Final grades were submitted last week, classrooms were packed up, and a breezy relief drifted through school corridors across Canada. Students, drained of all energy only a month ago, spilled out the doors with an excitement that can only come from knowing they won’t walk back through them again until September. For the next two months, many will revel in the unstructured, unsupervised, and gloriously unproductive time of summer.

We rarely stop to ask why they’re afforded this intermission, this pause, this extended recess. When we do, we often point to half-remembered answers — that summer break is a holdover from the agrarian calendar, when children were needed in the fields — but that turns out to be a myth. The real origins are historically contingent, though somewhat mundane: urban heat, the spread of disease, and the vacation habits of the nineteenth-century middle class. In other words, summer break wasn’t about kids.

But today, we defend it as though it is. If you ask any parent or teacher why kids should keep their summers, you won’t hear anything about harvests or illnesses. You’ll hear that children need time to play, to rest, to be with their friends and family, to be unproductive, and to simply be kids. We arrived at that conviction through our personal, almost sacred, experiences of summer growing up. The sentiment is as close to universal as you’re likely to find.

Even though we hold that conviction for two months of the year, we discreetly abandon it for the other ten. And the summer itself is no longer safe; we fill it with ultracompetitive travel sports, academic enrichment, and the kind of resume-building you’d expect from an ambitious young professional. When September comes and our students walk back through those doors, we make sure every interval of the school day is filled, every student maximally engaged, every moment turned toward something productive.

We all know that rest and play are good for children. But we stopped making room for them. If a fallow season is what summer offers, and it’s what we say we value about the break, then we have to ask ourselves why we’re working so hard to stamp it out.

Share

The Space Around the Task

I’ve written a lot about protecting and preserving the friction within educational tasks. Although students need to encounter and overcome developmentally appropriate challenges in order to grow, we’ve spent much of the past fifteen years smoothing the surface of learning. The best learning occurs when a task engenders a student’s sense of accomplishment, not because the thing is complete, but because its completion requires careful attention and effort.

Walled Garden schools don’t just defend the friction inside a task, however. They also defend the empty space around tasks — rest, play, idleness, boredom — as developmentally load-bearing. Marketplace Mirror schools — those that take their cues from economic imperatives in the name of “preparing kids for the future” — abhor empty time. Human attention, the most valuable raw material of the digital economy, is a function of time. So in an era when tremendous monetary worth is ascribed to attention, the commodification of time intensifies and our agency over its use diminishes.

If my student’s attention is captured every few minutes by Snapchat, TikTok, and WhatsApp notifications, surely my responsibility is to fight to win their attention back, right? When schools fail to control their students’ attentional environments, this is the logic that colonizes and ultimately extinguishes our children’s access to rest, idleness, and boredom. Under these conditions, time is a zero-sum commodity. Digital applications get most of it, teachers get some of it, and students get none.

Consider an analogy: In the absence of (environmentally damaging) industrial fertilizers, farmers periodically leave their fields fallow to restore soil fertility. Continuous cultivation for maximum yield depletes the soil.

Ever wonder why we have a crisis of engagement in our classrooms today? Our kids’ minds never have fallow seasons anymore. Their attention is the target of an uninterrupted competition by social media influencers, the devices we provide them in school, the demands of the teacher at the front of the classroom, their parents’ text messages, and their friends’ group chats.

To an economy that turns attention into revenue, empty time looks like waste — so it gets filled, every interval and every season, until there’s no fallow ground left.

Share

Subscribe now

Scholē: What School Forgot Its Name Means

The word scholē is the Greek root of our word “school.” It meant leisure. We have built an institution to fill every minute of a child’s day, and we named it after empty time.

Today, we struggle to see the connection between school and leisure. School is a place where work is done by students and teachers alike, not a place where people relax. But leisure and learning have a long and reciprocal history. Indeed, Aristotle argued that leisure is a prerequisite for the highest form of human fulfillment. It is in leisure, when we have space and time to reflect and contemplate — not the empty, passive relaxation we associate leisure with today — that we might develop habits that grow into virtues of character over time. To Aristotle, traits like generosity, honesty, and courage are not randomly assigned, but the consequence of conscientious habits only practised through leisure.

The twentieth-century German philosopher, Josef Pieper, furthered Aristotle’s thesis, explaining that leisure isn’t a time of recovery for more work; it’s a receptive, contemplative openness that is the very basis of learning and culture. In the absence of leisure, education becomes a shallow instrument of work, of the economy. The kind of deep engagement with ideas that characterizes our most valued contributions to science, the arts, and philosophy requires space for conscientious reflection, even reverie.

With that in mind, ask yourself: What opportunities do we offer for leisurely contemplation in our schools today?

Without leisure, Pieper warned of a “World of Total Work”, describing a society where human value is reduced strictly to economic utility and productivity, transforming people into mere “workers” rather than whole human beings. The Korean philosopher, Byung-Chul Han, updates this idea for the twenty-first century, calling ours as an “achievement society” in which even our leisure is enclosed by economic incentives to brand and sell ourselves. Sounds like the competitive and performative structures of social media, doesn’t it?

This matters for schools in two ways. The first is an opportunity: schools are uniquely positioned to create leisurely space for children at the very moment its developmental benefits have been suffocated everywhere else. The second is a guard against misreading what that space is for. Offering children leisure, properly understood, is not the same as claiming that “breaks make better students.” This isn’t about test scores. The academic payoffs may well be real, but a school that takes leisure seriously does so knowing the point is the holistic wellbeing of the child. It’s about human flourishing.

To extend the analogy, a field isn’t rested in order to be exploited harder next season; rest is part of what a healthy field simply is.

Share

Designing for Fallow Time

We know that children’s free, unstructured time to play has declined sharply in recent years. We know that this coincides with overscheduled activities that correspond with a rise in depression and anxiety. But we also know that boredom and mind-wandering can be developmentally beneficial, and that rest and recovery are good for learning. None of this is the reason to act — leisure is owed to children whether or not it pays a dividend — but it means we aren’t even getting the trade we thought we were making. So why, then, are we doing this to our children?

Sociologist Annette Lareau calls it “concerted cultivation” — the adult-organized model of enrichment that turns childhood into a competitive process of resume-building optimized for today’s college-application logic. In effect, it trades the developmental necessities of childhood away for career potential in adulthood, failing to notice that the latter is endangered without the former.

The orchestrated erosion of leisure at home, mirrored by its deterioration at school, and amplified by the incursion of algorithmic content on our kids’ private time means they don’t have a fallow field anywhere. Some of our children have hardly experienced boredom, let alone genuine leisure, in years.

Here’s what I suggest schools do about it:

1. Take back the environment. No phones, no smartwatches, no earbuds. Use computers only when the technology genuinely supports or transforms learning, never by default. This is the precondition for everything else because the screen is the instrument that abolishes fallow time, filling every interval with stimulus.

2. Return the intervals to children. Lunch, recess, and spare blocks should be theirs again for face-to-face conversation, for boredom, for imagination, and for the unhurried contemplation we’ve removed from the school day. A lunch table should be a site of laughter and eye-contact rather than one child gesturing to another to look at images on a screen. Recess should be a time when imaginations run wild rather than time spent watching a YouTube video in a classroom.

3. Bring families in. None of this holds if it stops at the school doors. The erosion of leisure is also happening at home, so schools can’t fix it alone. Hold parent meetings. Send letters home. Make the case for what’s at stake, and ask families to be partners in protecting it rather than bystanders to it.

Schools are already understood to be unique environments. We don’t think they should look like shopping malls. We don’t think they should look like movie theatres. We don’t think they should feel like living rooms, glowing with screens. A school is none of these — it’s the one place we set aside for the conditions learning requires. And learning requires attention, just as childhood requires room to breathe. Both need ground left fallow.

The Fallow Season

It’s summer. We love this season because it gives our children the one thing the rest of the year takes away: the space and time — the leisure — they need to flourish. The task is to stop treating these two months as an exception, and to notice the lesson in them.

We’re sentimental about this time because it offers a counterpoint to Han’s “achievement society.” As kids, it was a season of respite, when we didn’t have to perform for anyone or brace for what was coming tomorrow. We weren’t building a resume toward who we were supposed to become. We had the chance to simply be — here, and now.

Unfettered presence. It’s one of the most beautiful gifts of childhood — one we haven’t done a great job of protecting. But summer is our guide. The task is to stop letting it end in September.

If you teach, lead a school, or are raising a child, you're one of the people who decides how a young person's hours get filled, which makes you exactly who this piece is for. If it resonated, pass it to the teachers, administrators, and parents who set the conditions of childhood alongside you. Protecting fallow time isn't something any of us can do alone; the Walled Garden grows through exactly this kind of shared resolve.

Share

Subscribe now

Leave a comment

Interested in reading more?

Fintech companies like Affirm, Klarna, and Afterpay have revolutionized financial credit in recent years. Their “Buy Now, Pay Later” products are expected to become a US$180 billion industry by 2030. To put that into perspective, music streaming, dominated by media giants Spotify and Apple Music, is currently less than half the size.

Why does this matter? Because, as marketing professor Miranda Goode explains, BNPL schemes aren’t a new form of credit; they’re “debt in disguise.”

Goode asks us to imagine buying a $132 bean bag chair. Instead of paying the whole cost at checkout, we’re given the option to pay $12 per month for the next eleven months. The installment price seems more manageable to many consumers. But research suggests that deflating consumers’ initial “sticker shock” encourages BNPL customers to spend more than they otherwise would.

The industry likes to claim that BNPL is “credit on training wheels,” but because the wheels are built right into the checkout flow, it habituates consumers to a debt-intensifying behaviour. Those who manage their payments and cash flow poorly are susceptible to “debt stacking,” where customers find themselves juggling multiple BNPL products on top of traditional credit options.

Frictionless access combined with deflated sticker shock makes these companies the perfect Trojan horse for unintended and unnecessary debt.

But the costs they impose aren’t limited to their own customers. BNPL firms charge higher interchange fees than traditional credit card companies do. Ultimately, retailers may pass those additional costs on to all consumers through higher prices.

The mechanism is what should worry us, because it isn’t unique to retail. Anything that severs the moment we choose something from the moment we pay for it will pull us toward choices we’d otherwise reject. And that same dynamic has found its way into our classrooms.

Increasingly, schools make pedagogical decisions the way shoppers click “confirm” for installment payments. When given a choice, we often select the option that makes the day go smoother — the YouTube video that calms the room at lunch, the AI tool that grades the assignment for us, or the grammar extension that polishes students’ prose. The cost, however, is left for later. Each of those choices feels free because nothing is debited at the “point of sale,” when the decision is made. The lesson is engaging. The work is completed. The room stays quiet.

But what seems free is just a deferred cost. We haven’t escaped it.

When a shopper uses Klarna to put off a payment, the bill eventually comes back to him. But at least he made that decision himself. The student sitting in front of a YouTube video at lunch is pacified in the moment, but doesn’t know that her future self — the adult she’ll become — will not have benefitted from the creative attention one learns to harness from boredom. Unlike the Klarna customer, though, she didn’t make the choice herself; an adult made it for her.

I’ve called this “developmental cost” before, using it to help us sort which tools belong in which hands. Here, I want to look at the cost itself: what we’re trading away, why we’re struggling to see it, and why we keep choosing to reduce friction.

Share

What Gets Displaced: Three Kinds of Developmental Cost

A cost of what, though? When a tool carries out a task, or part of a task, the student doesn’t. Every time we ease a student’s workload or divert their attention from active engagement to passive consumption, we make a trade on their behalf, without their permission. While these choices buy some convenience — efficiency and a little quiet in our younger classrooms — those gains come with at least three kinds of costs:

The first is cognitive. When we use tools that outsource exercises that require thought and consideration — from something as simple as deciding where to place a comma to something as complex as imagining the outcome of an experiment — we cancel the cognitive rehearsals that develop deep and durable habits of mind.

The second is social. When schools stop building environments that require students to face each other and their teachers, when learning is “personalized” through an AI tool, or when a young classroom is placated by a video, the interactions that demand attention, patience, and tolerance give way to those a screen mediates. We displace the practices of collaboration, of belonging, of learning to read a room, and of working through disagreement.

The third is dispositional. When these tools become normal, students are habituated to them, just as shoppers grow accustomed to paying later. If using AI to do research becomes routine, reading primary and secondary sources starts to look inefficient and dull by comparison. A student who never pushes through the boring stretch of a difficult task lacks an important skill and they lose their taste for the effort.

None of these costs show up in a grade. Students can carry all three and still bring home an unblemished report card at the end of the semester. The surface of their academic performance is the last place where these costs appear, which is exactly the problem.

Share

Subscribe now

Invisible & Delayed: Why the Cost Stays Hidden

Why is that a problem, and not just an inconvenience? Because these costs share two characteristics. They’re invisible: the usual measures of academic performance don’t capture them. And they’re delayed: their developmental consequences persist and grow long after the decision that introduced them, by which point almost no one thinks to connect the harm to its cause. A cost with only one of these would be manageable; we catch hidden costs eventually, and we plan for delayed ones when we can see them coming. It’s the combination that does the damage.

Invisible: The Finished Work is Not the Evidence

A 2025 study by Fan et al. published a finding in the British Journal of Educational Technology found that university students who used ChatGPT produced better work on the surface, but their knowledge transfer didn’t improve. The visible thing — the submitted essay — got better, but the invisible thing — whether they could apply their knowledge across contexts — was unaffected. If AI-enhanced texts can mask the thinking of a motivated undergraduate, it conceals more of a child’s, whose capacities are still forming.

Were K-12 students to produce work for the collective benefit of a popular audience, encouraging the use of a tool that enhances the quality of the final product makes sense (ethical quandaries aside). But in my experience, very few people are interested in reading an eighth grader’s review of The Outsiders.

I’m interested in that book review because I want to assess the quality of their writing before it’s been subjected to the compositional equivalent of a Botox injection. Even more importantly, I want to evaluate whether my student can, with any level of sophistication, distinguish a gang fight that advances the plot from the underlying class division — the one that animates the conflict, provides characters with identity, and reveals difficult questions about loyalty and innocence.

If ChatGPT were involved, I’d be reading a review with a facelift. I’d know nothing of my student’s independent ability to discern plot and theme, track character development, or notice the subtleties of authorial craft. AI makes their cognition, and its development, invisible to me. That’s a problem because their review isn’t going to be published in The New York Times. What I’m assessing is the thinking, and that’s exactly what the polished product hides.

Share

Delayed: The Bill Arrives Later

In 1983, the cognitive psychologist Lisanne Bainbridge published a pioneering research paper titled “Ironies of Automation.” She explained that when a task is automated, the operator stops practising the relevant skill, so when the system fails and the human must take over, they’re rusty and struggle to act at exactly the moment they’re needed the most. When an automated system is running smoothly, the human’s rust stays hidden. In other words, everything looks fine until it doesn’t.

This is what’s happening in our classrooms en masse. When a student uses an AI tool to seek feedback for their written work, they interact with a technology designed for attachment. That accommodation is the problem. A tool tuned to flatter and adapt demands no social flexibility in return: no negotiating a disagreement, no reading a mood, no tolerating friction. AI sycophancy has been well-documented over the past couple of years. A recent study found that reliance on AI companions could lead to “the potential transformation of relational norms in ways that may render human-human connection less accessible or less fulfilling.”

Tools that are engineered to recognize and adapt to the idiosyncrasies of an individual demand very little, if any, social flexibility from that person. When we automate social interaction, even for academic purposes, the cost is the corrosion of their social capacities.

In this sense, the classroom isn’t separate from the rest of the students’ day. Whether they seek frictionless feedback from a chatbot at their desk or distract themselves with the video game that lives in their pocket, the accommodating screen offers them an always-available recess from discomfort.

When I first launched The Walled Garden Education, I recounted a time when I took my class outside, without their phones, to read a chapter of a novel on a sunny day. We finished early, and I gave my students some free time:

“As expected, many eagerly took to the playground. Yet, a quarter of my class lingered, clearly hesitant. I tried to entice them with suggestions: ‘Why don’t you borrow a ball from the gym? Have you ever played Capture the Flag?’ But my appeals fell on deaf ears. For half an hour, these 13-year-olds sat — and in some cases, merely stood — barely interacting with one another.”

This was the delayed cost made visible. Whether traded away to a screen at their desk or at home, the social reps they were missing surfaced at thirteen; they no longer knew how to fill thirty empty minutes with each other.

Their age matters. In Bainbridge’s analysis of automation, operators lose a skill they once had. These students are failing to build it in the first place, which is worse because there’s no prior competence to fall back on.

We already have clear evidence of this dynamic. For roughly fifteen years, our kids have proceeded through their primary and secondary education with near-unfettered access to phones and school-supplied laptops that house software engineered to capture and monetize their attention. That generation is now arriving at university. Professors from community colleges to the Ivy League report that their students struggle to complete readings already modified to be shorter. The number of young people who read for pleasure has collapsed. What post-secondary instructors report is students who not only have lost a skill, but lost a lost appetite. They feel unequipped to grapple with difficult texts, and uninterested in trying.

All three costs are visible in today’s lecture halls. The cognitive capacity to read critically is underdeveloped. The shared and social practice of deliberative analysis expected in an undergraduate seminar is lacking. And the learning disposition — the simple desire to push through difficulty in pursuit of personal growth — has gone missing.

But notice what we do with this evidence. We argue about whether it’s the phones, the pandemic, or the curricular choices we made to read excerpts instead of whole texts (it is, of course, all of these). We won’t settle those debates because we can’t. The cost arrived fifteen years after it was incurred, in settings where no grade recorded it, and it’s so far downstream that identifying a single decision that can be labelled “The Cause” is impossible.

This is what “invisible and delayed” looks like in education. When the bill finally comes, all we see on the receipt is a vague, system-wide deficit that we can’t assign blame for. The damage is undeniable, while the cause is deniable. And the young people at the centre bear the greatest burden.

Share

Subscribe now

The Loss and Its Receipt

We arrive at these costs, which are neither agreed upon nor understood by those who have to bear them, through what the sociologist Diane Vaughan called the “normalization of deviance.” After the loss of the Challenger Space Shuttle in 1986, she explained that clearly substandard practices become accepted as normal inside institutions that don’t immediately produce catastrophes — “non-event feedback,” in her words.

The absence of a disaster makes small deviations commonplace until the lowered standard becomes invisible to the people working inside the institution. It’s an elegant idea, and it tells us why the logic of Buy Now, Pay Later subtly took hold in schools: each frictionless choice failed to affect report cards, and the unblemished report cards represented the non-event feedback that authorized the same practices.

Knowing what we know now, we shouldn’t turn our backs on technology, but we should become far more serious about how it’s mobilized in our classrooms and managed in our homes. If you’re a school administrator and your students still spend their lunch breaks on Snapchat, a bell-to-bell phone policy is long overdue. If you’re an elementary teacher relying on Netflix to mollify an oversized class, please stop; collect evidence of harm, galvanize your colleagues, raise the concern with parents, and demand the conditions that make better practice possible. If you’re a high school teacher, the take-home assignment that assesses only a final product is due for a rethink: consider collecting Artifacts of Attention alongside the finished work.

Consider this: if you’re making a choice on behalf of children that makes things easier in the moment, and that choice wasn’t available to you fifteen years ago, you may be buying now and asking the kids to pay later. The adults those children will become aren’t in the room when the decision is made, but they deserve consideration. So before playing another Kahoot, ask: Will the adults these kids grow into thank us for this?

The crisis is already underway; you only have to read the statistics and the professors’ accounts to see it. Mortgaging our students’ futures for a quieter afternoon is the Klarna of the classroom. It’s time to stop.

If you teach, lead a school, or are raising a child inside all of this, you already know which convenient choice you'd reach for tomorrow. This piece is an argument for pausing before you do, and for passing it to someone who makes those choices too.

Please like, comment, and share if this piece resonated with you.

Share

Subscribe now

Leave a comment

Interested in reading more?

I never expected a 42,000-word papal document to converge so neatly with a K-12 newsletter. And yet here we are.

Two weeks ago, Pope Leo XIV published his first encyclical, Magnifica Humanitas, on safeguarding the human person in the age of AI. You may have seen the headlines, but I’d encourage everyone to read it for themselves. You don’t need to be Catholic or religious to find it valuable; it is a tremendous and refreshing piece of moral reasoning about technology and people.

Leo XIV signed Magnifica Humanitas on May 15, deliberately marking the 135th anniversary of Leo XIII’s Rerum Novarum, which directly addressed the first industrial revolution’s threat to the worker, insisting that the Church could not ignore “the concrete lives of people.” Leo XIV is consciously doing the same for the human person against the disruption of AI.

The pope’s encyclical calls for institutions to protect the vulnerable against a disruptive technology rather than mirror the market that produces it, which is precisely what The Walled Garden Education has argued schools should do. My work builds its case from empirical and developmental evidence; the pope’s rests on theology and a 135-year-old body of social teaching. The logic is the same in each case: an institution entrusted with the vulnerable owes them protection, not optimization.

Both refuse to let efficiency become the measure of a person.

Share

The Tool Is Not the Verdict

Here, Leo refuses to reduce our discernment about technology to the straightforward binary we often reach for. So often, the conversation collapses into an all-too-simple question: Are we using the tool well or badly? But that question rests on the assumption that the tool is itself neutral. The pope disagrees. He insists that any system embeds assumptions about people — in what it decides to measure and what it chooses to ignore.

I’ve argued the same thing, but from the classroom rather than first principles. The question of whether a tool is good or bad, or whether we use it properly, obscures a better framing: Who is the tool for and what does its use cost them? The same tool that can expedite a teacher’s work can rob a student of the executive function she’s still building.

Technology isn’t bad. The pope calls it “a profoundly human reality.” The more important point is that a tool is never just a tool. Its design is shaped by others’ ideas about what people need, what people are for, and those ideas act on us whether we notice them or not.

Share

Subscribe now

Friction, and the Desire to Ask Questions

Elsewhere in the encyclical, when Leo writes about work and the economy, he treats AI as a tool to be governed wisely, to be regulated in pursuit of the common good. Under those circumstances, it’s a tool to be used. But when he turns to education, his argument shifts. He doesn’t just ask us to teach youth how to use it well. He insists that we teach them when not to use the tool.

The spark of curiosity in a young mind is easily extinguished when one can summon a polished paragraph in seconds. Doing so eradicates encounters with “desirable difficulties” — with the discomfort of not yet knowing something. But it’s in that discomfort, in the friction of temporary struggle and confusion, that curiosity is born.

Leo identifies the casualty precisely: “the desire to ask questions, which is a process that bears fruit only over time.” He reaches back to Plato to underscore the same idea — that the deepest and most valuable things are learned only slowly. I’ve reached for the research of Robert and Elizabeth Bjork, as well as my decade of classroom experience to make the same point. A machine that removes the struggle removes the education.

Share

Schools Should Not Mirror the Marketplace

I could have written that sentence. In fact, I have written it many times over the past ten months.

This is the central argument of The Walled Garden Education in a single line. A school that mimics the digital world by reflexively adopting its tools and replicating its disruptions is what I’ve called a Marketplace Mirror: an institution that takes its cues from the economy outside its walls and reflects them back at children. The pope is describing its opposite: a Walled Garden school. He explains that schools exist to offer what the marketplace, by definition, cannot: occasionally inefficient but developmentally necessary experiences, and relationships built on trust. Those aren’t “nice-to-haves” alongside academic work; they are the precondition for academic achievement.

Earlier in the encyclical, when Leo writes about markets and labour, AI is cast as a potentially “valuable tool”. When he turns to education, his language becomes unconditional: “schools are not called to follow the pace of the digital world…” When you read the whole work closely, it’s clear that his position is graded. Its caution is thinnest where the subject is a formed adult navigating an economy, but it tightens steadily and deliberately as it moves toward children, which is the appropriate response to a simple fact: the more vulnerable the person, the higher the standard of protection we owe them. It’s the same logic a hospital follows, and the same logic I’ve argued for in schools.

The marketplace is free to move fast only as long as it serves the common good. A school entrusted with the wellbeing and development of our youth is held to a higher standard still.

Share

Subscribe now

Protect First, Then Teach

Children are “entrusted to our care.” The relationship between a school and a child is custodial; teachers, administrators, and support staff are responsible for our kids. That is the foundational logic that underpins the case I’ve made for the Precautionary Principle in schools. Our students’ attendance is compulsory and they can’t consent to their participation, so when a technology has the potential to impose harm that might not surface for years, the burden of proof for its efficacy and safety belongs with the tool and its vendor, not with the child using it. Leo makes the same point. The burden shouldn’t be placed on children or their families; it belongs to “service providers”.

In the second half of the quote above, the pope adds another responsibility. He insists that our kids are taught to recognize and resist manipulation in order to defend their dignity. This requires genuine AI literacy rather than mere proficiency — the difference between operating a tool and knowing what it is, who built it, and what its use might cost you.

Protection is the correct first step, but it must be complemented by education. Creating the boundary that allows for children to critically investigate the tool is the responsible, fiduciary approach. By doing so, we keep developing minds safe long enough that they can grow strong enough to defend themselves.

Share

Pope Leo reasons from a theological tradition and centuries of Church doctrine. My rationale emerges from classroom experience, philosophy, cognitive science, and developmental research. We may not share the same starting points, but we end up in the same place.

A school is not a mirror of the marketplace. A child is not “a project to be optimized.” Our youth are owed healthy and protected developmental conditions. They deserve the slow, friction-filled space to learn to think for themselves.

I take the convergence in these arguments to mean that they point to something real and important. The case against handing childhood to machinery is neither religious anxiety nor secular panic. It’s what you tend to conclude when you actually work with children, pay attention to the experiences they’re afforded, and notice what might be missing.

We live in a moment when perfunctory political rhetoric routinely subordinates moral clarity to the promise of efficiency, when “innovation” is treated as its own justification. Pope Leo rejects that subordination. Humanity, and all its imperfections, is worth defending. If this encyclical makes anything clear, it’s that remaining human is something we either protect or surrender.

So I’m glad that Pope Leo XIV calls for a Walled Garden Education even if he never uses the name.

The case for protecting children’s development can be made from different directions, which is exactly why it’s worth passing along. If this piece resonated, share it with the educators, administrators, and parents in your circle. The argument only gets stronger when more people recognize they’ve been making it too, from wherever they happen to stand.

Share

Subscribe now

Leave a comment

Interested in reading more?

When a teacher posts a new assignment to Google Classroom, they know that their students have been notified. They know they’re reaching students who happen to be absent that day. They know the assignment description is accessible anywhere, at any time. They know their students can’t lose it. They know they saved ten minutes by not writing it on the board.

For each of those wins, there are losses. Students who receive notifications for new assignments learn that paying attention to in-class instructions is optional. They learn to rely on software to stay organized because they never have to do so independently. They learn that their presence in the classroom isn’t required.

The convenience these tools afford is real, but so are the costs, and they don’t fall on the same person.

That asymmetry points toward the questions schools should be asking. Not simply: Is this tool appropriate? But: Who is using it, and what does its use cost them developmentally? When you ask those questions, teacher-facing tools and student-facing tools emerge as categorically different objects requiring categorically different standards of evaluation. And that difference becomes most consequential in cases like Google Classroom, where a single tool shapes the experience of both.

Share

The Triangle and Its Blind Spot

One of my favourite resources for evaluating classroom technologies is Everyschool’s EdTech Triangle. It asks teachers to determine if a technology is Disruptive, Restrictive, Supportive, or Transformative before employing it in their classrooms. At the framework’s least desirable end, Disruptive tools needlessly expose students to harmful content, encourage habitual use, and require screen time that serves no clear pedagogical purpose. At the most desirable end, Transformative tools produce outcomes that wouldn’t be possible without the use of the tool: coding, robotics, computer animation, etc. The Triangle is a valuable, research-based hierarchy that I’d recommend to any teacher or administrator trying to think carefully about student-facing technologies.

But the Triangle was designed to evaluate tools in students’ hands. It doesn’t have a category for teacher-facing tools, and importantly, it doesn’t evaluate the impact of technologies that start in teachers’ hands but end up shaping student behaviour as well.

When a digital gradebook saves a teacher two hours of arithmetic per week, its utility is obvious. Students’ cognitive processes haven’t been disrupted or displaced, and the teacher’s math skills aren’t a developmental capacity we’re trying to cultivate, so the Triangle’s categories don’t apply. This is the clear-cut example — the gradebook lives exclusively on the teacher’s desk. The more difficult cases are those that migrate off of it.

Learning Management Systems like Google Classroom are the clearest example of this migration. The tool is clearly marketed to teachers on the grounds of efficiency and succeeds in delivering it. But it does so by routing student behaviour through the platform and subtly requires them to outsource their executive function, which they could otherwise build themselves.

Everyschool’s framework doesn’t catch this dynamic. The EdTech Triangle isn’t wrong, but it wasn’t designed to ask who the tool actually faces. In many cases, there isn’t a clear line to be drawn between teacher-facing and student-facing tools at all.

Share

Developmental Cost and Who Bears It

Identifying who a tool faces is only the first question. The second is harder: Whose development does this tool cost, and can they afford it?

Unlike the children we work with, teachers are formed professionals. Their administrative efficiency isn’t a developmental capacity in formation. Time saved on grading logistics, electronic communication, or lesson design can be time returned to the work that requires astute human judgement such as planning, relating, noticing, and adjusting. When a teacher uses a gradebook that automates the calculation of averages, we needn’t be concerned that a cognitive capacity is being atrophied. The teacher’s development isn’t at stake.

Students are in the middle of building the cognitive architecture that underpins everything else. When a tool robs a student of the opportunity to sustain their attention, to exercise their working memory, or to organize their own work, it forecloses the repetition needed to build those key capacities. Digital prostheses don’t just complete tasks, they pilfer our students’ chances to exercise the capacities they need to develop.

Teacher-facing tools should be evaluated for efficiency and professional utility while student-facing tools must be evaluated for developmental cost.

Convenience is an appropriate metric for professional administration, but it’s an error when applied to students. Students’ development must supersede administrative convenience in cases where a tool blurs the line between teacher-facing and student-facing effects.

Share

Subscribe now

Two Tiers: Who the Tool Faces

But first, a clarifying question: Even if this tool is nominally teacher-facing, does its function depend on shaping or replacing student behaviour?

Many tools are sold to administrators as teacher efficiency products, but only become functionally productive by routing student behaviour through the platform. In these cases, the teacher is the customer, but the student’s cognitive process is the tool’s operating environment. We’ll return to the hardest of these cases once the framework is in place.

For now, the clarifying question above helps us determine which set of criteria applies. A tool that passes through students must be evaluated as a student-facing tool, no matter what the vendor calls it. With that established, here are the criteria:

Tier One: Teacher-Facing Tools — Evaluate for Professional Utility

The governing question for Tier One products: Does this tool free the teacher to do more of the work that requires human judgement or does it displace the professional thinking that makes a teacher better over time? Most administrative tools clear this bar.

• Gradebook software (e.g. PowerSchool, JumpRope, iGrade, etc.) — These products are purely administrative when the student has limited access to them. They record and calculate what a teacher has already assessed, and increase productivity by reducing or entirely eliminating the need for tedious arithmetic. A gradebook can make a student’s performance transparent to them without managing that awareness on their behalf.

• Communication platforms (e.g. email, parent-facing apps) — These tools help manage and expedite the logistics of school-home communication. They’re entirely appropriate so long as they don’t become the mechanism by which students receive information that they could be tracking and recording to the developmental benefit of their executive function.

• Lesson planning and resource tools — This is the Tier One category most exposed to AI. With important caveats — ethical considerations, energy consumption, environmental consequence, etc. — AI is best understood in the context of a school as a productivity tool for expert professionals rather than an experimental plaything for developing minds. Lesson planning and resource tools are legitimate when they free time for relational and judgement-based work, but they warrant scrutiny, as outlined above, when they begin to displace the thinking that makes a teacher’s practice develop over time.

• Assessment design tools — This, too, is appropriate in principle. Outsourcing the creation of an entire assessment is likely to stall, or reverse, a teacher’s professional practice over time. But using a tool to format a rubric or organize a test bank that was independently created by a teacher is legitimately productive.

Tier Two: Student-Facing Tools — Evaluate for Developmental Cost

Drawn from the Pedagogy of Cultivated Attention, the following questions govern the evaluation of Tier Two tools:

• Does this tool displace a cognitive capacity the student is still building: sustained attention, working memory, executive function, metacognition?

• Does it remove an opportunity for students to encounter a desirable difficulty?

• Does it make the student’s thinking visible, or does it produce an output that obscures whether thinking occurred at all?

For example:

• Electronic texts replacing print — An e-text removes the comprehension advantages of print, removes the opportunity for physical annotation, and exposes students to the possibility of device-enabled distraction. The displaced capacity is sustained attention to a text.

• Grammar and writing extensions (e.g. Grammarly, Quillbot, etc.) — These tools remove opportunities for editorial thinking, which builds writer’s self-awareness and the capacity to improve independently over time. They produce a clean output that obscures whether that thinking ever happened, which masks deficits rather than resolving them.

• AI writing or research tools — These outsource the very cognitive processes students need to exercise for their personal development. Turning to AI for research may seem reasonable, but doing so circumvents the essential processes of discernment that more inefficient methods of research help students develop.

• Audiobooks — An audiobook displaces the decoding and sustained visual attention that reading print requires. For a student who could read the text themselves, that displacement is a cost. For a student with a documented accessibility need, it removes a barrier rather than a desirable difficulty, and is wholly appropriate.

• Coding, design, and production tools — When the tool is the subject of learning, none of the three questions raises an alarm because the student is building capacity, encountering genuine difficulty, and producing work that makes their thinking visible. This is an example of student-facing technologies that can be transformative, in Everyschool’s language.

  Share

  Subscribe now

Two Hard Cases

In the extremes, the framework is easy to use. A gradebook on a teacher’s screen and a grammar extension in a student’s browser are easy to sort. Things are tricker when the tool lies somewhere in the middle, affording teachers some conveniences that impart developmental costs on students. Two such examples are worth examining closely, because they are normalized and highly consequential.

Learning Management Systems

Most learning management systems are the clearest examples of a tool that fails the clarifying question — offering teacher efficiency by routing student behaviour through the platform. There are fewer lost assignments, streamlined submissions, and the LMS centralizes communication, but students are required to check a feed, receive automated notifications, and submit work through the platform rather than developing their own systems for managing these things. The teacher’s convenience and the student’s developmental cost are not separable under these conditions. They are produced by the same features.

A practical recommendation for procuring an LMS: treat the product as a teacher-facing archive and parent communication tool. For students, assignment tracking should require active cognitive engagement; provide them with physical agenda books, verbal announcements they’re expected to record, and deadlines they have to manage rather than are managed by. This looks different in primary and secondary classrooms, but the principle holds across both. Schools can still provide their teachers some of the conveniences of an LMS while preserving the developmental opportunities students need.

Supplementary Resources

The routine posting of slides, notes, and even lesson recordings online is a practice with developmental costs hidden behind the language of access and support. Of course, the intention is generous and kind: remove barriers, extend access, and support diverse learners. But the effect, in practice, is often to make the cognitive work of the classroom structurally unnecessary.

While some students — those with documented accessibility requirements — benefit from these practices, providing all students with these resources is both unnecessary and self-defeating. When students know that slides will be posted, the need to listen carefully, synthesize spoken instruction, and record it in one’s own words — all of which build working memory and comprehension — becomes optional. The resource intended to support learning becomes the reason why learning doesn’t have to happen in the room, and as a result, attendance itself begins to feel like it’s negotiable.

Having to be present, attentive, and responsible for capturing information is, in and of itself, a desirable difficulty, not an inconvenience. When schools remove this requirement in the name of access, they often fail to notice what they’ve traded away.

Supplementary resources, appropriately marshalled, are targeted and conditional. They are for students who missed class for legitimate reasons, for review prior to assessments, or for students with accessibility needs. The normalization of routinely posting resources for all students as a matter of convenience replicates the same logic in schools that has made attention scarce everywhere else. If we want our students to attend to something, we have to create the conditions that make their attention necessary.

Share

Making the Case to Administrators

Administrators resist reducing EdTech on the grounds that it affords conveniences, and they’re not wrong that convenience matters. Where teachers are stretched, tools that ease that load are important. Although the instinct is correct, it misses a key distinction: that a standard which makes perfect sense for teacher-facing tools can misfire when applied to tools that pass through to students. The clarifying question — whether a tool’s function depends on student cognitive outsourcing — can help schools make better decisions for all stakeholders.

In order to reframe the conversation, we need to understand that teacher convenience and student development aren’t in competition unless we conflate the tools that serve each. So our first task in broaching this subject is to clearly delineate which tools are teacher-facing, which are student-facing, and which face both. We can explain that protecting teachers’ administrative efficiency and protecting students’ developmental conditions are compatible goals, but their tools must be evaluated by different standards.

Here are a few concrete talking points for teachers and parent advocates:

1. “I don’t think we need to choose between supporting teachers and protecting students, but it helps to ask, for each tool, which parts are saving teachers time and which parts are doing work students should be doing themselves.”

2. “The gradebook genuinely saves us time and I wouldn’t want to lose it. What I’m less sure about is the assignment reminders — when the system does that for students, are they still learning to remember for themselves? That’s the piece I’d love to think through with you.”

3. “Maybe the question to put to every tool isn’t just whether it’s convenient, but convenient for whom, and whether that convenience is costing students something we won’t notice until later.”

4. “When we post the slides automatically, I worry we’re actually telling students that being present and taking notes is optional. I’d love to figure out together when posting them helps and when it lets the real work slip.”

   Share

The Right Tool

This piece was inspired by a reader who wrote to me asking if it was possible to keep the conveniences of EdTech for teachers while returning to pen and paper for students. The framework outlined above is my answer because a school that learns to separate teachers from students when evaluating their tools can protect both.

But it’s worth noting the deeper current that runs through this exploration. The tools that genuinely help teachers are almost never sold on that basis alone. They’re almost always marketed as catalysts of student learning, but that claim rarely survives scrutiny. When you strip away the advertising, students are left with tools that impose developmental costs and offer little in return. But those costs aren’t just developmental. In exchange for tools that take more than they give, students also hand over the two things these platforms are really designed to collect: their attention and their data.

We began with an asymmetry and we’ll end with it. When schools understand this, they don’t have to choose between their teachers and their students. The convenience and the cost of these tools do not fall on the same person, and once a school sees that clearly, it can make better choices for everyone.

This piece exists because a reader asked a question. If something here clarified a problem you’ve been wrestling with, or if you’re facing a version of it in your own school, I’d love to hear about it. Reply to this post in your inbox, or send a message through Substack. The questions readers send me often become the pieces that follow.

And if this was useful, please share it with the teachers, administrators, and parents in your circle. The Walled Garden grows through exactly these conversations.

Share

Subscribe now

Leave a comment

Interested in reading more?

The Roundup is a monthly collection of my writing as well as the writing of others who have helped shape my thinking about education, attention, and technology. Below, you’ll find works that explore AI literacy, youth screen time, and the gamification of our classrooms.

Please enjoy:

From The Walled Garden Education This Month

This is the third and final piece in the series on evidence and technology in schools. Where the first two pieces examined the plausibility trap and the structural conflict of interest that prevents AI companies from teaching genuine AI literacy, this one proposes what a real alternative looks like. The framework has three elements — What It Is, Who It Serves, and What It Costs — and is grounded in an important distinction: the difference between proficiency and literacy. A student who can write an effective prompt is not AI literate in any meaningful sense. A student who understands why the prompt produces what it produces, who built the system and why, and what it costs them cognitively and the planet environmentally has developed some critical literacy. The piece also proposes a developmental sequence, grounded in cognitive science, for when each element of that framework can realistically be taught. Any approach that inverts the sequence isn’t preparing students for an AI-saturated world; it’s preparing them for dependency.

Two new technologies have arrived in classrooms, and they share a common mission: to attempt to make the illegible legible by measuring it. The first of these is emotional surveillance: platforms that monitor students’ attention and affect via camera during screen-based work. The second is integrity surveillance: platforms that log keystrokes, track browser activity, and generate originality scores to detect AI use. Drawing on Onora O’Neill’s distinction between trust and control and C. Thi Nguyen’s concept of “Value Capture”, I argue that both technologies produce convincing imitations of the qualities we hope young people develop: honesty, engagement, and self-awareness. The alternative is more humane and trusting; it requires us to redesign the conditions of learning: assessment that reveals process rather than surveillance that monitors compliance.

I joined a panel hosted by The Dais, a public policy think tank at Toronto Metropolitan University, to discuss best practices in school phone restrictions, the case for going further than most schools currently have, and what the research says about outcomes. I was joined by two administrators — Matthew Shapiro of Westmount High School in Montréal, Québec and Tim Mushumanksi of Trafalgar Middle School in Nelson, British Columbia — who have both implemented bell-to-bell restrictions in their schools and can speak to what the resistance looks like and what their schools feel like now. The conversation is grounded in the Canadian policy landscape but the practical questions it covers travel well.

Share

Subscribe now

Two Voices from Substack Worth Your Time

“Schools Don’t Need an AI Plan. They Need a Teacher Support Plan.” by

Requarth explores New York City’s recently released AI guidelines for its one million students, but he thinks they ask the wrong question. The document asks how to manage AI inside schools when the question it should be asking is how to redesign the classroom now that AI outside schools has made many traditional methods of assessment obsolete. His modest proposal: for every dollar spent on AI procurement, spend a dollar helping teachers redesign their teaching for a world that has changed underneath their feet through no fault of their own.

“Too Fast, Too Confident, Too Digital?” by

Kirschner returns with a close reading of a new study on how seventh and eighth-graders read the same texts in print and on screen. The finding is that when students read digitally, they read faster, became more overconfident, and understood less. Drawing on Gavriel Salomon’s classic work on the effort learners invest in different media, Kirschner explains that screens signal ease, and learners respond accordingly. Students glide through digital text, don’t pause, don’t reread, and mistake fluency for understanding. The authors propose better-designed digital environments as the solution. Kirschner has a simpler one: print it.

Share

Three Great Reads from Beyond Substack

“New Surgeon General’s Advisory Raises Alarm About Screen Time Risks for Kids and Teens” by Jacqueline Howard in CNN

Released this week, the advisory declares excessive screen time among children and teenagers a public health concern. By adolescence, children average four or more hours of screen time daily, and nearly half admit they lose track of how long they’ve been on their phones. The advisory recommends no screens for children under 18 months and a maximum of two hours daily for ages 6 to 18. That it was issued at all by a leaderless public health institution suggests the evidence has become impossible to ignore.

“Canada Is Considering an AI Chatbot Ban for Kids” by Laura Osman and Martin Patriquin in The Logic

Three sources with knowledge of the government’s plans confirmed to The Logic that Ottawa will proceed with a social media ban for youth and is now actively debating whether to extend that restriction to AI chatbots. At their recent policy convention, Liberal Party members voted to set the minimum age at 16. Unlike social media, however, some believe that AI tools have educational uses, making a blanket ban thornier. But the political direction is clear, and the Tumbler Ridge shooting — in which a B.C. teen who killed eight people was found to have had troubling conversations with ChatGPT beforehand — has increased the urgency considerably.

“My Son’s Math Homework Is Essentially Just Pokémon” by Will Oremus in The Atlantic

Will Oremus opens with a scene that will be familiar to many parents: his 11-year-old, supposedly doing math homework, was instead battling cartoon monsters on Prodigy, a platform his teacher had assigned. In ten minutes of gameplay, his son spent less than thirty seconds answering math questions. Oremus documents how gamified EdTech platforms like Prodigy, Blooket, and Gimkit have become fixtures in many classrooms, promising engagement by delivering something similar to Candy Crush. The deeper problem his piece identifies is that screen time in schools has become a default rather than an intentional choice, and that phone bans and content filters, however necessary, haven’t proven effective enough. Software has colonized K-12 schools and unwinding that will require more than a Yondr pouch.

This month’s reading identifies a gap between awareness and action. The surgeon general’s advisory, Canada’s proposed chatbot restrictions, and the flaws of gamified homework all illuminate the vast distance between our concerns and our institutional practices. It’s time that we address that disparity.

If you agree, I’d love to hear about it in the comments. These conversations matter.

Until June,

Andrew

Share

Subscribe now

Leave a comment

Trust and Its Enemies

Ralph Waldo Emerson wrote those words in 1841. Nearly two centuries later, trust remains an essential element of any healthy community, which is why I was concerned to read about the rise of AI-enabled “emotional surveillance” technologies in The Atlantic. These products monitor people’s disposition, affect, and concentration during screen-based work. They can tell you if someone appears interested or impatient, agreeable or skeptical. They surveil to produce insights about performance and productivity. And they’re being marketed to schools.

This technology arrives at a moment when societal trust has reached a nadir, which is a problem because “[trust] is key to building and maintaining social cohesion, as it promotes cooperation between groups of people.” Surveillance produces the opposite effect. When people feel they are being watched, they report feeling more stress and anxiety, they feel pressured to work longer and accomplish more, causing burnout, and they explicitly report feeling that their trust is being violated.

So why surveil people? In a word, data.

As is the case for all digital technologies, data is king. The more information one has, the better they can “optimize” for a preferred outcome. Want to maximize productivity? Wonder how your students feel about this activity? Interested in knowing if an employee is a “good fit” for your corporate culture? The answer is always data. And the best way to collect data is to measure everything.

But measurement requires observation, and sometimes that comes at a cost. The consequences of disaffection can outstrip the benefits of what we decide is worth optimizing.

Schools can’t afford to engender feelings of mistrust in their students. At a time when children already feel alienated from their schools and young people are phenomenally pessimistic about the future, sowing distrust is inadvisable.

Instead, schools can be sites of institutional trust. They should be places where young people feel hope.

The Logic of Surveillance

Two distinct technologies have arrived in schools from different directions, but they share a single underlying logic.

The first is Emotional Surveillance, as described above. Sometimes called “emotion AI” or “affective computing”, these platforms monitor schoolchildren’s attention and emotional state. Students complete tasks while the platform accesses their computer’s camera to watch and record them.

“On completion of the learning task the learner gets a summary of their attention and an optional reward,” per one such product called Focus Pocus. The company’s pitch proceeds from the premise that “Learning online can be hard for kids because it’s so easy for them to get distracted from a screen.” Its solution is to surveil them and incentivize sustained attention with rewards rather than to simply offer children learning activities that don’t require distracting screens in the first place.

Another product, MorphCast, focuses on student engagement. “You keep shipping content that doesn’t change behavior,” their website states — as if your students’ disinterest in the civil rights movement means you might consider ‘shipping’ another historical event to them instead. But MorphCast goes further: “Rolling out programs without proof of performance across cohorts makes ROI impossible to link to next-step actions.” Yes, because teachers design their lessons primarily concerned about whether it will yield a good ‘return on investment’ from their students. Most revealing of all, the following image takes centre stage on their website:

What these platforms share, beyond surveillance, is an impossible promise: that a camera can tell you something true about what’s happening inside a child. According to Lisa Feldman Barrett, a neuroscientist and psychologist interviewed by The Atlantic,

“Your movements, whether it’s on your face or in your body or the tones you emit, don’t have inherent emotional meaning. They have relational meaning… There are — I mean, literally, at this point — hundreds and hundreds of studies involving thousands and thousands of people to show that when it comes to emotion, variation is the norm.”

In other words, emotions and their expression are so irreducibly diverse and context-dependent that these products simply can’t work the way they say they do. The instrument doesn’t reveal a student’s inner life; it simply produces a number where the inner life used to be.

Share

Subscribe now

The second technology is Integrity Surveillance. These products have emerged as a direct response to AI-assisted cheating. Rather than assess writing after the fact, like AI detection tools attempt to do unreliably, these platforms require students to complete assignments inside the software. They log keystrokes, record pauses in workflow, and track browser activity. The student is being monitored at all times.

One such example is Packback, a platform that offers both students and instructors a range of features from research aids to grading assistance. But it also tracks how long students spent planning, drafting, refining, and reflecting on their writing. It produces a report that reveals how many writing sessions went into producing the work, how long each session was, and an “originality fingerprint” that estimates the likelihood that AI may have been used. Ironically, the platform itself is powered by AI.

In contrast to my own education, I can’t help but wonder how awful it must feel to try to learn anything from within a digital Panopticon. Surveillance stifles trust, but it also reduces education to a box-ticking exercise. When someone — when some-thing — is looking over one’s shoulder at all times, there’s no space left for curiosity, let alone the slower work of genuine inspiration.

Whether we monitor emotion, attention, or integrity, we can’t forget that the very choice to measure those things exacts a penalty on students. That data, faulty as it may be, is gathered through means that are extractive, invasive, and mistrusting. Any teacher can tell you that there is a hard limit on how much their students can learn — and how they feel about their learning — when they don’t trust that you trust them.

These two technologies look different on the surface because the first concerns itself with wellbeing and the second with honesty, but they’re both expressions of the same institutional reflex: the Marketplace Mirror’s response to the illegibility of genuine learning. Genuine understanding can’t be photographed. You can’t ask an instrument to capture something that can’t be measured. And when you try, you erase the dignity of learning.

This is explained by the philosopher C. Thi Nguyen’s idea of “Value Capture”. The basic premise is that when we attempt to capture an immeasurable complex value, like one’s passion for learning, we inadvertently reduce it to a simplistic and measurable proxy, like engagement. A passion for learning and mere engagement are not the same thing. When we begin to measure the proxy because the complex value is impossible to capture, we slowly start to optimize for the proxy instead.

In Nguyen’s framework, the surveillance technology that measures attention displaces the thing we actually care about: developing students’ genuine passion for learning. We can feel, intuitively and relationally, when our students are developing a passion for what they’re learning, but we can’t measure it.

Value Capture, to the person doing the surveillance, feels like achieving clarity. But that’s precisely what makes it dangerous. If we care about the fullness of educational experience, we have to accept that we won’t be able to measure everything, nor should we.

Share

What the Watched Student Learns

The strongest argument against surveillance in schools is not unreliability — though that’s real enough. It is what surveillance models. Our core objective as educators is not to ensure compliance toward an easily measured goal; it’s to assist in the formation of young people so that they may become trusting, caring, and capable members of a healthy society.

The philosopher Onora O’Neill draws a distinction between trust and control. Trust requires vulnerability and the acceptance of risk. She says, “Where we have guarantees of proofs, placing trust is redundant.” In other words, if a system uses watertight monitoring to ensure that someone performs perfectly, you aren’t actually trusting them; you’re just managing their compliance. Trust only exists where we give up control.

Surveillance produces compliance, not character. If we wish for someone to be trustworthy, we have to, as Emerson suggested, open up the space for trust to take root. A student completing an essay inside keystroke-monitoring software isn’t learning to be honest; they’re learning to perform honesty for the system. This is a different skill entirely, and it’s not one that schools should be teaching. A classroom that surveils its students teaches them that they are suspect, that their inner processes are a liability, and that the school’s relationship to them is adversarial.

O’Neill’s characterization of trust and control is amplified by Nguyen’s thesis. A student whose behaviour is optimized for an integrity score develops the capacity for score-management, not integrity. A student whose emotions are measured continuously develops performance awareness, not self-awareness. Ironically, surveillance produces convincing imitations of the qualities we hope young people develop while stifling their actual formation.

A camera or an algorithm can’t replace the relational — and immeasurable — knowledge that a teacher develops about a student over time, through repeated observation, exchange, and authentic care. As Barrett explains, trying to measure and analyze a student’s emotions actually displaces the opportunity to build relational trust that only occurs between people, not people and machines.

The Walled Garden’s answer to the illegibility of genuine learning isn’t surveillance, but redesigned conditions. Artifacts of Attention — handwritten drafts, annotated sources, and in-class work periods — don’t monitor students for compliance; they create the conditions under which authentic student engagement becomes more likely and more visible. A teacher who reads a student’s essay outline, subsequent drafts, and their final product doesn’t need a keystroke log to know whether thinking and growth occurred. They created the conditions that made thinking possible, and with it, genuine interest in the process.

There is a stark distinction to be made here: assessment that reveals process versus surveillance that monitors compliance. The first treats students as trustworthy learners. The second treats them as untrustworthy liabilities. Both can produce a document. Only one produces a student.

Share

Subscribe now

Schools Built for Trust

Consider what young people are inheriting:

• According to the 2025 Edelman Trust Barometer, an annual global report, only 36% of people believe things will be better for the next generation. 61% believe that government and business make their lives harder and serve narrow interests. And 53% of 18-34 year-olds approve of hostile activism: “attacking people online, intentionally spreading disinformation, threatening or committing violence, damaging public or private property.”

• According to a UN DESA Policy Brief from December of last year, “more than half of the world’s population reports little or no trust in their government.”

Young people in classrooms right now are forming their foundational sense of what institutions are, what they do, and whether they deserve engagement. They’re forming those opinions through their lived experience, not through civics lessons.

The good news is that schools, among institutions, are in a unique position. According to Edelman’s 2026 report, teachers are trusted by 70% of people, second only to scientists. Their 2023 report noted that 64% considered teachers “a unifying force”, higher than any other profession. If we do the math — eight hours a day, across twelve years — it’s clear that what schools model through their practices, rather than their stated values, shapes civic dispositions at scale.

The AMP State of Global Youth Report (2025) reinforces this claim: “the thread that runs through all of these is that the youth trust people they know or people that work directly with individuals far more than they trust systems, platforms, or any political structure.” This makes sense when we consider what we know about trust — that it’s built through relational experience: through fairness, by being heard, and through small acts of consistent care. This is what good teachers do.

Schools, and the professionals who work within them, need to remember that they aren’t passive mirrors of social conditions. Their design choices, the metrics they record, and the software they license are pedagogical and civic acts. Fashion assessment in a humane manner and watch trust grow. Outsource surveillance to an algorithm and watch it erode.

If we want students who will grow into citizens capable of trusting and being trusted, that capacity has to be practised somewhere. The surveillance classroom can’t produce it. The Walled Garden can.

If this piece resonated with you, please consider sharing it with the educators, administrators, and parents in your network. The question of what schools model — not just what they teach — deserves the widest possible audience. And if you’re not yet a subscriber, the Walled Garden grows through exactly this kind of conversation. Join us.

Share

Subscribe now

Leave a comment

Interested in reading more?

The discussion was moderated by Travis Saunders, a professor of applied human sciences at the University of P.E.I. and an advocate for reducing screen use in classroom settings. The other panelists included Matthew Shapiro, a Vice Principal at Westmount High School in Montréal, Québec, and Tim Mushumanksi, the Principal of Trafalgar Middle School in Nelson, British Columbia.

Both Matthew and Tim have been instrumental in advancing successful bell-to-bell phone restrictions in their schools. The discussion covered practical concerns about what makes for a good phone policy, why phone restrictions are necessary, and why we should consider going even further.

Although the conversation occasionally refers to Canada’s unique policy landscape, there is plenty of practical advice that will be useful to teachers, administrators, and parents everywhere.

I hope you enjoy:

Share

Subscribe now

Leave a comment

Interested in reading or viewing more?

Are students using AI to cheat? Can AI improve student learning? How should we deploy it in the classroom?

I think these are the wrong questions.

There’s a Turkish proverb can help us ask better ones:

The forest was shrinking. But the trees kept voting for the axe. The axe was clever — it had convinced the trees that since its handle was made of wood, it was one of them.

A proverb works the same way a good poem works. It asks you to sit with it before you reach for its meaning. So hold that image for a moment. We’ll come back to it.

Proficiency Is Not Literacy

As I wrote a couple of weeks ago, there are two ways to introduce students to AI. The first treats it as a tool to be used. The second treats it as a subject to be understood. The central distinction between these approaches is the difference between proficiency and literacy. Proficiency is knowing how to use something. Literacy, on the other hand, is understanding what something is — how it works, whose interests it serves, what it costs, and what it does to the person using it.

A student who can write an effective prompt is not AI literate in any meaningful sense. A student who understands why the prompt produces what it produces, who built the system and why, and what it costs them cognitively and the planet environmentally has developed critical literacy. The former proceeds from the assumption that the tool’s use is necessary, that its existence is inevitable, and that its utility is unassailable. The latter affords the student critical distance — the distance required to question those same assumptions.

Consider the following. Would you say a student is literate about computer science because she knows how to operate a web browser? Would you consider her literate about contract law because she signed a contract for a summer job?

Neither would I.

AI literacy is no different. The dominant approach to AI in schools today — learning to write good prompts, evaluating outputs for accuracy, and integrating AI assistance into one’s workflow — develops a kind of proficiency (rather than literacy) that normalizes the product’s use, serves AI companies’ interests, and forecloses the possibility of understanding AI as a legitimate subject of critical inquiry.

Genuine AI literacy requires a different curricular approach — one that centres the principles of critical literacy, that isn’t organized around the commercial interests of private corporations, and that belongs in all schools regardless of resources, ideology, or technological infrastructure.

Share

The Examined Tool

Such a curriculum would examine AI using a framework with three constituent elements: What It Is, Who It Serves, and What It Costs.

What It Is — The Architecture

Before students can think critically about AI, they need to understand what it actually is, not at the level of a software engineer, but at the level of an educated person engaging with a consequential technology. This requires them to understand, in plain terms:

Read more

April’s reading converged on the question of evidence-based practices in education. The pieces below document this pattern from multiple angles — the research that gets ignored, the classrooms where someone finally ran the experiment, the students who stopped seeing the point of showing up at all. Taken together, they make the case that our approaches to research and evidence ought to be strengthened.

Below you’ll find what I published this month, plus eight pieces from other writers doing great work.

From The Walled Garden Education This Month

I opened this month’s writing by returning to a conversation I overheard between two students who invoked their “learning styles” to explain their difference in their feelings toward a lesson. Neither student realized they had rationalized their experience using one of education’s most persistent and consequential myths. That exchange is the entry point for a piece about what I’m calling the plausibility trap: the pattern by which ideas that sound right, feel student-centred, and align with the profession’s humanistic values get adopted at scale despite little or no evidential support.

The piece examines three case studies: Whole-Language Reading Instruction, Learning Styles, and the EdTech industry’s expanding catalogue of classroom tools. It explains how each exploited a partial truth and then made specific, testable claims the evidence doesn’t support. The through-line is that plausibility is a poor substitute for evidence, and that the profession’s best instincts — its commitment to individualization, its wariness of what feels mechanical or reductive — are precisely the instincts the plausibility trap exploits. The science of reading movement offers some hope, but it took decades, significant accumulated harm, and considerable institutional resistance before change came. The evidence for making better decisions is available to us now.

The second piece in the series grew out of a brief exchange with an EdTech developer who had built an AI product to teach students about AI on the grounds that training teachers to do it would be too slow and too expensive. I wanted to give the argument a fair hearing before questioning its premises.

The charitable reading is that AI literacy is underdeveloped in most schools, and the pace of change makes the gap feel urgent, but the proposal has a structural problem. A student learning about AI through an AI product built by an AI company with a financial interest in AI adoption is not in a position to think critically about AI. It’s something like contracting a casino to teach about gambling. The piece also pushes back on the assumption that teachers are deficient, or that AI is somehow beyond the reach of human instruction. Teachers teach about nuclear power plants without being nuclear physicists. There’s no principled reason AI should be different. What makes the assumption revealing is who benefits from it. Genuine AI literacy requires the ability to examine a thing from beyond its own borders. When the tool of instruction and the subject of instruction are the same thing, that critical distance doesn’t exist.

The third piece in the series, which will address what genuine AI literacy actually requires, and the institutional conditions that make it possible, is coming early next month.

Share

Subscribe now

Four Voices from Substack Worth Your Time

“Education Research is Weak and Sloppy. Why?” by

This piece is a necessary companion to everything I wrote this month about the plausibility trap. Piper opens with a detailed account of researcher Jo Boaler’s work on discovery-based math instruction, which has been influential enough to get algebra removed from middle schools across the Bay Area. The subsequent investigation that found her methodology to be, in Piper’s assessment, among the most incompetently or dishonestly conducted research she had encountered in a decade as a journalist. Boaler compared the top two quartiles of students at the school she studied to the middle quartiles at other schools; her assessments contained material years below grade level; her claim to have closed the gender gap in math performance was contradicted by every external measure taken during the study period. Piper is clear that this is not an isolated case. The replication crisis that reshaped psychology and medicine has largely passed education by. The mechanisms for correction are weaker, the feedback loops are longer, and the consequences for children are enormous.

“What the Studies Say About How AI Affects Your Brain: A (Very Big) Compilation” by

Romero has compiled more than thirty studies on what AI does to human cognition and synthesized them into a single readable document. The finding that emerges is what Romero calls a performance-competence dissociation: AI reliably improves the quality of immediate outputs while just as reliably degrading the cognitive processes that build durable knowledge and independent reasoning over time. The outputs get better; the person producing them gets worse at producing them. He also identifies the most alarming gap in the entire research base: children are almost unstudied. One fMRI study of thirty-one children exists. Meanwhile, sixty-four percent of American teenagers use AI chatbots and thirty percent use them daily. The developmental neuroscience of AI use is, in Romero’s words, the most urgent gap in the field.

“What Happens When Students Stop Believing Their Work Matters” by

Watkins opens with a student named Drew, who attended every class, sat in the front row, contributed to every discussion, and never submitted a single assignment. Not the ones that invited AI use, and not the ones that forbade it. Watkins has been thinking about Drew ever since, and this piece is his attempt to explain what Drew experienced: not laziness, not burnout in the conventional sense, but something more corrosive. When a machine can now produce a credible version of almost any task a student might be asked to complete, some students don’t respond by outsourcing their work to it; they respond by questioning whether their work was worth doing in the first place. This is a different and in some ways more troubling problem than academic dishonesty. The question for students that Watkins keeps returning to is: “Why am I even here?” Schools need to be able to answer it.

“It’s Not Just a Notification“ by

Kirschner returns with a close reading of a new study on what social media notifications actually do to cognitive processing. The finding is that a single notification slows cognitive performance for approximately seven seconds, regardless of whether the recipient even looks at their phone. Kirschner explains why the number matters at scale: twenty-five students, multiple devices, multiple apps, more than a hundred and fifty notifications received per day on average. The disruption shows up as an invisible loss of processing capacity, as students who appear engaged but are losing the mental thread in ways that don’t register until you look at outcomes. He also introduces the concept of attention contagion — the finding that inattention spreads socially in a classroom — which reframes the individual phone as a collective problem. The piece ends with a policy conclusion that is simple and apparently still controversial: anyone who wants students to learn must protect their attention.

Share

Four Essential Reads from Beyond Substack

“What Will It Take To Get A.I. Out of Schools” by Jessica Winter in The New Yorker

Winter documents what it looks like when a parent discovers that the creepy neighbour they’d been warning their children about has taken up residence inside their children’s school — but in this case, the neighbour is AI. The piece traces how Google’s Chromebook dominance has created a captive market for Gemini and made AI in schools a near-universal prospect, often without meaningful consent from parents or teachers. The piece identifies conflicts of interest embedded in the officials writing AI guidelines, a coalition of parents and educators pushing for a moratorium, a teachers’ union president who says plainly that the more people rely on AI the less they think, and a co-founder of an AI literacy organization who lost sleep when Google made Gemini available to all ages because, she said, they clearly knew it was going to be bad for kids. The piece ends with a question: What if the answer to “What do you want from this?” is simply nothing?

“What Happened After A Teacher Ditched Screens” by Jenny Anderson in The Atlantic

Anderson’s piece follows a seventh-grade math teacher in Colorado who spent more than a decade as an early adopter of classroom technology and then, in January, took the Chromebooks away for a month. He found that assignment completion among his least engaged students rose from forty-five to sixty-two percent when he switched to paper. He noticed that screens had given students cover — a way to appear engaged without sustained effort — and had given him cover too, watching data points move across a screen rather than actually teaching.

“Is Schoolwork Optional Now?” by Lila Shroff in The Atlantic

Shroff takes up what may be the next stage of the same problem. Where early AI concerns centred on students using chatbots to write essays, the emergence of agentic AI tools means students can now outsource entire workflows without meaningful intervention. A tool called Einstein, designed to provoke exactly this reaction, completed an entire online statistics course in under an hour. Its creator released it deliberately, arguing that the technology was already in use and that educators needed to know what they were facing. The piece explains that even students who aren’t using AI are affected by its existence, and what many of them are absorbing is that the work they’ve been asked to do may not mean what they thought it meant.

“I’m a Grade 11 student. This is why attendance is so low” by Gracie Ceriz in The Toronto Star

As a Grade 11 student, Ceriz offers a student’s own account of why attendance is falling. She’s direct about it: when everything is posted online, when lessons are uploaded before class begins, when grades come only from tests and assignments rather than participation, the school building becomes optional in a way it simply wasn’t before. She’s asking adults to consider what the design of the current system actually incentivizes. It’s worth reading alongside everything else in this month’s Roundup, because it’s the closest we get to hearing what the students living inside these systems are actually experiencing.

Share

A Closing Thought

What runs through April’s reading is the gap between what a tool promises and what it actually does, compounded by the institutional reluctance to close that gap until the evidence is impossible to ignore.

The plausibility trap I described in my own writing this month is operating right now, at scale, in the form of AI tools deployed in classrooms where the cognitive costs are still accumulating and the research base is still thin. Romero’s compilation shows us what we know thus far. The fMRI study of thirty-one children shows us how much we don’t. And the student in The Toronto Star, explaining why she doesn’t see the point of showing up, shows us where those costs eventually land.

The corrective is a robust culture of evidence, institutional honesty, and the willingness to ask — before adoption, not after — what a tool actually does to the people who use it. Some teachers are asking that question now, sometimes by removing the tools and watching what happens. Some parents are asking it by organizing. Some students are asking it simply by not showing up.

These questions deserve better answers than the ones most institutions are currently offering.

Until May,

Andrew

Share

Subscribe now

Leave a comment

“I have developed a platform to teach [students] AI literacy because it’s not being taught in schools and [I] believe the growing educational gap in this area is going to leave them behind.”

When asked why AI literacy can’t be taught by a teacher, they responded:

“Unfortunately, it’s not scalable without massive resources and is not happening any time soon…”

There are three claims worth surfacing from this exchange:

1. AI literacy isn’t being taught in schools.

2. It needs to be.

3. Training teachers to do it is too slow and expensive, so an AI product is a practical solution.

The first claim is partially correct. The second is undoubtedly true. The third is where the conversation gets interesting, and where the undefined terms begin to matter. What counts as “AI literacy”? What constitutes “massive resources”? What does “any time soon” actually mean?

I had questions. But before I get to them, the developer deserves a fair hearing.

Share

The Charitable Reading

They identified something real and important. AI literacy is crucial and it’s underdeveloped in most schools right now. According to reporting in Education Week:

Less than half of teachers (48%) have participated in any training or professional development on AI provided by their schools or districts; and less than half of students (48%) said someone at their school provided information to students on how to use AI for schoolwork or personal use… To make matters worse, the content of the training or information provided on AI doesn’t always cover all of the basics.

Training teachers — and doing so with the right information — can be a slow process relative to the pace of technological change. The desire to close that gap quickly is understandable.

AI is publicly available. Students are already using AI. AI can be harmful. Traditional methods of assessment are vulnerable to AI use. We aren’t on top of these problems and risks, as evidenced above. So, can we leverage technology to alleviate these concerns?

If I were a technologist, I’d be asking the same question.

But sometimes, we have a tendency to apply technological solutions to problems that technology created.

Share

Subscribe now

The Casino

That’s the real dilemma. Mobilizing AI to teach students AI literacy is something like contracting a casino to teach about the risks (and benefits?) of gambling.

Imagine what a casino’s gambling curriculum might look like. What would it emphasize? What would it leave out? I doubt that the house would lie outright. It might teach the rules accurately, explain the odds honestly, and even acknowledge the risks of addiction. But the environment of instruction would shape everything. The absence of clocks. The engineered sounds of slot machine bells. The near-miss psychology. The curriculum would exist inside a system designed to produce a specific relationship between the student and the subject — one that serves the house.

An AI company’s AI literacy curriculum has the same structural problem. Even if we afford them positive intent, the problem is about structural design. The environment of instruction shapes what can be thought inside of it, while normalizing the very environment itself. A student learning about AI through an AI product, built by an AI company, with a financial interest in AI adoption, is not in an ideal position to think critically about AI. The walls of the casino are still the walls of the casino regardless of what’s being taught inside of them.

This is the epistemic problem at the heart of the proposal, and it’s one that good intentions can’t fix.

Share

The Structural Problem

The environment within which one learns matters more than EdTech developers tend to acknowledge, but the casino analogy only gets us so far. There are at least three more structural problems with the proposal.

The first is about our ability to scale AI literacy with human, teacher-led infrastructure. The scalability argument is an economic argument dressed up as a pedagogical one. When someone says a solution “isn’t scalable without massive resources,” we must ask ourselves who benefits from the scale. Teacher training is expensive. So is building, marketing, and maintaining an AI product. The difference is who pays, who profits, and who the solution is actually designed to serve.

The second is a structural conflict of interest. In many of our most important institutions — from journalism to medicine and law — we weigh conflicts of interest seriously and without controversy. We don’t allow pharmaceutical companies to design clinical trials for their own drugs without oversight from independent regulatory bodies. This principle exists because good intentions are insufficient when financial interests are fundamentally misaligned with the interests of the people being served. Education needs the same norms.

Finally, the proposal rests on the assumption that teachers are deficient — that AI is somehow beyond the reach of human instruction. Teachers teach about nuclear power plants without being nuclear physicists. They teach about the printing press without operating one. There is no principled reason why AI should be different. What makes this assumption particularly revealing is that it comes from an industry that routinely admits it doesn’t understand how its own product functions. The implicit argument is this: AI is too complex for teachers to explain, but not too complex for an AI product to teach. That’s not a pedagogical claim; it’s a sales pitch.

A teacher doesn’t need to understand how a large language model works to teach students about sycophancy, hallucinations, companionship, privacy, persuasive design, environmental consequence, cognitive effect, mental health, and the ethics of training models on data without permission or compensation.

One has to wonder if such topics would find their way into a literacy curriculum delivered by an AI product.

Share

Subscribe now

The Conditions of Literacy

What, then, does real AI literacy look like?

There are two ways to introduce AI into schools. The first treats AI as a tool to be used. The second treats it as a subject to be understood. The developer’s proposal assumes the first, but genuine literacy requires the second. Only when students understand AI as a subject of inquiry can we responsibly consider its use as a tool.

Technologists often frame critical distance as a luxury subordinated to expediency. Teachers — trained specifically in how knowledge and skills are acquired — know better. It isn’t a luxury; it’s a foundational condition of literacy itself. Literacy requires the ability to examine a thing from beyond its own borders. When the tool of instruction and the subject of instruction are the same thing, those borders and that critical distance don’t exist.

A student whose introduction to AI is mediated entirely by AI isn’t developing literacy, they’re developing familiarity, which, unsurprisingly, is to the benefit of the AI developer. Literacy and familiarity aren’t the same thing. The difference matters enormously.

When we allow vendors to define what students need to know about the vendors’ own products, we’ve confused two fundamentally different relationships: the one that exists between a company and its market, and the one that exists between a school and its students. The first is transactional. The other is fiduciary.

The AI developer is solving a real problem — matching a product to an identified need. When running a company, that’s what responsible development looks like. Schools, however, need to recognize the difference. Their mandate isn’t to reach for privately marketed solutions to their problems. The students in their care are not a market. They never have been.

This piece is the second in a three-part series on evidence and technology in schools. The first examined how plausible ideas without evidential support have quietly shaped classrooms for decades. Next: what genuine AI literacy actually requires — the framework, the content, and the institutional conditions that make it possible.

If this piece resonated, share it with the educators, administrators, and parents in your network. These conversations grow stronger with more voices. And if you find value in this work, please consider subscribing — it’s what makes the research and writing possible.

Share

Subscribe now

Leave a comment

Interested in reading more?

“Honestly, I didn’t love it.”

“No?”

“No, I’m lowkey a visual learner.”

“Valid… Yeah I never thought about it, but maybe I liked it ‘cause I guess I’m like an auditory learner.”

Now, I don’t know the specifics of the activity they were talking about, but it’s safe to say that it didn’t involve watching a video or reading text. Rather, they must have been asked to attend to something by listening.

They clearly had different feelings toward the lesson, but what these students didn’t realize is that they’d rationalized their disagreement using one of education’s most persistent and consequential myths: Learning Styles.

Learning Styles live alongside a raft of other educational theories, practices, and products that command tremendous allure despite evidence that is shallow at best or non-existent at worst. Whole-Language Reading Instruction, for example, promised a naturalistic and low-friction method for teaching early literacy, and the EdTech industry’s expanding catalogue of classroom tools claims to resolve problems schools have always faced, ranging from student engagement to curricular personalization. But research consistently casts doubt on these claims.

The students I overheard proceeded from their exchange believing they’d arrived at genuine self-knowledge. That belief, however, comes at a cost — two costs, in fact.

The first is epistemological. They couldn’t accurately describe their experience because the vocabulary they’d been given was wrong. “I’m a visual learner” explains nothing and eliminates the perceived need to ask more useful questions: “Was the content too difficult?“; “Was the pacing off?“; “Was I distracted?“; “Did I find the topic uninteresting?“ The myth of Learning Styles substituted a tidy label for genuine self-reflection.

The second is developmental. If a student comes to believe she’s a visual learner, then lectures, seminar discussions, podcasts, and conversations with other people are all working against her nature. She now has a permission slip to disengage from an enormous range of human experience and instruction. That permission slip was issued, at some point, by her school.

The responsibility for this doesn’t lie with these students. It lies with the adults and institutions that presented this framework as fact. And Learning Styles are only one example of a pattern that has quietly shaped classrooms for decades — a pattern with consequences that are anything but quiet.

Share

Subscribe now

The Ideas That Felt True

Whole-Language Reading Instruction

Immerse children in rich, meaningful text. Read to them, read with them, surround them with books. Literacy will emerge naturally from exposure in much the same way spoken language does. Drilling kids on phonemes is mechanical, joyless, and reductive. It treats reading as decoding rather than meaning-making.

Or so goes the logic behind Whole-Language Reading Instruction.

Read more

March’s reading converged on two questions that I’ve been working through in my own writing this month: what conditions allow genuine learning to flourish, and what happens when the institutions responsible for creating those conditions prioritize other interests instead? The pieces below document these trends and emerging responses — from Dutch smartphone bans to Ontario grade inflation to the growing parent revolt against screens in kindergarten classrooms.

Below you’ll find what I published this month, plus eight pieces from other writers doing excellent work that matters.

From The Walled Garden Education This Month

When an Oxford student writes that studying literature feels absurd “when the planet already feels like it’s in ruins,” she captures something many young people experience but struggle to name. The percentage of American high school seniors who agree that life often feels meaningless held steady around 15% for decades until approximately 2012, when it began surging toward 27%. Schools have responded to this with an unprecedented focus on preparation: future-readiness rhetoric, skills training, constant vigilance about a changing job market. Yet preparation appears to be compounding the problem rather than resolving it because schools are organizing themselves around the wrong question. The antidote to meaninglessness isn’t more adaptability. Drawing on Simone Weil’s The Need for Roots, written during the Second World War’s most profound uncertainty, I argue that what students need now is what she identified then: real, active, and natural participation in the life of a community that preserves particular treasures of the past and particular expectations for the future. This piece develops a framework for three dimensions of rootedness that schools can deliberately cultivate: cognitive, narrative, and embodied.

The school day described in most North American classrooms is one of efficient passage: students move through subjects, complete tasks, and leave at the end of the day without having had any hand in maintaining the place they just spent seven hours. Drawing on four years working in South Korean schools, where students clean their own classrooms and serve lunch to their classmates as structural features of the school day, not as occasional programmes, I argue that belonging is not produced by proximity; it requires mutual responsibility. Japan’s sōji and Korea’s cheongso dangbeon satisfy Gordon Allport’s four conditions for social cohesion: equal status, common goals, cooperation, and institutional support. Schools that ask students to care for their shared environment are doing something structurally different from schools that simply seat them in the same room. A school is a commons. Commons require stewardship. And stewardship is learned by doing.

I was invited by Smartphone Free Childhood US to participate in its expert panel series alongside two people who together represent the evidence and the execution. Jared Cooney Horvath, neuroscientist and author of The Digital Delusion, makes the most rigorous available case for why EdTech has underdelivered, and why the biology of learning was always going to make that outcome likely. Inge Esping is the principal of McPherson Middle School in Kansas — a school featured in The Economist‘s recent coverage — who has acted on this evidence with striking results. My role was to offer a framework for thinking about what schools can deliberately choose to do differently. The conversation covers attention as curriculum, the Marketplace Mirror model, and the wager every school is quietly making about the kind of people it is trying to produce.

I was a guest on HeightsCast, the podcast of The Heights School in Potomac, Maryland — a school that has thought more seriously about the relationship between education and attention than almost any I’ve encountered. Rich Moss, their Director of Outreach, invited me to discuss what it means to cultivate focus, depth, and genuine thinking in young people at a moment when all three are under serious pressure. We covered the cognitive science of sustained attention, the concept of desirable difficulty, what schools can actually do to protect the conditions for learning, and what a thoughtful institutional approach to AI in K-12 looks like.

Share

Subscribe now

Three Voices from Substack Worth Your Time

“Draft Bill of Research Rights for Educators” by

Kirschner reposts a document by cognitive psychologist Daniel Willingham that deserves far wider circulation than it has received. Its premise is simple: when educators are asked to implement an intervention in the name of research, they have the right to ask what problem it solves, how improvement will be measured, when results are expected, and what evidence exists that it will actually work. The phrase “all the research supports it” has become a blunt instrument used to silence questions rather than a meaningful claim about evidence. Willingham’s six rights are a framework for putting the right information on the table. This document should be pinned to the wall of every staffroom where a vendor has recently given a presentation.

“Don’t Blame Technology” by

Korbey has been an EdTech skeptic for some time, and this month she writes something braver than most: a caution against the emerging consensus that simply removing screens will fix what ails our schools. The backlash against classroom technology is real and warranted, she argues, but the pattern it follows is familiar: we’re reaching for the next silver bullet. Removing screens matters. But if the only thing we do is remove screens, the underlying problem remains. The public lacks a foundational understanding of how learning actually works, and that ignorance is what makes us vulnerable to whatever compelling narrative arrives next. Korbey offers six principles from learning science — the kind of simple, durable knowledge that could act as what she calls a vaccine against future silver bullets. It is as much a piece about epistemics as it is about technology.

“AI Can’t Fix Student Engagement” by and

Only one in three American students is highly engaged at school, a number that has remained stubbornly flat for a decade and predates the arrival of generative AI. Anderson and Winthrop, co-authors of The Disengaged Teen, argue that the current scramble to prevent AI-assisted cheating is addressing a symptom while ignoring a cause: students who are disengaged don’t need better detection tools, they need genuine agency. The piece describes their four-mode framework — Passenger, Achiever, Resister, and Explorer — that gives teachers and parents language for what they’re observing and practical tools for responding to it. The deepest insight here connects directly to the rootedness work I’ve been developing: AI chatbots promise to remove the friction of learning, but that friction isn’t a design flaw. It’s where learning happens. A school that cultivates Explorer mode is building something AI cannot replicate.

Share

Five Essential Reads from Beyond Substack

“America’s Math and Reading Scores Tanked After Schools Ditched Textbooks for Screens — And AI Could Worsen the Brain Rot” by Sasha Rogelberg in Fortune

Rogelberg synthesizes recent Congressional testimony from Jared Cooney Horvath, who noted that Gen Z is the first generation in modern history to be less cognitively capable than their parents, with the broader pattern of EdTech adoption. The piece traces the failure of automated learning from Sidney Pressey’s 1924 teaching machine through Google’s Chromebook rollout to the present moment with AI, identifying a consistent pattern: students perform adequately while using the tool and lose the skill as soon as they set it down. The most clarifying line belongs to Horvath: the tools experts use to make their lives easier are not the tools children should use to learn how to become experts. When you give novices the offloading tools of experts, they don’t learn the skill. They learn dependency.

“iPads in Kindergarten, YouTube Videos at Snack Time: Parents Are Pushing Back on Screen Time in the Early Grades” by Jackie Mader in The Hechinger Report

Mader documents parents across the United States discovering, often to their horror, that their youngest children are spending significant portions of the school day watching YouTube ads, playing gamified apps with no demonstrated learning benefit, and absorbing influencer culture instead of learning to make friends. One kindergartner came home knowing jingles from diaper and car commercials. Another, after having her dance filmed, looked at the camera and said, “If you like what you saw, click below to subscribe.” The piece also documents the educators and districts beginning to push back — the third-grade teacher in New York who replaced her smart board with a paper easel and says she can see students “detoxing under my eyes” — and the growing legislative movement in states like Utah, Vermont, Massachusetts, and Missouri. The youngest students have no say in these experiments. This is what it looks like when we stop treating that as incidental.

“Teens Are Using AI-Fueled ‘Slander Pages’ to Mock Their Teachers” by Owen Carry in Wired

Students on Instagram and TikTok are using AI video tools to insert teachers’ faces into degrading scenarios and distribute them at scale — posts that in some cases accumulate over one hundred thousand likes. These accounts have been dubbed “slander pages,” and they represent something more than a digital mutation of the high school prank. They document what happens when young people are socialized through a culture of constant content, where a face becomes the viewer’s to laugh at rather than a person’s to respect. One researcher describes it as the knock-on effect of being reflected through the internet rather than a mirror. School administrators are beginning to pursue disciplinary and legal consequences. The piece is difficult to look away from because it illustrates what the attention economy, given enough time and enough access to developing minds, actually produces.

“I’ve Taught School for 30 Years. Yes, Grades Are Being Inflated. No, It’s Not Being Fixed Yet” by J.D.M. Stewart in The Toronto Star

A veteran Ontario teacher writes plainly about a problem that everyone in education knows and almost no one addresses directly. In his first year of teaching, a handful of students achieved grades in the 90s. In his last, more than half did, and virtually none finished below 80. CBC analysis of Council of Ontario Universities data tells the same story: average entry grades at McMaster, for example, rose from 84.1% in 2006 to 92.6% in 2021. The piece proposes four concrete remedies: moderated marking, exemplars, moving away from percentage-based grading, and renewed professional development in assessment and evaluation. The most uncomfortable observation is the simplest: grade inflation helps no one, and we have known this for years. The gap between what grades communicate and what students have actually learned is not a peripheral issue, it is a direct measure of whether schools are willing to tell the truth.

“Has banning phones improved performance at Dutch schools?” by Anna Holligan in BBC News

Two years after the Netherlands implemented a nationwide smartphone ban in schools, the early data is encouraging. A government-commissioned study of 317 secondary schools found that roughly three-quarters reported better concentration, nearly two-thirds noted an improved social climate, and around a third saw better academic performance. Teachers describe less friction in classroom management and calmer hallways. Students, including ones who were initially resistant, report talking more and socializing in ways they hadn’t before. Now the Dutch government wants to go further, pushing for an EU-wide minimum age of 15 for platforms like Instagram, TikTok, and Snapchat. Perhaps most striking: a Unicef survey of more than a thousand Dutch children and teenagers found that 69% supported a social media ban for under-18s. The piece challenges the assumption that young people want to be permanently online, and offers a model for what institutional resolve, rather than legislative battle, can accomplish when schools, parents, and government move together.

Share

A Closing Thought

What strikes me most about March’s reading is that it circles, from many different directions, around a single problem: the gap between what institutions say they are doing for young people and what they are actually doing to them. Grade inflation papers over the gap between performance and achievement. Passive screen time fills the gap where learning could be. AI-generated slander fills the gap where respect and human dignity could be. Future-readiness rhetoric fills the gap where rootedness and belonging could be.

None of these gaps opened overnight, and none will close quickly. But there is hope in the growing willingness to name them — in the parent who testifies before a state legislature, the teacher who puts the devices in a cart and covers it with an easel, the veteran educator who writes publicly about what has been quietly understood for years. These are acts of institutional and personal honesty that instigate positive change.

The conditions that support genuine learning have always been the same: sustained attention, meaningful human connection, time with complex ideas that demand careful thought, and the experience of belonging to something you are responsible for. None of those conditions are technologically mediated. All of them are within reach of any school willing to prioritize them.

Thank you, as always, for reading and for engaging with this work. It matters because you bring it into your classrooms, your schools, and your conversations.

Until April,

Andrew

Share

Subscribe now

Leave a comment

If you’d prefer, the episode is also available on Spotify and Apple Podcasts.

If you found this useful, please pass it along to a colleague, a department head, a parent, or anyone thinking seriously about education today. These ideas travel furthest through the people who care about them most.

I’m available for speaking engagements at schools, conferences, and professional development events. To inquire, contact me here on Substack or via LinkedIn.

Share

Subscribe now

Leave a comment

Interested in listening, viewing, or reading more?

The Meaninglessness Surge

Two weeks ago, Natalie Conboy, an undergraduate student at the University of Oxford, published a reflection in Cherwell, the student newspaper, that captures something many young people feel but struggle to articulate:

When the planet already feels like it’s in ruins, emerging with a Masters in some niche corner of French literary history does seem like a somewhat absurd endeavour. The guilt of not contributing towards a better future with each passing moment can lead to inertia: not feeling able or willing to do small things because I can’t single-handedly save the world.

Conboy’s paralysis has a popular name: ‘Doomerism’. This nihilism or hopelessness is a growing sentiment among youth today. In the Monitoring the Future study, the percent of American high school seniors who agree that ‘life often feels meaningless’ held steady around 15% for decades, until approximately 2012. Then it surged, nearly doubling to ~27% by 2023.

This surge in meaninglessness tracks closely with the well-documented collapse in youth mental health.

But Conboy’s worries aren’t held without reason. She and her generational peers consistently report concerns, grounded in evidence, about the climate, economic uncertainty, and political instability. 37% of Canadian teens, for example, report that climate change impacts their mental health, and “nearly half of Gen Zs (48%) and millennials (46%) say they do not feel financially secure,” according to an international survey from Deloitte.

Schools have responded to this uncertainty with unprecedented focus on preparation. Future-readiness. Skills training. Constant vigilance around a changing job market. We’ve never worked harder to equip students for what’s coming.

So why do they feel more hopeless than ever? Why does preparation create fragility?

The answer lies in what we’re preparing them for, and what we’re failing to provide now.

Share

Building on Quicksand

Edward Osborne Wilson, the American biologist, famously observed that humanity suffers from “paleolithic emotions, medieval institutions, and godlike technology”. This discord, between what we are, how we organize ourselves, and what we’ve invented, lies at the heart of many contemporary problems.

Education faces a similar dilemma with equally consequential stakes. The developmental needs of children operate at the pace of biology. Human formation hasn’t changed: adolescents still need stable relationships, opportunities for sustained attention, and time to develop a coherent identity. But the economy moves at the pace of technological innovation. And when schools look to that economy for guidance — when they adopt what I call the Marketplace Mirror model — they shackle biological development to technological disruption.

The result is predictable: students feel perpetually off-balance, as outlined above. Each wave of innovation demands new skills, new competencies, new ways of being. Yesterday’s essential skillset becomes today’s obsolete knowledge. The pathways that seemed secure dissolve. And under these conditions, young people are trained not to develop deep expertise or stable identities, but to be perpetually reactive, always adapting and never rooted.

The Marketplace Mirror model rests on a well-intentioned but calamitous assumption: that preparing students for a changing future means immersing them in constant change now. But a mirror facing the marketplace isn’t guaranteed to reflect something stable to orient toward. It can, instead, reflect a maelstrom.

What students actually need is not to be trained (exclusively) for constant adaptation, but the cultivation of what persists in spite of external conditions. They need stability precisely because the external world can, at times, offer very little. They need what philosopher Simone Weil, writing during the Second World War’s profound uncertainty, called “the most important and least recognized need of the human soul”: rootedness.

Share

Subscribe now

What Uncertain Times Actually Demand from Schools

Read more

In the first, students proceed through a sequence of activities designed to facilitate their individual development. They open laptops, complete tasks, and submit work. They eat lunch in small groups, or sometimes alone, scrolling through their phones. Many leave at the end of the day having had no hand in caring for the place that they just spent seven hours. The school is a building they pass through.

In the second, the school day is inhabited by its students. They assist in the preparation and service of lunch to their classmates. At the end of the day, they help clean the corridors, wipe down desks, and discard trash. The institution belongs to them, not rhetorically, but in practice. They demonstrate this belonging every day, with their hands.

I spent four years working in South Korean schools. The second version is not a thought experiment. It’s simply what school looks like there.

Share

Atomization as Design Outcome

Most of the essays I’ve written under the Walled Garden framework have focused on the cognitive dimensions of what schools owe their students: protecting opportunities for sustained attention, cultivating working memory and executive function, and preserving the “desirable difficulties” through which their capacities are built. These arguments rest on the claim that schools must be sanctuaries, deliberately designed environments that protect the developmental conditions children actually need, rather than Marketplace Mirrors that reflexively adopt whatever the broader culture offers.

But there are further dimensions of the framework worth articulating. A few weeks ago, I quoted Paulo Freire:

No one liberates anyone else, nor do they liberate themselves alone. People liberate themselves in communion.

I used that line to highlight the social costs of relying on AI in K-12 education, whereby students may be siloed and isolated into hyper-individualized pathways. But Freire’s insight can serve as something more: a design principle.

A Walled Garden Education isn’t only a sanctuary for cognitive development; it’s a community that students help to build and maintain. That community membership — the experience of belonging to something you are responsible for, alongside others who bear that responsibility equally — is itself developmentally fruitful.

In North American schools, three forces that I’ve touched on in the past, converge to undermine this possibility:

1. The first is the algorithmic personalization of learning. Adaptive platforms, which I examined through McLuhan’s tetrad in “The Civic Cost of Algorithmic Pedagogy”, create individualized pathways that deteriorate opportunities for shared intellectual struggle. What they render obsolete, when pushed to extremes, is negotiated understanding — the face-to-face work of deliberation and forging meaning with others. Students in these environments learn to justify answers to algorithms, which tend to be a less demanding audience than other human beings.

2. The second is the collapse of embodied experience. In “When Childhood Conditions Change, Schools Must Adapt”, I documented the measurable decline in physical play, unstructured time, and face-to-face peer interaction. I argued that schools have failed to counteract this trend, often replicating its conditions through one-to-one device programmes and screen-mediated instruction. In one of my early essays, I described an impassive student standing in the schoolyard on a beautiful spring afternoon, completely indifferent to the physical world in front of him. His experience isn’t merely anecdotal, it’s an outcome of our design choices.

3. The third is the reduction of school to a service-delivery institution. When schools are designed primarily to transmit content and assess its acquisition, students are understood as recipients, or as customers, rather than participants. Nothing in this model asks them to take responsibility for the community they inhabit. The relationship between student and school is transactional, and transactional relationships do not produce belonging.

What these three forces produce together is not only the risk of cognitive atrophy, as I’ve documented elsewhere, they lead to social atrophy. Students practiced in the private mastery of strictly defined skills, but not in public responsibility, are optimized for individual metrics without consideration for collective obligation.

Share

Subscribe now

Looking East: School as Commons

In South Korea, and in Japan — where the practices are formally codified in the national curriculum under the framework of tokubetsu katsudo, or tokkatsu — students carry responsibilities that extend well beyond academic performance. Every day, on a rotating basis, students clean the school. They sweep floors, wipe tables, and maintain the common spaces they share. In Japan, this practice is called sōji. In Korea, students rotate through 청소 당번 or cheongso dangbeon (cleaning duty). In both counties, students also serve lunch to their classmates. They don aprons, collect food from the kitchen, and eat the same meal together. Teachers eat alongside them.

These are not service-learning initiatives. They are not character education programmes scheduled once a month. They are structural features of the school day, embedded in the timetable and treated with the same seriousness as academic instruction. These practices extend throughout a student’s entire school career. The pedagogue Yutaka Okihara describes sōji as “an essential school experience that encourages a child’s sense of responsibility, cooperation, cleanliness, interpersonal and socialization skills and duty towards the community and society,” and adds that cleaning is “much more than a goal in itself, but rather a tool to reach a more important goal.” In the Japanese curriculum, sōji is intended to inculcate a sense of responsibility within a social context.

I watched this logic operate in practice for four years. What struck me most wasn’t the cleaning itself, but what it communicated. Students didn’t see themselves as guests in an institution maintained by others. Through their experience, the school was theirs, and they understood this because they were asked to demonstrate it every day.

What These Practices Actually Do: The Pro-Social Mechanics

Three things happen when schools are structured this way:

1. The first is that students begin to see and understand each other outside of a hierarchy of individual academic performance. When a student serves lunch to a peer, or sweeps a hallway alongside a classmate they wouldn’t have otherwise spoken to, they encounter each other as equals who depend, cooperatively, on each other in a shared task. Gordon Allport theorized that social cohesion emerges when a group has equal status, common goals, cooperation, and institutional support. Japan’s sōji and Korea’s cheongso dangbeon satisfy all four conditions. The school that asks students to clean together is doing something structurally different from the school that simply seats them in the same room.

2. The second is that distributed responsibility brings community to life. Civic capacity, as I argued in “The Civic Cost of Algorithmic Pedagogy”, is not produced by private mastery. It’s produced by public participation, by contributing repeatedly to something beyond oneself. These can be small acts, but the habit of contribution, practised from a young age, is the foundation of civic capacity later in life. It’s difficult to expect people, let alone students, to care for institutions they haven’t actually been required to care for.

3. The third is that communal responsibilities ground the experience of school in the body. In a world that can seem bent on the disembodiment of childhood — the displacement of physical, unstructured experience by screen-mediated alternatives — sōji and cheongso dangbeon offer a bulwark against this trend. They require presence, effort, and physical coordination with others. Unlike screen-mediated interactions, these routines offer students an experience of having done something real, together, with their hands. As I’ve written previously, philosophers like Charles Taylor and Hubert Dreyfus have argued that the body is not incidental to identity and community membership; it’s the site where both are formed. When students use their hands to care for a shared space, they’re enacting that membership.

   Share

The Walled Garden as Inheritor of this Tradition

A Walled Garden Education, as I’ve explained across my work, is a framework of conscientious, intentional design. It asks what young, developing human beings actually need, rather than privileging what the market currently values. What this piece adds to that framework is the claim that belonging is a developmental need, one that is not produced automatically by proximity to others, and not produced at all by siloed personalization, but by the repeated, embodied experience of mutual responsibility.

The school day as I described at the beginning of this piece exists. It’s not a utopia, but it is a design choice that hundreds of millions of students in East Asia experience as ordinary. It doesn’t need to be replicated wholesale in North America (or wherever you may be); culture is not easily transplanted, and the argument is not about cultural imitation. It’s about the underlying logic: that a school is a commons, that commons require collective stewardship, and that stewardship is learned by doing.

Years after leaving Korea, I often think of an image that became an ordinary daily sight for me: a group of 13-year-olds sweeping a hallway at the end of the day, without supervision, laughing and joking while finishing their work, because that’s simply what you do. They weren’t being prepared for community. They were already in one.

That is a design outcome. And design outcomes can be chosen.

Share

Subscribe now

If this piece resonated with you, I’d be grateful if you’d share it with the educators, administrators, and parents in your network.

If you’ve spent time in a school system outside North America, or if you’ve witnessed something like what I’ve described here, I’d love to hear about it in the comments. This is the kind of piece that gets richer with other people’s experience added to it.

And if you find value in this work, please consider becoming a paid subscriber. It’s what makes the research, the writing, and the ongoing development of this framework possible.

Leave a comment

Interested in reading more?

I was joined by two remarkable people. Dr. is a neuroscientist and educator whose latest book, The Digital Delusion, makes the most rigorous and readable case for why EdTech has underdelivered, and why the biology of learning was always going to make that inevitable. Inge Esping is the principal of McPherson Middle School in Kansas, who has put these ideas into practice with striking results. Between the two of them, you get the evidence and the execution. My role was to offer something in between: a framework for thinking about what schools can deliberately choose to do differently.

The conversation covers attention as curriculum, the Marketplace Mirror model, what protective school environments actually look like, and the wager every school is quietly making about the kind of people it’s trying to produce. I’m proud of how it came together.

If you’re interested in bringing these ideas to your school, board, or organization, I’m available for speaking engagements. You can reach me by replying directly to this email, sending me a message via Substack, or on LinkedIn.

Enjoy the discussion below:

Share

Subscribe now

Leave a comment

Interested in reading more?

February was consumed by a single question: What happens to a generation of students when the institutions responsible for their development look the other way?

From The Walled Garden Education This Month

Most research on technology’s harms arrives after the damage is done. The Brookings Institution attempted something different: a premortem on AI in K-12 education, drawing on more than 500 stakeholders across 50 countries and a review of more than 400 studies. Their headline conclusion: at this point, the risks of AI in education overshadow its benefits. This digest examines what the report gets right, where its optimism strains against reality (particularly its fixability assumption, its lack of developmental precision, and a paradox at the heart of its equity framing), and what schools must add where Brookings leaves things implicit: sunset clauses, genuine audit processes, and institutional protection for educators who judge a tool harmful and refuse it.

(reposted from the archive for its thematic relevance)

Alpha School compresses core academics into two hours of AI-mediated instruction and calls what remains “personalization”. Running McLuhan’s tetrad against that model reveals what the pitch obscures: algorithmic pedagogy enhances efficiency while obsolescing the negotiated, contested, argumentative practice through which civic capacity is built. When correctness becomes what a screen-based AI tutor privileges rather than what a community tests in public, students learn a new grammar of legitimacy — one that serves private interests and erodes the democratic habits schooling helps to form.

Paulo Freire understood that literacy wasn’t just a skill, it was the foundation for political participation, critical consciousness, and the capacity to resist oppression. That insight has never been more urgent. Drawing on Isaiah Berlin’s distinction between negative liberty (freedom from interference) and positive liberty (freedom to act according to one’s own reason), this piece argues that AI threatens both: it strips students of the protected conditions development requires, and allows the capacities that make genuine freedom possible to atrophy through disuse. The emancipatory wager — the institutional bet on capacity over efficiency — is a commitment to the kind of society we ought to build.

Share

A Bonus Appearance on the After the Bell Podcast

I was invited by “The Stunt Brothers”, and Roy Hunt, to talk all things ‘Walled Garden’ on their K-12 education podcast. Martin and Roy are veteran teachers from the north shore of Vancouver. This free-ranging and lighthearted discussion covered everything from student attention to the carbon cost of artificial intelligence. I hope you find it valuable:

Three Voices from Substack Worth Your Time

“Common Arguments About AI in Education — And Why They Fail” by Lily Abadal

Abadal dismantles five claims that have become slogans in debates about AI in education: that AI is just a calculator, that meaningful assignments prevent cheating, that students only cheat themselves, that AI-completable subjects are obsolete, and that AI frees teachers for “the human side” of their work. Each argument, she shows, misunderstands what education is actually for. The deeper question is about what kind of human beings we’re trying to form, and whether our practices remain mindful of that.

“Silicon Valley’s Mythology of Human Amplification” by Tim Requarth

When Silicon Valley upgraded Steve Jobs’s bicycle metaphor to a steam engine, something important was lost: a cyclist travels, a passenger is travelled. Requarth traces this distinction through GPS adoption among Inuit hunters in the Arctic, where a tool that bypassed wayfinding skills rather than extending them produced measurable erosion of intergenerational knowledge — a pattern a recent Anthropic study confirms in software developers using AI. The executives pushing AI as amplification ask what output a human produces. The Inuit elders asked what kind of person a tool produces. That difference in values is what the steam engine metaphor asks us to concede.

“How Much Cognitive Damage Does A Phone Notification Actually Do?” by Carl Hendrick

A new study confirms that a single notification disrupts cognitive processing for approximately seven seconds, which may seem trivial in isolation, but is damaging at scale when the average person receives more than 150 per day. The most counterintuitive finding: total screen time barely predicts cognitive disruption, but notification frequency does. Notifications are engineered to feel personally relevant, and brains conditioned through thousands of daily reward cycles respond accordingly. Every notification, Hendrick concludes, is a small vote for shallowness.

Share

Four Essential Reads (and One Listen) from Beyond Substack

“Reading — and Thinking — in a Digital Age” on CBC IDEAS (podcast)

This episode of CBC’s long-running IDEAS podcast brings together book historian Adriaan van der Weel, neuroscientist Maryanne Wolf, and three other scholars to ask what we lose when reading becomes shallow. The average person consumes more written words than ever before — scrolling, skimming, scanning — and yet the kind of deep, sustained reading associated with critical thinking is in measurable decline. The answer that emerges across each conversation is the same: the medium shapes the mind, and a culture optimized for speed and fragmentation produces thinkers optimized for exactly that. As our information landscape fills with disinformation and AI-generated text, the capacity to read carefully may be the most important thing we’re losing.

“It’s the Most Urgent Public Health Issue’: Dr. Rangan Chatterjee on Screen Time, Mental Health — and Banning Social Media Until 18” by Emine Saner in The Guardian

Chatterjee, a former GP turned podcaster, makes the case that children’s screen use is the defining public health crisis of our time, one that clinicians have recognized for over a decade while governments have looked away. He calls for raising the minimum age for social media to 18, abolishing screen-based homework, and ending the expectation that tech companies will regulate themselves. The most powerful moment in the piece is also the simplest: a suicidal 16-year-old arrives at his surgery already prescribed antidepressants, and Chatterjee addresses his screen use instead. Two months later, his mother writes to say he’s a different boy.

“A Landmark Social Media Addiction Case Puts Big Tech on Trial” by Varsha Bansal in Wired

Approximately 1,600 plaintiffs, families and school districts from across the United States, have filed suit against Meta, Google, Snap, and TikTok, alleging that these companies knowingly designed addictive platforms that drove children toward depression, self-harm, and suicide. The first bellwether trial is now underway in Los Angeles. The plaintiffs’ strategy sidesteps Section 230 by targeting design decisions — infinite scroll, autoplay, algorithmic amplification — rather than user-generated content. One attorney who spent 25 years representing asbestos victims calls it asbestos all over again. He means it as a warning about what happens when commercial interests are permitted to define acceptable risk on behalf of children who can’t consent.

“Stop Meeting Students Where They Are” by Walt Hunter in The Atlantic

Hunter, an English professor at Case Western Reserve University, walked into his American literature survey last fall convinced by the prevailing wisdom that students could no longer read whole books and emerged convinced the wisdom was wrong. He assigned Faulkner, Cather, Harriet Jacobs, and Thoreau anyway, replaced take-home essays with timed in-class “flash essays” to foreclose AI assistance, and discovered that his students rose to meet the demand. The reaction to declining reading skills, he argues, should not be to negotiate or compromise but to double down on the cure. Don’t meet students where they are. Stop somewhere ahead and wait for them to catch up.

“‘Lobotomised by AI’: Lecturers Blow Whistle on AI Cheating Crisis as Universities Refuse to Help” by Ros Thomas in The Australian

Thomas spent months interviewing students, lecturers, and academics across Australia, and what she found is difficult to look away from. Senior academics estimate AI fraud rates above 80 percent; the students she spoke to put it closer to 95. One graduate outsourced his entire final year to ChatGPT — every assignment, every exam — and finished with a High Distinction. A 13-year-old uses it for everything and doesn’t bother faking wrong answers anymore. Universities, meanwhile, have largely chosen institutional silence over enforcement, leaving detection and consequences to exhausted individual staff. The piece ends where the crisis began: with a system that has commodified learning to the point where students see cheating not as a moral failure but as rational adaptation.

What strikes me about this month’s reading is how much of it documents not just the problem but the response. Chatterjee is campaigning. Hunter is assigning whole books. A professor in Perth is bringing students back into rooms for supervised exams. Awareness is growing, and with it, the willingness to act. The question of what schools owe their students is being answered, one classroom at a time.

Share

Subscribe now

Leave a comment

This week’s piece begins with inspiration from a Brazilian educator:

On Literacy and Liberty

Paulo Freire came to understand education not as a method for filling empty vessels with knowledge, but as a tool of liberation. He developed this philosophy in 1960s Brazil, working among impoverished, illiterate peasants (camponeses) while his country stumbled toward military dictatorship. When the coup came in 1964, approximately 40% of the country’s population could neither read nor write.

Freire’s philosophy rested on a simple observation: illiteracy wasn’t just an educational deficit, it was a tool of oppression. In Brazil at the time, illiterates couldn’t vote, so keeping peasants illiterate was a way of keeping them politically powerless. But his insights went a step further — the people he worked with had internalized their own oppression. They believed they were incapable of learning, that education was for other people, and that their own knowledge and experience had no value.

In 1968, while in exile in Chile, Freire wrote:

“Education either functions as an instrument which is used to facilitate integration of the younger generation into the logic of the present system and bring about conformity or it becomes the practice of freedom, the means by which men and women deal critically and creatively with reality and discover how to participate in the transformation of their world.”

Education, in its most essential form, is emancipatory. It builds the capacities that make genuine freedom possible: the ability to think independently, express oneself coherently, resist manipulation, and participate meaningfully in civic life. But emancipation requires capacity, and capacity requires practice, an effortful, friction-filled practice.

Today, when a ninth-grader submits an essay written by ChatGPT — his teacher unable to tell the difference — we must ask ourselves if that student has been afforded, or stripped of, an opportunity to build his emancipatory capacities. Is his a path toward liberation, or does it point in the direction of dependence? Has the appearance of learning replaced its substance? Which kind of education are we building?

When we allow students to outsource cognitive work to algorithms, we don’t make them freer; we make them reliant and vulnerable. The question before schools now is whether we will build capacity or ensure that a generation is defined by its dependence on tools it doesn’t understand and can’t do without.

Share

Freedom Requires Capacities, Not Just Opportunities

In the decades that followed, Freire’s ideas reverberated domestically and across the globe. At home in Brazil, literacy rates grew and democracy returned to the country. Abroad, his philosophy played a central role in South Africa’s anti-apartheid movement, and research institutes bearing his name were founded in the United States, the United Kingdom, Canada, and elsewhere.

Through education, Freire’s theories expanded freedom around the world. But what does it mean for freedom to grow? To answer that question, we turn to another scholar: Isaiah Berlin.

Read more

Marshall McLuhan, the famed media theorist, philosopher, and fellow Canadian, offered us a simple but consequential insight: technologies don’t only do things for us, they restructure the conditions under which we experience the world. His tetrad, four simultaneous questions about any medium (what it enhances, what it renders obsolete, what it retrieves, and what it flips into if pushed to extremes), is a diagnostic instrument. When applied to a school that places artificial intelligence at the centre of pedagogy, the results warrant careful scrutiny.

Alpha School, a new “AI powered private school”, sells a radical promise. Core academics are compressed into two hours of AI-mediated learning in the morning. Afternoons are reserved for “life skills” and “passion projects”. Adaptive software serves one on one instruction at scale, adults in the room are “guides” rather than certified teachers, and Alpha advertises dramatic gains in efficiency and measurable mastery. The school’s pitch, “Learn 2x in 2hrs”, is insistently attractive to parents and investors alike. Tuition is not incidental: in many locations Alpha is a high-cost private option, and its roll-out is accompanied by heavy capital and prominent backers.

We should be clear: the question here is not whether algorithms can perform certain instructional tasks. The question is whether a model that makes algorithms the primary pedagogical intelligence is compatible with democratic education — an education meant to form people capable of responsible disagreement, to deliberate in public, to suffer contradiction, and to repair and maintain the civic fabric. Read through McLuhan’s four questions and the structural concerns become visible.

Subscribe now

What Alpha enhances

Read more

Most research examines harms after the fact. It took decades to establish evidence-based regulations on the use of asbestos in construction and leaded gasoline in the auto industry. Now, fifteen years after smartphones put social media in every adolescent’s pocket, we’re beginning to respond to the risks it imposes on children. But a landmark report from The Brookings Institution attempts something new: a “premortem” on the risks and benefits of a new technology affecting K-12 education.

In an effort to anticipate the potential harms and advantages of using artificial intelligence in the classroom, Brookings conducted a yearlong study surveying more than 500 stakeholders including students, teachers, caregivers, and technologists from more than 50 countries. They reviewed more than 400 studies on the use of AI in education, and assembled a Delphi panel to gather expert opinion. Their headline conclusion?

“Ultimately, we find that at this point in its trajectory, the risks of utilizing AI in education overshadow its benefits.”

While Brookings recognizes that AI is an evolving technology with risks and benefits that are likely to fluctuate over time, they chose to demonstrate institutional responsibility by acting according to the Precautionary Principle: “when an activity raises threats of harm to the environment or human health, precautionary measures should be taken even if some cause and effect relationships are not fully established,” especially when the vulnerable cannot consent.

This digest examines what the report tells us, where its findings align with classroom realities, and what schools must prioritize as they respond.

The Doom Loop: Why These Risks Undermine the Benefits that AI Purports to Offer

Cognitive Offloading as Developmental Disruption

Brookings’ central insight is that the risks that AI introduces to education erode the bedrock of learning; they impede the very capacities needed to realize any benefits. Specifically, the report states, “these risks undermine children’s foundational development.” AI produces a kind of developmental doom loop:

• Students offload cognitive work to AI → capacities atrophy → students become more dependent → further offloading required → capacities deteriorate further

While technologies like calculators or spell check assume students have basic number sense and literacy, AI operates at a different level. Unlike those other tools, AI can complete the entirety of a given cognitive task rather than support an aspect of it. Under those circumstances, it takes only one difficult assignment, one moment of frustration, or one stressful deadline for a student to find themselves caught in the whirlpooling effects outlined above.

In this new paradigm, teachers find that their responsibilities extend beyond their traditional scope of practice. The dissemination of content knowledge and the acquisition of defined competencies now rest upon a deeper, more fundamental responsibility: the preservation and protection of at-risk developmental conditions.

Suddenly, the educator’s task is to protect their students from hazards to their most basic cognitive capacities: working memory, sustained focus, executive function, etc. AI threatens to remove encounters with “desirable difficulties” during the precise developmental windows when brains need those challenges the most.

Unlike gaps in knowledge or skill that may be remediated later, developmental windows close — what doesn’t develop during those critical periods may not develop at all.

But the doom loop doesn’t stop at cognition; it cascades into every dimension of development.

Share

The Full Spectrum of Harm

Beyond cognition, Brookings identifies other cascading risks:

• Social-Emotional: Reduced face-to-face interaction means less practice with emotional regulation, conflict resolution, negotiated understanding, and empathy development

• Relational Trust: Teachers unable to verify whether students’ work is their own, and the displacement of pedagogical authority from teacher to machine

• Equity: Emerging evidence suggests that privileged students retain access to human teachers while others are more likely to receive algorithmic instruction

• Privacy: Commercial surveillance embedded in learning; data extraction as the cost of “free” tools

• Safety: Age-inappropriate content, manipulation risk, inadequate guardrails for vulnerable users

These are unacceptable trade-offs for convenience. As I wrote a couple of weeks ago, vulnerable populations like children require protective boundaries, not optimized engagement.

Where the Report’s Optimism Meets Reality

Brookings concludes that AI’s harms are “fixable” and distinguishes between “AI-diminished” learning (current reality) and “AI-enriched” learning (aspirational possibility). The report identifies two categories of potential benefits:

1. AI as a productivity tool, freeing teachers from “routine tasks” such as “lesson planning, grading, assessment design, [and] administrative work.”

2. AI for personalization, offering “learning benefits directly to students through adaptive platforms, writing tools, and tutoring programmes that tailor instruction to individual needs.”

These aspirations rest on three critical assumptions worth examining: design feasibility, developmental precision, and equity. Yet some of the proposed benefits contain inherent tensions. AI-generated lesson plans appear as both a productivity gain (freeing teacher time) and a potential risk (teachers offloading professional judgement about curriculum sequencing, learning objectives, and pedagogical choices). The report doesn’t reconcile when teacher use of AI becomes teacher dependence on AI.

Share

The Fixability Assumption

The report acknowledges that, “While AI’s potential negative risks and the damages it has already caused are daunting, they are fixable.” It points to a description of enriched learning where AI tools are designed with the following principles in mind:

• The AI is “purposefully designed for learning for children and youth.”

• The AI has “guardrails to mitigate attachment, overuse, and cognitive offloading.”

• The AI is “built on principles of learning.”

• The AI is “designed to teach, not tell.”

These are reasonable aspirations, but they describe AI as Brookings hopes it could be, not AI as it currently exists. More critically, the report provides no assessment framework for evaluating whether existing tools meet these standards.

The danger is in schools reading these criteria and assuming their chosen tool qualifies. Design aspirations are necessary but insufficient. Schools also need assessment criteria for evaluating existing tools:

• Does it preserve productive friction (desirable difficulties)?

• Does it make cognitive process visible (drafts, annotations, artifacts of attention)?

• Does it require human interaction as primary, AI as supplementary?

• Does it build capacity (teaching) or provide answers (telling)?

The vagueness creates cover for adoption without evaluation; schools can point to the “enriched” category while students experience the “diminished” reality.

Crucially, despite acknowledging existing harms, the report recommends no waiting period — no pause for more independent research to emerge, no requirement that vendors prove safety and efficacy before adoption. Schools are left to implement now and troubleshoot problems later, precisely the pattern that produced fifteen years of social media harms before institutional response. As I argued in “Education Against Disruption”, the Precautionary Principle demands proof of safety before deployment, not apologies after damage occurs.

The report suggests problems are fixable if we act now, but underestimates path dependency. Once teachers’ workflows rely on AI-generated lesson plans, grading, or assessment design — all listed as benefits — reversal becomes institutionally costly. This creates a parallel risk to the student doom loop: teacher cognitive offloading. Despite extensive concerns about atrophying student capacities, there is no mention of the same possibility in teachers. New research from MIT suggests cognitive costs aren’t limited to youth; adults who rely heavily on LLMs show similar patterns of reduced “neural, linguistic, and behavioural” capacity.

Share

The Developmental Window Problem

Beyond implementation challenges around design and institutional lock-in, a second problem emerges: developmental precision. While the report provides age-differentiated guidance for teaching about AI (through curricula like MIT’s “Day of AI”), it offers little in the way of a developmental framework for students using AI as a learning tool. The same recommendations apply whether discussing eight-year-olds building foundational literacy or seventeen-year-olds with established skills.

But the risks aren’t equally distributed across ages. Research from Toronto’s Hospital for Sick Children shows that early screen exposure predicts lower academic achievement through Grade 6, while UC San Francisco studies demonstrate that even moderate social media use among adolescents produces measurable cognitive costs. These findings reflect the brain’s heightened plasticity during critical developmental windows.

Brookings’ recommendations are limited to the following: “As students mature, they can progressively experiment with AI systems, building both technical familiarity and critical understanding through hands-on engagement.” But when do students mature enough? At age 8? 12? 16? And for which uses: writing assistance, math tutoring, research? Without age-specific thresholds, schools are left to guess during the precise years when cognitive foundations are being built.

Compounding these concerns about timing, a third problem emerges: equity assumptions.

The Equity Trap

Brookings identifies a concerning pattern:

“A critical finding emerging from the analysis of data to date is that a large and growing AI divide is widening and deepening what has long been an existing digital divide. This divide operates through a self-reinforcing cycle: as AI integration deepens within educational systems, already-existing barriers to accessing technology become increasingly entrenched.”

But the report’s equity framing contains a paradox. On one hand, it worries that underserved communities lack AI access, framing access as a problem to solve. On the other hand, Delphi panelists “expressed worry that the poorest education systems will employ AI to teach students while students in wealthy education systems will be taught by both AI and human teachers,” framing AI-mediated instruction as the inequity itself.

The report never reconciles this tension: Is the problem that poor students can’t access AI, or that they’ll be taught primarily by AI while wealthy students retain human teachers? By framing the issue as unequal access rather than questioning the appropriateness of AI-mediated learning, Brookings implicitly accepts a two-tier system; the question becomes only which students get algorithms and which get humans.

This obscures the deeper question: Is AI-mediated learning pedagogically sound for anyone during critical developmental periods? The equity solution isn’t “better AI for poor students”, it’s human teachers for all students.

All children deserve what research shows they need — face-to-face interaction, sustained attention practice, and relationships with adults who know and care for them. The equity question isn’t whether poor students can access AI; it’s whether any students should learn primarily through it during the years when human relationships form the foundation for cognitive and social development.

Share

Subscribe now

What the Report Gets Right, and What Schools Must Do

Despite the tensions in its equity framing, the report offers recommendations worth building upon, provided schools add the protective mechanisms Brookings leaves implicit.

Recommendations Worth Amplifying

On regulation and procurement, the report calls for “comprehensive regulatory frameworks” and urges schools to “procure technology that protects students’ privacy, safety, and security.” These recommendations recognize that market forces alone won’t produce safe tools. Schools need elevated procurement standards: privacy by default, data encryption, and safety requirements that shift the burden of proof to vendors.

On human agency, the report clearly states that “thinking and learning are not tasks to be outsourced to AI — they are how students build their identity, agency, and dreams.” This principle contradicts the AI-enriched learning optimism elsewhere in the report, but it’s the right starting point. If we agree that cognitive work shouldn’t be outsourced, then AI must be supplementary, not substitutive.

And finally, on research, Brookings correctly notes a dearth of evidence around AI in education, and calls for “real-time evidence development to help inform current AI practices before norms are entrenched.” But this recommendation undercuts the report’s fixability assumption; if we need more research before norms are entrenched, why not institute a waiting period now?

These protective measures require institutional commitment. Schools must operate as Walled Gardens — bounded environments with elevated standards designed to protect developmental conditions — rather than Marketplace Mirrors that reflexively adopt whatever vendors offer. Without this foundational shift in how schools understand their responsibility, even strong recommendations become aspirational rather than operational.

What Brookings Leaves Implicit

What’s missing? The report provides no mechanisms to prevent adoption before safety is proven, no accountability when tools fail, and no protection for educators who resist harmful implementations. Schools must add what Brookings leaves implicit:

1. Institute waiting periods: 12-24 months before adopting new AI tools, or as long as is required for substantial evidence to emerge; vendors must prove safety and pedagogical efficacy before adoption, not after harm emerges,

2. Establish sunset clauses: All AI pilots expire automatically; require evidence-based renewal.

3. Audit current use ruthlessly: Is AI genuinely pedagogical or administratively convenient? Is it being used during critical developmental windows for foundational skills?

4. Protect assessment integrity: If educators can’t verify that cognitive work occurred, the tool undermines education’s core function. Make process visible through drafts, annotations, and artifacts of attention.

5. Support teacher resistance: Normalize professional refusal of tools judged pedagogically harmful; protect educators facing administrative pressure to adopt tools that undermine learning.

Share

The Path Is Narrow

While there are transparent contradictions in the Brookings report, its overarching conclusion is correct: we’re using AI in education before we’re prepared to do so. Market pressures, administrative convenience, and technological optimism are driving adoption while children bear the developmental costs. There is a lack of technical understanding, a lack of thoughtful reflection about AI’s purposes and consequences, and a lack of informed, coherent, educational policy.

The premortem methodology demonstrates what institutional responsibility should look like: anticipating harms before they become entrenched. The evidence already supports protective action. We know that early screen exposure predicts lower achievement, that adolescent social media use produces cognitive costs, and that sustained attention predicts academic success. We know that developmental windows close, that cognitive capacities require exercise, and that what doesn’t develop during critical periods may not develop at all.

Schools must respond with pedagogy that emphasizes process, human relationships, and embodied experiences — a bulwark against the harms that have already emerged. Let’s make the patient choice rather than the reactive one. Let’s operate according to evidence rather than prediction. Let’s heed the premortem so we don’t need a postmortem.

Developmental windows don’t wait for more research. Neither should we.

If this research digest clarified the stakes around AI in education, please share it with educators, administrators, and parents in your network. These conversations matter, and the evidence deserves the widest possible audience.

The work of analyzing educational research and building frameworks for protective practice continues through reader support. If you find value in these digests, please consider subscribing.

Share

Subscribe now

Leave a comment