Chapter 13

The Home

The one room that is not a workplace — and the curriculum nobody thinks they are teaching.

“Children have never been very good at listening to their elders, but they have never failed to imitate them.”

— James Baldwin

Every decision in this book so far has been a professional one. This one is not, and it may be the one that matters longest — because the most consequential lesson about artificial intelligence in any household is not being delivered in any conversation about artificial intelligence. It is being delivered by what the adults reach for, dozens of times a day, in full view of people who are still deciding what thinking is for.

What children take from watching

In 2017, three researchers at MIT published an experiment in Science with infants barely past their first birthday. Babies who watched an adult visibly struggle at a task before succeeding then tried harder and persisted longer at their own difficult task. Babies who watched an adult succeed effortlessly gave up sooner. The effect was measurable in children too young to speak.

Persistence, in other words, is transmitted by demonstration long before it can be taught by instruction. Children calibrate how long to stay with a hard thing substantially by watching how long the adults around them stay with hard things.

Now place a capable AI assistant into that picture. An adult who reaches for the tool at the first flicker of not-knowing is not modelling laziness — and this is exactly what makes it difficult to see. It looks like competence. It is competence, in the narrow sense that the problem gets solved. But what a watching child absorbs, across thousands of unremarkable repetitions, is a rule about what to do when something is hard: resolve it externally, immediately. No single instance matters. The accumulated pattern becomes their default orientation toward their own mind.

Every time an adult takes the shortcut, they are not only outsourcing the task. They are teaching a watching child what thinking is for.

It is worth being precise about what this does not mean, because the obvious conclusion is the wrong one. It does not mean hiding these tools from children, which would disadvantage them in the world they are actually going to work in. It does not mean performing struggle, which children detect instantly and discount entirely. It means practising the sequence this book has argued for throughout — thinking first, reaching second — visibly, and occasionally working through something the slow way because the working-through is the point.

The homework question

The practical question most households are actually arguing about — whether children should be allowed to use AI for schoolwork — is largely settled by the world they are growing into. A child raised without fluency in these tools will be disadvantaged in the way a child raised without literacy would be. The argument is over.

The real question is not access. It is sequence — and sequence gives families a rule far more useful than any ban.

That third step is the one that carries the weight. A field experiment with nearly a thousand high-school students found that those who practised mathematics with an unrestricted AI tutor performed dramatically better while they had it — and scored below students who had never used it at all once it was withdrawn. The version engineered to give hints rather than answers largely avoided the harm. The distinction between those two tools is, in a household, entirely a matter of how the thing is used. A ten-year-old who has practised the compare-out-loud habit is holding the hint version, whatever software is on the screen.

Write the rules together

Households that handle this well tend to have something households that struggle do not: an explicit, short, jointly written agreement. Five lines on the refrigerator, drafted by everyone including the children, beats any policy imposed from above — because a rule a child helped write becomes part of their identity, and a rule handed down becomes something to route around.

The content matters less than the authorship. We try first. We use it to learn, not to skip. We say when we used it. Some things we do the slow way on purpose. Screens are parked during family time. The adults are bound by all five, visibly, and are the ones most likely to break them.

The half of home life worth automating

There is a second half to all of this, and here the argument runs the other way entirely: hand it over.

Meal planning, scheduling, the endless logistical coordination of a household — delegate all of it without guilt. Nothing developmental is happening in the negotiation of a grocery list. This is exactly the efficiency this book has argued for in every chapter, and the hours it returns are the budget everything else is spent from.

The risk is subtler, and it is the domestic version of the whole thesis: letting the efficiency migrate from the logistical layer into the relational one. Connection between people is built almost entirely in unoptimised margins — the aimless afternoon, the unplanned conversation, the boredom that turns into something someone has been carrying around for days. A generation ago the sheer labour of running a household imposed that shared time by default. As the tools remove the labour, the togetherness has to be created deliberately, because the moments that matter arrive on their own schedule and only land in unhurried time.

A household can be impressively managed and relationally thin, and the difference does not appear on any calendar. So take some of the hours the tools return and spend them, deliberately and inefficiently, on the people they were always for. One protected block a week, unstructured, devices parked, nothing planned. Defend it the way the deep-work block is defended — because that is precisely what it is.

The tools should buy a household time. The question is whether anyone decided in advance what the time was for.

Why this chapter is here

The professional argument in this book operates on a horizon of years. This one operates on a horizon of a generation, and it runs in the opposite direction: the depth a child develops is not primarily a product of what they are told about technology. It is a product of what they watched the adults around them do when something was difficult.

Which is a heavier responsibility than the professional version, and also a simpler one. There is no protocol to install and no assessment to take. There is only the question of whether, in a house where someone is watching, the adults still sometimes choose the harder path on purpose — and whether anyone said out loud why.