resistance encountered when reality does not cooperate with the first idea. Clear that away too quickly and the task may remain impressive while the learner remains unchanged.
The categories are not fixed forever. What begins as cognitive friction can become mechanical once fluency develops. A child learning multiplication may need repeated unaided practice; an engineer does not prove seriousness by refusing a calculator. The educational judgment lies in knowing what the learner is still forming and what can now be safely accelerated.
Friction for friction’s sake is no more educational than convenience for convenience’s sake. The point is not to preserve the labour. It is to preserve the learning that sometimes occurs within it.
If AI writes the essay, organises the evidence, selects the examples, and settles the conclusion before the learner has entered the problem, it has removed the cognitive encounter the assignment was meant to provoke. If it helps compare two structures, exposes a contradiction, or questions an assumption after the learner has attempted the work, it may deepen that encounter. The question, then, is not whether the tool made the task easier. It is whether the difficulty that remained was the difficulty that mattered.
6. Embodied Experience and Experiential Capital
Human beings are physical. This should be obvious, but education has increasingly behaved as though learning were mostly the movement of information from one location to another. It is not. We develop through bodies, environments, relationships, responsibilities, risks, and consequences.
As a young person, I climbed mountains, trained in karate, worked on a farm, travelled, made mistakes, held jobs, and participated in practical tasks with adults whose knowledge was expressed as much through action as through language.
My grandfather’s tool shed
My grandfather taught me in his tool shed. The memory comes as a whole: the tools, the sharpening stone, the preparation, his hands, the act of watching, and eventually the construction of a mousetrap. The object was small. The learning was not.
It placed me inside a relationship between generations. It showed that tools require preparation. It made visible the movement from understanding to function. The trap either worked or it did not.
Reality answered.
The rock walls
My father’s rock walls taught something else. I gathered stones. He placed them. At the time, I may have thought my part was simply to make the pile larger. Yet the wall depended upon that repeated contribution. He then fitted each irregular piece into something he could already partly imagine but could only complete through contact with the stones themselves. The wall was not imposed upon the material. It emerged through relationship with it.
Embodied lessons
Mountain climbing taught the importance of where one is standing. Karate taught restraint, timing, and the difference between force and aggression. Work taught consequence. An accident with a motorcycle at thirteen led to the expectation that I would earn money and assume responsibility. I became a dishwasher in a Chinese restaurant and continued working through my teenage years. None of these experiences can be replaced simply by a description of them. That does not mean mediated experience is worthless. Simulation, virtual reality, carefully designed scenarios, and AI-supported tutoring can create forms of rehearsal that would otherwise be impossible, dangerous, or inaccessible. A flight simulator can develop real skill. A virtual patient can permit repeated diagnostic practice. A language learner can enter conversations that would not otherwise occur. Those are experiences, and they can matter enormously. But they remain designed encounters with different consequences, textures, and limits from the world they represent. The distinction is not real versus unreal so much as what kind of experience is being provided, what responsibility it carries, and what the learner can transfer beyond it. AI can explain a garden. It cannot take responsibility for keeping something alive. It can describe engineering. It cannot make a structure bear weight. It can generate a climbing route. It cannot feel weather changing, fatigue accumulating, or rock shifting beneath a hand. It can write about courage. It cannot stand in the place where courage becomes necessary.
The educational responsibility
This is why gardens, hydroponics, engineering labs, makerspaces, performance, art, outdoor education, service, and physical challenge are not decorative additions to schooling. They bring learners into contact with reality. A school cannot equalise every childhood.
It cannot change the stars under which each child is born. Some students arrive with travel, books, adult conversation, tools, nature, language, and responsibility already embedded in their experience. Others do not. Schools often mistake access for aptitude.
A child who has never used a tool may appear less capable than one who has spent years around adults making things. A student who has never travelled may have less contextual knowledge than one who has crossed cultures since infancy. A learner who has never been expected to care for something may initially struggle with responsibility. These are not necessarily differences in potential. They are differences in experiential capital. AI may narrow an information gap and, through simulation and guided practice, may narrow parts of an experience gap as well. It cannot erase the difference between information about a responsibility and being responsible, between observing consequence and carrying it, or between rehearsing an encounter and living with what follows. That is where schools carry a particular responsibility. We cannot reproduce every possible childhood, nor should we manufacture hardship for its own sake. But we can widen access to experience.
Students should plant, build, test, repair, travel, perform, negotiate, contribute, and care for things that depend upon them. They should encounter materials that resist them and people who see the world differently. They should experience the distance between an idea and a thing that actually works. The more powerful artificial systems become, the more deliberate we may need to be about protecting direct encounters with the world.
7. The Generational Spiral
There is another concern that predates artificial intelligence but may now become easier to hide. Over the past several decades, education has sometimes reduced difficult knowledge or practices because students found them uncomfortable, discouraging, or contrary to immediate preference. Some of those changes were necessary. Education has often been unnecessarily harsh, repetitive, exclusionary, and humiliating. Difficulty alone does not make an experience worthwhile. But the rejection of harmful difficulty can gradually become a rejection of difficulty itself.
The spiral
A subject is experienced as demanding. The demand is reduced. Students receive less of the underlying knowledge. Some later become teachers. Because they never developed confidence in that knowledge, they feel less comfortable teaching it. The content is simplified again. The next generation receives even less. This is not quite a cycle. It is a spiral because the capability available at each return may be lower than before. Of course, curricular erosion has other causes – policy, funding, workload, testing regimes, institutional fashion, and changing social expectations among them. I am not trying to squeeze the whole history of education into one explanation. I am naming a mechanism I have watched operate. The danger is not merely that certain facts disappear. What one generation ceases to teach, the next may cease to know. What it ceases to know, it may eventually cease to value. What it ceases to value, it will no longer feel responsible to preserve. AI can disguise this erosion.
Teachers and students may both produce polished materials that create the appearance of competence beyond what either could independently demonstrate. The presentation improves while the underlying knowledge thins. That possibility should concern us. Schools need to think about adult development as well as student development.
Professional learning cannot focus only on new tools, trends, and strategies. It must also strengthen what teachers themselves know, how they reason, and how confidently they can guide students through demanding material. Educators cannot preserve foundations they have not been supported to develop. Care does not mean removing every struggle. Care means making struggle meaningful, supported, and possible to overcome. There is a difference.
8. Assessment in the Age of AI
The final artefact can no longer be treated as sufficient evidence of the learner. That does not require abandoning projects, research, creativity, collaboration, or sophisticated production. Students should be allowed to create excellent work through the full legitimate range of resources available to them, including AI. They should see what is possible, experience the pride of ambitious creation, and learn to use the tools that will shape the world they are entering.
The polished artefact simply has to sit within a wider body of evidence. Assessment now needs to see three things: what the learner can do independently, what the learner can create with amplification, and whether the learner remains present inside the work.
The classroom and the studio
A useful way to organise that relationship is through a rhythm between the classroom and the studio. These are contextual locations rather than necessarily separate physical rooms. The same space may become one or the other within a single lesson.
The classroom is where foundational capacity is practised, observed, and strengthened. Work here is often short, regular, low-stakes, and at times deliberately unassisted. Students write, calculate, explain, recall, map, question, and respond. The purpose is not surveillance. It is to make development visible and to give the learner enough internal structure to judge what larger tools later produce.
The studio is where that foundation is given reach. Students pursue sustained and ambitious work using the legitimate ecology of augmentation: teachers, peers, families, libraries, field experience, specialists, collaboration, technology, software, AI, and whatever other resources are appropriate. Here they design, synthesise, create, test, revise, and produce something that may exceed what they could have made alone. The classroom establishes and checks the foundation. The studio gives it reach.
Neither is a reward for enduring the other. The relationship is reciprocal. Classroom practice prepares students to enter the studio responsibly. Studio work gives purpose to classroom learning and exposes new weaknesses, questions, and possibilities. What emerges returns to the classroom as a stronger baseline: something the learner can now explain, transfer, repair, or do independently. The rhythm is therefore classroom practice, studio production, ownership and defence, and then a return to the classroom with enlarged capacity. Studio does not mean unsupervised output. It needs visible intentions, a few decision points, contact with the teacher or peers, evidence of revision, and some final act of explanation or transfer. None of this has to become a bureaucratic parade. A studio should feel open, but open is not the same as unobserved.
Independent demonstrated capacity
Independent demonstrated capacity asks what the learner can do without assistance under conditions in which authorship and understanding are visible. This need not mean constant high-stakes examination. A ten-minute handwritten response can show whether a student can organise thought. A short oral explanation can reveal whether the vocabulary is understood. An unfamiliar problem can test transfer. A concept map can expose the relationships a learner actually sees. A brief calculation can reveal fluency. One repaired paragraph may tell more than another full essay.
Dipstick checks and the reality of teaching
Any assessment proposal that ignores classroom scale deserves suspicion. A secondary teacher may be responsible for 150 students. No serious framework can depend upon conducting a thirty-minute oral defence after every assignment or adding another mountain of paperwork to work that is already difficult to sustain. Fortunately, that is not necessary. Teachers do not need to inspect everything. They need to sample intelligently.
A two-minute dipstick conversation can be enough. Ask a student to explain one decision, define one term, repair one weak section, sketch the logic of the argument, or apply the central idea to an unfamiliar case. Rotate the checks. Listen while students work. Collect one unassisted paragraph rather than another full submission. Ask for a concept map before the project begins and again after it ends. Small samples, chosen well and accumulated over time, have high predictive value.
Teachers already make judgments this way. They notice how a student speaks, hesitates, asks questions, solves a problem, changes direction, or responds when an answer does not fit. AI makes that professional observation more important; it need not turn it into bureaucracy.
Augmented production
Augmented production asks what the learner can create responsibly through the full range of legitimate resources available. AI belongs here, but it does not define the category. So do collaboration, teachers, tutors, family knowledge, books, libraries, software, specialists, cultural experience, travel, fieldwork, and other forms of assistance that have always surrounded student work. This is where ambition, synthesis, design, communication, creativity, and application become visible. Where AI is involved, students should understand prompting, comparison, verification, revision, disclosure, and the responsibilities that accompany generated material. In every case, they should understand the nature of the help they used and the decisions that remained theirs.
Good AI use is not measured by how quickly the system supplies an answer. It is measured by how intelligently the learner directs, questions, evaluates, and improves what is produced.
Ownership and defence
Ownership and defence ask whether the learner remains present inside the work. Can the student explain the decisions that shaped it? Identify where AI and other significant resources contributed? Recognise a limitation? Repair a weak section without asking the system to start again? Explain why one source deserved more trust than another? Respond to an unanticipated counterargument? Transfer the underlying idea into another context? Point to something the AI produced that should have been rejected?
Ownership is the bridge between independent capacity and augmented production. Without it, amplification becomes substitution. With it, AI may become a genuine extension of the learner.
These are not necessarily three unrelated grades. They are three views of the same developing person. One reveals the foundation. One reveals the reach of the tools. One reveals whether the learner remains present inside the work.
From detection to design
Many schools are treating AI primarily as an academic-integrity problem. That response is understandable. Teachers want to know whether work is authentic, schools want fair assessment, and students need clear expectations. Detection, however, will become increasingly unreliable, adversarial, and exhausting. Better design is the more durable answer.
A task that can be completed convincingly by placing its prompt into an AI system and copying the result may no longer be an adequate task. That does not mean every assignment must become more complicated. Probably the opposite.
Keep it simple
Ask for visible thinking. Ask students to explain decisions. Build a few checkpoints into the process. Use provisional drafts, short conversations, practical demonstrations, reflection, and unfamiliar transfer tasks. Make AI use transparent rather than automatically shameful. Clarify which parts of the work may be assisted and which must be independently demonstrated.
Most importantly, know why the task exists. What capacity is it intended to develop? Where does the learning take place? Which forms of assistance deepen the process, and which bypass it? Without those answers, AI policies will become collections of restrictions built around uncertain purpose.
Do not build endless hedges around a weak structure
Schools have a habit of building hedges around deficient structures. A system stops working properly, so we add another rule, then another form, another declaration, another exception, perhaps a committee and a layer of surveillance for good measure.
Sometimes those measures are necessary. Even so, every hedge should raise a question: what deficiency is this protecting us from, and can the deficiency itself be redesigned? Managing a weak structure is not the same as creating a stronger one. AI may require schools to reconsider assessment at its foundation rather than police an older model more aggressively.
9. The Educator’s Responsibility
The educator’s role is not to compete with AI or to protect students from every difficulty. It is to understand the learner well enough to know what must develop next. That requires attention, but it should not require a second administrative life. A teacher who regularly sees a student write, speak, reason, calculate, build, collaborate, and respond to difficulty develops a trustworthy understanding of that student’s capacity. The dipstick checks described earlier formalise something good teachers already do: they sample, notice, compare, and remember.
The teacher can then interpret polished work intelligently. This is more humane than constant suspicion and more accurate than trusting the artefact alone. The educator must know when to explain and when to ask, when to support and when to withdraw support, when the classroom needs unassisted practice and when the studio should open fully, when augmentation widens access and when it begins to conceal a gap, and when AI is simply one useful resource among many rather than the subject of the lesson itself.
None of this is simple. Students should not be divided too quickly into fixed categories of honest and dishonest, capable and incapable, good and bad. People are shaped by complex experiences and circumstances. Actions still have consequences. Understanding is not justification.
Accountability in education should nevertheless remain developmental. It should ask not only what happened, but what conditions, capacities, incentives, pressures, or misunderstandings allowed it to happen. A student who misuses AI may be responding rationally to a system that rewards finished products more visibly than learning. That does not make the action acceptable. It does mean the school should examine the structure it created. The educator’s task is not only to identify failure. It is to understand what must change so another response becomes possible.
10. Trust, Agency, and Purpose
One of the phrases I often use, or at least think, is trust the process. That can sound empty. Sometimes it is. A process does not deserve trust simply because it exists. Trusting the process does not mean assuming everything will work out or continuing along the same path regardless of evidence. A process deserves trust only while it remains accountable to reality. The process I trust is iterative: Observe. Act. Receive feedback. Reflect. Adjust. Continue. That is how learning develops. It is how organisations develop. It is how schools develop. It is how the handbook developed. It is how people develop. We often demand certainty before action, particularly from leaders and institutions. We want exact answers about what something will become, when it will happen, and what every future stage will look like. But anything genuinely developmental remains partly undefinable before we arrive. Version 2.0 cannot be described completely from Version 1.0 because the iterations between them will reveal information that does not yet exist. The path contributes to the destination. Creation is not simply an idea imposed upon reality. It is a conversation with reality. The rock changes the wall. The weather changes the climb. The learner changes the lesson. The school changes the handbook. The process teaches us what the original plan could not know. This matters in a culture pulled repeatedly towards fads, counterfads, and binary certainties. The fashionable position may contain truth. So may the reaction against it. Neither deserves authority simply because it has momentum. Education should help students act responsibly before certainty exists. That requires knowledge, judgment, courage, humility, and self-belief. Self-belief does not mean assuming that one is already right. It means trusting one’s capacity to learn, adapt, be corrected, and become more capable. Difficulty is not proof of inadequacy. Correction is not humiliation. An unfinished state is not a failed identity. Students need to experience themselves as developing. AI can support this when it helps learners cross barriers that once prevented participation. It undermines it when it teaches them that success means hiding the present self behind a more impressive artificial product.
Human beings have an odd habit of losing sight of their own development at the moment it should give them confidence. We struggle through a problem, find a route, and absorb the answer so completely that the capacity developed in reaching it begins to disappear from view. What once required courage, uncertainty, error, persistence, and help is soon recategorised as something we simply know how to do. Our internal picture of ourselves tends to lag behind what we have actually become. Then we arrive at the next gap. Because no route is visible yet, we see only the absence of an answer and feel helpless once again.
We remember the answer and forget the becoming.
What should carry forward is not necessarily the previous solution. The next problem may look nothing like the last. What transfers is the deeper capacity formed in the struggle: the ability to observe, remain present within uncertainty, ask for help, revise, collaborate, retreat when necessary, and look for another route. Not every gap can be crossed in the way first imagined, and some cannot be crossed alone. Even so, unfamiliarity is not the same as incapacity. The very fact that we have reached a new edge is evidence of the ground already crossed.
This is true of learners, and perhaps of humanity itself. Fire did not end vulnerability; it enlarged possibility and consequence. Artificial intelligence will not settle the human problem either. It places us at another edge, and because we cannot yet see the complete route, we are tempted to imagine ourselves newly powerless. The more responsible response is neither certainty nor helplessness, but recognition: we have developed capacities before, we remain responsible for how they are used, and the next stage will require us to become capable in ways that cannot yet be fully described.
Agency and purpose
The best use of artificial intelligence is not to replace the individual. It is to increase the individual’s capacity to act with agency and purpose. Agency is not merely the freedom to choose. Without knowledge, competence, judgment, and some understanding of consequence, choice can become little more than reaction. Purpose is not merely a goal. It gives direction to capability. It connects what a person can do with why it should be done.
In the redevelopment of the handbook, AI served agency by allowing us to accomplish more within the resources available. It did not stand above the wider ecology of knowledge and contribution; it helped us work across it. It served purpose only because the result was intended to strengthen communication, honour the people represented, guide the school community, connect users with the right sources of knowledge, and clarify the commitments through which the institution understands itself. Nothing meaningful existed without that prior human purpose. The same must be true in education. A student’s agency is not increased simply because a tool produces on demand. Agency grows when the learner becomes more capable of deciding what should be produced, why it matters, whether it is accurate, what should be rejected, how it should be improved, and what responsibility accompanies its use. The aim is not to help students produce increasingly impressive work while they remain unchanged behind it. The aim is to help them become more capable of meaningful participation. AI should help the learner reach further. It should not make the learner disappear.
Conclusion: Learning to Live
The deepest purpose of education is not learning to use technology. It is not even learning to learn. It is learning to live.
We arrive through a sequence of events we did not choose. We inherit cultures, relationships, strengths, limitations, opportunities, difficulties, and perspectives. We do not control our beginnings. What follows depends increasingly upon how we respond. Education is part of that response. It should help people enter into more truthful relationships with reality, with others, and with themselves. It should develop the capacity to act, the judgment to choose, and the purpose to contribute. Artificial intelligence belongs within that larger responsibility. As with controllable fire, the leap in capacity does not arrive with a purpose attached; it makes our choices more consequential. It should help us move more quickly towards worthwhile goals, test ideas more freely, express possibilities more clearly, and extend the reach of genuine human capability. But advancement should improve who we are. It should not merely increase what we depend upon. The measure of AI in education will not be the polish of the student’s work.
It will be whether the student becomes more capable of understanding, judging, creating, contributing, and continuing when the tool is absent, wrong, or insufficient. The question is not whether AI can produce the answer. It can. The question is what kind of person is developing beside it.
We should allow students to accomplish more than they could accomplish alone. We should encourage ambition, experimentation, creativity, collaboration, and responsible use of every tool available. But we must not mistake amplification for foundation. We must not confuse the quality of the product with the development of the person. We must not allow efficiency to erase the distinctive source point from which individual meaning, responsibility, and contribution arise. And we should remember that everything AI can amplify depends upon something human existing first.
The classroom and the studio give this responsibility a practical rhythm. They also move AI out of the spotlight where current debate too often places it. In the classroom, the learner’s foundation remains visible. In the studio, AI may be present, absent, central, or incidental depending upon the work, alongside every other legitimate source of augmentation. The classroom keeps the foundation visible. The studio permits ambition, amplification, and real creation. Dipstick checks reconnect the two without burying teachers beneath an impossible assessment burden. Students move outward with powerful tools and return with something more than a polished product: a stronger capacity they can still call their own.
The responsibility of education is to cultivate the individual and then help that person extend their capabilities without surrendering ownership of them. Foundation before amplification. Process before performance. Understanding before dependency. The person before the artefact. And throughout it all, the continuing responsibility to learn how to live.
July 2026, Osaka