What if the most important question about tomorrow’s schools is not how intelligent their technology becomes, but whether their students and teachers are still allowed to exercise judgment?
Imagine a school where a student can ask an AI tutor for a patient explanation at any hour. Every difficult word can be translated; every algebra problem can be broken into steps; every draft can receive instant suggestions. The building is connected, the platform is impressive, and the dashboards are full.
Now imagine that the same student must decide whether a source deserves trust, whether a shortcut is honest, or whether a classmate has been treated unfairly. A system may provide an articulate answer. But has the school taught the student to judge?
That distinction is the center of The 22nd-Century School, a Learning Coach Editions book that attempts something more ambitious than forecasting educational software. Its central contribution is a design argument: rather than rebuilding schools around each new technological disruption, build an institution around enduring human purposes—intellectual independence, responsible judgment, meaningful relationships and character—then make technology serve those purposes. The title points to a century-long horizon, not a literal prediction of classrooms in 2100.
This is an appealing thesis. It is also more demanding than the slogan “keep humans in the loop.” To evaluate it seriously, we must ask what judgment means, which decisions belong to whom, what evidence would show that learners are developing it, and whether schools can provide such education fairly.
A school can be technologically advanced and educationally confused
In the book’s introduction, a twelve-year-old sits between a paper worksheet, an unreliable school portal and an AI assistant that can solve the task. The scene matters because it exposes a distinction that technology procurement often obscures: availability of assistance is not the same as growth in capability.
A school can improve the supply of explanations without improving students’ capacity to decide which explanation is relevant. It can accelerate feedback without creating time to act on that feedback. It can measure activity without detecting whether a learner understands. These are not arguments against digital tools; they are arguments against mistaking a tool’s output for an educational outcome.
The manuscript calls the accumulation of incompatible reforms a Frankenstein School: a curriculum from one era, an examination regime from another, platforms from a third, and overlapping responsibilities that nobody has designed into a coherent whole. The image is deliberately provocative. But its strongest point is practical. If a school cannot state what students should ultimately become capable of doing, the next technology purchase cannot supply the missing purpose.
An American school district considering a new AI platform, for example, should first establish the instructional need: is it helping multilingual learners access explanations, freeing teachers for better feedback, identifying prerequisite gaps, or merely creating polished assignments faster? These are different objectives. They require different evidence.
The question is not whether AI belongs in schools. It is whether schools know what educational work they want AI to help accomplish.
Judgment is not a decorative “soft skill”
The word judgment can sound imprecise, an inspiring value that disappears when the timetable, curriculum and accountability metrics arrive. The remedy is to make it observable without reducing it to a simplistic score.
A student exercises disciplinary judgment when she can identify which evidence supports a historical claim, choose a mathematical method for a novel problem, or recognize that a scientific result does not justify a sweeping conclusion. She exercises practical judgment when she knows when to ask for help, which digital suggestion to reject, or when a faster route compromises the integrity of the work. She exercises civic and ethical judgment when she considers competing interests, articulates reasons and accepts responsibility for a decision.
These capabilities depend on knowledge, not on vaguely encouraging students to “think critically.” A child cannot assess the plausibility of a medical claim without some understanding of biology, evidence and uncertainty. An AI answer does not become suspect merely because it was produced by AI; it becomes suspect because the learner has learned to examine the reasoning and sources behind it.
This point aligns with the OECD Learning Compass 2030, which treats student agency alongside knowledge, skills, attitudes and values. It describes learners as people who must navigate unfamiliar contexts rather than simply follow a preset route. But the OECD’s framework is aspirational; it is not proof that any particular coaching-based school design will deliver these outcomes.
Judgment must therefore be taught through actual subjects, tested through authentic decisions and developed with feedback. Otherwise it risks becoming another abstract competency printed on classroom walls.
The division of labor that the book proposes
Chapter 20 of The 22nd-Century School, “The Coach-AI Partnership,” gives the theory its operational shape. It assigns AI assistance to explanation, practice, first-pass feedback, pattern tracking and the identification of potential issues for educators to investigate. The proposed human Learning Coach, meanwhile, knows the learner over time, exercises professional judgment, helps develop the person, coordinates support and provides continuity.
That is the book’s most concrete institutional proposition. It is not merely that teachers should use AI. It is that schools should redesign roles so that automation does not consume the attention required for distinctly human work.
The proposal has merit when viewed as a division of responsibility, not a claim that algorithms can perfectly diagnose learning or that every human teacher automatically exercises wise judgment. AI-generated explanations can be wrong, feedback can be shallow, and analytics can mistake differences in access or language for differences in ability. Teachers can also misjudge learners. Neither side deserves blind trust.
The real design challenge is to establish protocols: What may an approved system suggest? What information may it collect? When must a teacher check its recommendation? How can a student challenge an error? Who is accountable when an automated inference harms someone?
UNESCO’s 2024 AI Competency Framework for Teachers emphasizes human agency, ethics and professional competence alongside technical knowledge. Its earlier guidance on generative AI in education similarly warns that deployment requires pedagogical validation, attention to privacy and protection of learners’ autonomy. Those principles support a human-accountability approach, though they do not endorse this book’s specific organizational model.
A useful rule follows: AI may inform an educational decision; it should not quietly become the unaccountable author of that decision.
The hardest assessment question: what did the student actually learn?
A beautifully written assignment has become a less reliable indication of who performed the underlying thinking. But this does not mean every traditional examination is obsolete, or that schools should abandon individual performance.
Chapter 24, “Assessment Redesigned,” distinguishes four purposes that schools frequently ask a single test to perform: guiding learning, certifying competence, motivating effort and monitoring the system. The book argues that AI makes these purposes more urgent to separate.
That is a productive way to escape the exhausted debate between banning AI and permitting it everywhere.
Consider a high-school history assignment. A student may be allowed to use an approved AI tool to outline competing interpretations of an event. She then checks its claims against primary and scholarly sources, chooses an argument and drafts her own explanation. A teacher can review the finished essay, but can also ask the student to defend an evidence choice, revise a claim after criticism, and explain which AI suggestion she rejected.
These additional steps reveal more about reasoning than the final artifact alone. They do not necessarily require constant surveillance, expensive software or a high-stakes oral defense for every assignment.
A better assessment ecology might combine low-stakes independent demonstrations of knowledge, documented revision, discussion, supervised application in new contexts, and carefully bounded AI-supported tasks. Each method has limitations. Oral defense can privilege fluent speakers; portfolios can reflect unequal home support; timed tests can be distorted by anxiety or accessibility barriers. Fair assessment requires multiple kinds of evidence, accommodations where appropriate and explicit criteria.
The crucial shift is not “from exams to projects.” It is from treating one polished product as sufficient evidence toward examining what a learner understands, can explain and can transfer.
No autonomy without instruction
There is a seductive but dangerous interpretation of human-centered schooling: because judgment is important, teachers should talk less and students should figure everything out for themselves. This confuses agency with abandonment.
A novice needs clear explanations, worked examples, practice, feedback and a gradual increase in responsibility. What an expert can decide intuitively may be invisible to a beginner. It would be inequitable to demand independent reasoning before providing the foundations needed to reason.
The Education Endowment Foundation’s 2025 guidance on metacognition and self-regulated learning recommends explicitly teaching pupils to plan, monitor and evaluate their learning, using modelling and scaffolding embedded in subject teaching. The evidence supports teaching learners how to manage their learning; it does not support simply telling them to become independent.
Imagine a science teacher who models how to separate observation from inference. Pupils then compare two explanations, check evidence and finally evaluate a fresh case with fewer prompts. The teacher has not chosen between direct instruction and judgment. Direct instruction has made increasingly responsible judgment possible.
This exposes a limitation of any broad “teacher-to-coach” narrative. Expertise in subject matter remains indispensable. A Learning Coach role could complement excellent subject teaching, coordinate support and make strategic learning more visible. It should not become a pretext for replacing subject experts, weakening disciplinary knowledge or asking one adult to substitute for an entire network of specialist services.
An uncomfortable question: whose judgment counts?
If judgment becomes the school’s most valued outcome, someone must decide what good judgment looks like. That power is never neutral.
One school might prize assertive debate; another might recognize careful listening, collective decision-making or thoughtful restraint. A culturally unfamiliar style of speech can be mistaken for weak reasoning. A student with a disability may demonstrate sophisticated decision-making through assistive technology rather than through a conventional oral performance. A multilingual learner may understand an idea deeply while lacking the vocabulary to defend it in English at speed.
Chapter 31 confronts an especially serious threat to the proposed 22nd-century model: affluent families are often best positioned to take advantage of new educational resources. Personalized coaching, sophisticated AI tools and project-based credentials could become another set of advantages purchased outside the public system.
The antidote cannot be the assurance that technology is “democratizing” because it is widely available. Access requires appropriate devices, connectivity, language support, privacy protections, trained educators and the time to use them productively. Likewise, if independent learners receive praise while students who need scaffolding are stigmatized, the school has turned a developmental aim into a hierarchy of worth.
A credible school built around judgment should make five commitments:
Teach the foundations to everyone. Literacy, numeracy, disciplinary knowledge and explicit reasoning instruction are preconditions for meaningful choice.
Make support an entitlement, not a reward. Coaching and expert help should reach learners who need them, including those least able to advocate for themselves.
Keep decisions contestable. Students and families should be able to understand and question consequential recommendations.
Treat accessibility as part of design. More than one form of demonstrating understanding may be necessary.
Audit the distribution of benefits. Ask which pupils actually gain time, attention, feedback and opportunities, rather than counting only platform logins.
These are design principles, not a validated five-point assessment scale. Their value lies in the questions they force schools to ask.
The philosophical claim deserves scrutiny, too
At its most forceful, The 22nd-Century School argues that AI cannot care, bear moral responsibility or provide the human witness through which a young person develops a sense of identity. Chapter 43 presents attention, care, witness, authority, meaning-making and presence as forms of human contribution worth protecting.
As a moral and institutional position, the argument is compelling. A system cannot be the accountable teacher of record merely because it generates reassuring language. Schools need adults who can listen, explain decisions, admit mistakes, stay involved and be answerable to students and families.
The stronger metaphysical claim—that no future artificial system could ever have experiences or morally relevant capacities—is less settled than the manuscript sometimes suggests. We do not need to resolve the philosophy of machine consciousness to establish responsible school policy. The practical case is sufficient: current AI systems should not be treated as substitute guardians, trusted adults or independent decision-makers for children.
Likewise, the observation that an educator can care is not evidence that every educator is given the conditions to do so. Overload, poor staffing, punitive accountability and inadequate preparation can make humane teaching difficult. Rebranding teachers as coaches without changing those conditions would merely place a new title on an old burden.
The book is strongest when it turns the problem from a technological contest into an institutional responsibility. It is weakest wherever a large philosophical contrast risks concealing the difficult questions of staffing, financing, subject expertise and implementation evidence.
A practical test for schools today
No school needs to wait until the 22nd century to ask whether its routines support human judgment. A modest experiment within an existing department can reveal where the present design helps or hinders learners.
Start with one curriculum objective—for instance, evaluating evidence in a science explanation. Define what students should know before making the judgment, which parts of the task may use approved AI assistance, and which decisions must remain visible in the student’s own work.
Teach the content. Model the reasoning. Invite pupils to compare a tool-generated explanation against a credible source. Ask them to identify a limitation or revise a claim. Then give a similar but unfamiliar task with less assistance.
Review three things separately: the quality of the disciplinary understanding, the student’s explanation of the choice, and the conditions of access. Check which students benefited, which needed additional support, and whether staff workload actually changed. Avoid claiming a causal effect from a single class trial.
For school leaders, this is a more useful first step than a technology showcase. It asks the institution to make the learning process legible, while leaving room to revise the intervention when evidence disappoints.
Beyond the next gadget
Perhaps the most interesting feature of The 22nd-Century School is the length of its horizon. New tools will change. The difficult questions about what people should know, whom they should trust, when they should act and how they should live together will not disappear with the next software release.
The school of the future cannot be built by preserving every old classroom ritual. Nor can it be built by outsourcing more and more decisions to systems whose fluency makes them appear authoritative. It must keep the intellectual difficulty of education alive: acquiring knowledge, confronting error, choosing responsibly and learning to live among other people.
That is why human judgment is a plausible design center—but not an all-purpose substitute for curriculum, evidence or institutions that protect fairness. The ideal is worth defending precisely because it forces hard questions about the present.
A school may eventually be full of astonishing machines. The decisive question will still be whether the young people who leave it can say, with knowledge and reasons of their own: I understand the choice before me, and I can take responsibility for it.
About the book and editorial disclosure
The 22nd-Century School is a Learning Coach Editions book, registered in the publisher’s English-language book records as a published title. This article draws on the manuscript’s introduction and Chapters 13, 20, 24, 31, 34 and 43; it is a critical editorial debate, not a claim that the book’s school model has been experimentally validated. Books & Ideas is published by Learning Coach, and its relationship to the book is disclosed rather than presented as independent third-party reviewing.
Explore Learning Coach’s English-language book catalogue for the title and related works. The individual book-page URL could not be independently confirmed during this editorial check.
Sources and further reading
OECD — The Learning Compass 2030, framework on agency and competencies.
OECD — Future of Education and Skills 2030/2040, including the Teaching Compass.
UNESCO — AI Competency Framework for Teachers (2024), human-centered professional competence.
UNESCO — Guidance for Generative AI in Education and Research (2023), pedagogical, ethical and privacy safeguards.
Education Endowment Foundation — Metacognition and Self-Regulated Learning, second edition (2025), explicit instruction and scaffolding.
Learning Coach — From Curriculum Compliance to Learner Agency, a related applied leadership discussion with a distinct national-policy focus.
