Attention in the classroom in the age of phones

Twenty-five minutes into second period. You're explaining something you've explained many times before and know well. Two students by the window have their notebooks open and are looking at them, but clearly not taking anything in. At the back, one student holds a phone under the desk so it's not obvious, and the student next to them is watching along. In the middle sits a girl listening with total focus, and you notice she's the only one left. You raise your voice. It works for twenty seconds. Then everything drifts back to where it was.

Most teachers read this situation as a discipline or motivation problem — and respond accordingly: a warning, a stricter rule, a confiscated phone. Sometimes it helps briefly, usually not. Because the real cause is somewhere else. At minute twenty-five, nobody is misbehaving. At minute twenty-five, attention has run out, which is a physiological state, and physiology doesn't respond well to being told off.

This chapter rests on one idea that looks unassuming but changes nearly everything about how a lesson gets planned: a class's attention isn't a trait students either have or don't — it's a shared, renewable, and exhaustible resource the teacher manages. A phone in a pocket, or a language model inside that phone, doesn't change any of this. It just makes it more visible and more costly.

Attention as a shared, exhaustible resource

Human attention is inherently narrow. At any given moment we can actively hold only a few elements, and any switch costs something — this mechanism is covered in the chapter on the science of productivity using adult work as the example, but it holds in a sharper form for twelve-year-olds. The capacity to sustain deliberate focus develops over a long time, and in any given class it's spread across a wide range: what's a comfortable twenty minutes for one student is a maximum of seven for another.

A second layer applies in a classroom that doesn't show up in individual work: attention is contagious in both directions. One student who disengages and starts doing something else doesn't just lower their own attention — they become a stimulus for the four people around them. Conversely, a class where everyone is intensely engaged holds the attention even of those who couldn't sustain it alone. That's why a lesson can't be managed as a sum of individuals. It has to be managed as an environment.

A third thing worth saying out loud: attention gets spent, but it also regenerates. It doesn't regenerate, though, by letting students "just be" for a while — an idle class picks up momentum, and pulling a class that's picked up momentum back into focus costs more energy than the break saved. It regenerates through change. A different type of activity, a different modality, movement, a brief social contact, something that uses a different part of the head. Not switching off, but switching over.

The practical consequence: planning a lesson around the material ("what I need to cover") is only half of prep. The other half is planning around attention — when the class will be freshest, so that's where the hardest content goes; when the slump will hit, and what will interrupt it. Teachers who do this have calmer lessons not because they have better classes.

Lesson rhythm: blocks, switches, movement

A lesson that works isn't an even stream. It has shape. The simplest usable model is a chain of blocks of ten to fifteen minutes, where the activity changes between blocks, not necessarily the topic.

A block of instruction, a block of independent work, a block in pairs, a block of shared review. Or: a short shared opening of a problem, an individual attempt, comparing with a partner, instruction that finally makes sense now, application. Within such a chain, attention can be sustained noticeably longer than in one continuous stream, because every switch is a small restart.

The key is that the switch has to be clear. When you drift smoothly from instruction into an independent-work assignment, half the class won't notice and will keep doing something else. Clear signals work: a change of position in the room, a short sentence like "five minutes on your own now, then we'll compare," writing the assignment on the board, a timer. A timer visible to the whole class is one of the cheapest tools you can add to a lesson — it bounds the time, makes the pace public, and takes you out of the role of the person constantly nagging.

Movement belongs in the rhythm too. Not as a reward or distraction, but as a tool: standing up to answer, moving to a classmate, walking to the board, picking up a material from a table. A short physical change does more for attention than repeated reminders. The chapter on the body and productivity describes the same mechanism in adults — in children, who are supposed to sit at a desk for forty-five minutes, the effect is even bigger.

One last note on rhythm: the hardest thing doesn't belong at the end of the lesson. It's the most common planning mistake, and it usually happens because prep proceeds by the logic of the material (basics first, then the hard stuff), not the logic of attention. When new, difficult material lands at the end, it meets the class at its worst possible state, and the teacher then reads that as the students being incapable.

Opening and closing rituals

A ritual is a repeated sequence that doesn't require deciding anything. That's exactly why it works so well: it saves attention for everyone, you and the class alike. A lesson that starts differently every time burns the first three minutes just getting everyone oriented. A lesson that starts the same way every time starts instantly.

The opening needs to accomplish three things: switch the class into subject mode, connect to what came before, and say what today is about. A short sequence handles it — two questions about the previous material, one sentence about today's goal written on the board. It helps if the ritual has a fixed form and variable content: always "three questions to start," always the same spot on the board. Within a month, students already know what's expected and switch on their own.

The ending is even more underrated. A lesson that ends mid-sentence with the bell never closes in students' heads, and nothing about it settles. Yet two minutes is enough: what we found out today, in the student's own words (not yours), one question that stays open, and physically tidying the space. Closing is also the cheapest form of review you have — and that review happens right when the memory trace pays off most. The connection to spaced repetition is developed in the chapter on productivity while studying.

Rituals have one more function that only shows up in a difficult class: they give the lesson a predictable skeleton to lean on when something goes wrong. If you know a familiar transition is coming in five minutes, you don't have to force your way through a situation that's spiraling — you just need to steer it to the next fixed point.

Phones in class: what works and what doesn't

The debate about phones in schools is usually loud and runs to two extremes: either an outright ban or resignation. It's more useful to ask differently: what problem are we actually trying to prevent? Because a phone isn't one problem, it's several different ones, and each needs a different answer.

A phone as a source of interruption (notifications, vibration, a glance at the screen) hurts attention even when the student isn't actively using it. Just the presence of a device you know might have something on it drains some capacity — this effect is well documented in adults, and there's no reason to expect anything milder in teenagers. A phone as a social tool is different: relationships, conflicts, and hierarchy get worked out on it, and taking it away is a social loss for a teenager, not a technical one. A phone as a work tool is a legitimate thing too: camera, dictionary, calculator, access to shared materials. And a phone as a path to answers is the subject of the next section.

That's why blanket bans often fail. A ban that can't be enforced mostly teaches students that rules are for show. A ban enforced by confiscation creates a conflict that costs more attention than the original disruption. And a ban with no justification runs into a teenager's natural resistance to authority without a reason.

What works better in practice is a rule that's visible, consistent, and justified. Visible: the phone is in a designated spot (a bag, a pocket organizer, face down at the edge of the desk) — not "hidden," because a hidden phone is an invitation to a game of hide-and-seek. Consistent: the same rule across the whole school, or at least across subjects, because otherwise every teacher's individual regime means the whole thing gets renegotiated every single lesson. Justified: not "because it's not allowed," but "because this activity needs twenty uninterrupted minutes, and nobody has that with a phone on the desk."

It's also worth explicitly naming when phones are allowed. When the rule includes the sentence "for this part, we'll get our phones out," the rule stops being repression and becomes an organization of work. And it's worth mentioning that students can manage their own attention too — setting a phone to grayscale or using app limits aren't punishments, they're tools adults set for themselves voluntarily. Practical how-tos are in the tips Phone grayscale mode and App limits; a broader frame is offered in the chapter on digital minimalism.

AI in your pocket: a reality that can't be banned

The last two years have added something qualitatively different to this equation. A phone gave students access to information. A language model gives them a finished output — an essay, a calculation with steps shown, a book summary, a translation, a word problem solved and explained. And it gives it to them instantly, free, and in a form that can't be reliably detected. Tools promising AI-text detection aren't reliable enough to base a grade or an accusation on, and a falsely flagged student is a serious problem.

That means one uncomfortable but liberating thing: homework whose output is text has stopped being evidence of a student's own skill. That's not a catastrophe. It's information that the assignment has to change — not that assignments have to stop.

Process instead of product

When you grade only the final product, you're grading what the model can do. When you grade the path, you're grading the student. In practice that means requiring intermediate steps: an outline submitted before the text, sources with notes on what the student found in them, a draft with their own crossed-out revisions, a version after incorporating feedback. This structure, incidentally, matches what quality work looks like even without AI — and as the chapter on grading without burnout shows, it shifts feedback to where the student can still act on it.

Defense

The simplest and most effective tool: talk about the submitted work for two minutes. Why did you choose this argument? Where did you get this figure? What would you change if the length doubled? A student who genuinely did the work answers without trouble. A student who had it generated gets stuck — with no accusation, detector, or conflict needed. A defense isn't a punishment, either — it's a skill they'll need for life.

In-class work

Where you genuinely need to know what a student can do on their own, the work belongs in class. Not because it's about control, but because you can see the process: where they get stuck, what they try, how long they persist. It's also the only situation where you can step in at the moment it actually matters. Shifting the center of gravity of creative work from home to the classroom is the most significant change AI has caused in schools, and schools that have made this shift agree it's benefited them for other reasons too.

And explicit rules

The last piece is naming what's allowed. "No AI" is unenforceable and dishonest. A graduated approach is more usable: for this assignment, don't use AI at all; for this one, you may use it for brainstorming and spellchecking; for this one, use it fully and submit the conversation too, noting what you accepted and what you rejected. The last variant is, incidentally, an excellent assignment in its own right — it forces the student to judge the quality of the output, which is exactly the skill they need. The tip AI in the classroom: how to teach its use walks through this step by step; for your own prep, the tip Lesson prep with AI fits well.

One principle applies to school just as it does to business: AI proposes, the human approves. For students and teachers alike. And tools without appropriate contractual data protection are not where students' personal data, health information, or family circumstances belong — that needs to be factored into every tool you bring into your teaching.

Media literacy as part of the subject, not a separate day

There's a temptation to solve all of this with a single project day on media literacy. It works about as well as a one-off lecture on healthy eating. The ability to assess where information comes from, who it serves, and how reliable it is isn't built through awareness campaigns, but through repeated practice inside a subject, where it has concrete content.

In history that means asking, at the source, who wrote it and why. In science, distinguishing a claim, a measurement, and an interpretation. In language arts, recognizing what devices a text uses to affect the reader. In math, showing how a graph with a truncated axis tells a different story than the same data would otherwise. None of this is an extra lesson — it's a way of teaching what you'd have to teach anyway.

You can do this directly with language model outputs. Have the model answer a question from the material you're covering, and have the class find what's correct, what's oversimplified, and what's flat-out wrong in the answer. It's a lesson where students learn the material and critical judgment at once, and it has one rare quality: it puts the student in the position of judge, not recipient. Checking a claim against a source is a topic outside school too — the process is described in the tip Fact-checking with AI.

The teacher as a model of behavior toward technology

The last and most uncomfortable part. Students learn their relationship with technology far more from what they see than from what they hear. A teacher who scrolls their phone during independent work has just demonstrated how it's done. A teacher who takes a call in the middle of a lesson has communicated that interruption is normal. And a teacher who keeps their phone face-down on the desk and picks it up only when they have a reason and say so out loud teaches self-control far more convincingly than any rule in the handbook.

The same goes for AI. If you use models for prep, for building tests, or for drafting messages to parents, say so. Not only is it honest — it's also the best possible demonstration of sensible use: you show what you asked for, what the model proposed, what you kept, and what you discarded and why. A student who sees a teacher working with AI like an assistant, not like an answer engine, takes away more than they would from a lesson about ethics.

And then there's one thing you model most powerfully without meaning to: what a focused person looks like. A calm pace, one thing at a time, a finished thought, silence that doesn't need filling. In a world where most students have few chances to see an adult in deep focus, the teacher is often the only such model around. That's exactly why it's worth protecting your own attention too — a lesson taught by someone frazzled and juggling two unfinished tasks looks different from a lesson taught by someone who arrived with their head in the room. How to build that kind of space is described in the chapter on the psychology of productivity and in the previous chapter, grading without burnout.

Key takeaways

• Attention is a resource, not a trait. In a classroom it's shared besides: one disengaged student pulls down four around them; conversely, an engaged class holds even those who couldn't sustain focus alone.
• Plan the lesson around attention, not just the material. Blocks of ten to fifteen minutes, clear switches, movement — and the hardest content in the first half, never at the end.
• Opening and closing rituals save attention for everyone. Fixed form, variable content. Two minutes to close a lesson is the cheapest review you have.
• With phones, be clear about which problem you're preventing. A visible, consistent, justified rule works — including stating when phones are allowed. An unenforceable ban mostly teaches that rules are for show.
• AI in students' pockets can't be banned, but it can be planned around. Grade the process instead of the result, add a two-minute defense, and do the substantive work in class. AI-text detectors aren't reliable enough to base a grade on.
• Name when AI is allowed and to what extent. An assignment like "use it and submit what you accepted and what you rejected" teaches judgment — the skill students need most.
• Media literacy belongs inside subjects, not on a project day. And you're a model: a focused adult with their phone face-down teaches more than a rule in the school handbook.