Build It Before You Hand It Over

The calculator took over a skill children already had. AI can take over one that never got the chance to form. The question is not whether, it is before or after.

In 1980 the National Council of Teachers of Mathematics published "An Agenda for Action," eight recommendations for American school mathematics in the decade ahead. The third one said that mathematics programs should take full advantage of the power of calculators and computers at all grade levels.

The country looked at a machine that could do a child's arithmetic, decided on purpose to hand it to children, and arithmetic did not die. Children still learn math, and the calculator let them go further with it. If that is how the calculator went, the argument runs, AI is simply the next one.

The calculator came after the counting

Look at when it arrived in a child's life. First a child counted on fingers. Then they learned to group things, to put numbers in order, to carry and borrow by hand. Only after all of that did the calculator show up, and it took over a job the child already knew how to do. It also gave one answer. If that answer was wrong, a child who knew the math could tell.

Issue 5 of On Judgment named the behavior underneath this: cognitive offloading, handing a mental task to an external aid instead of doing it in your head. Writing a number down is offloading. So is letting a calculator do the arithmetic. It is often useful. What that issue left open is when the offloading costs something, and the calculator is the clue. It cost little because it came second.

AI arrives differently. A child talks to it in ordinary language. It can be asked to do a whole assignment from start to finish. It can arrive before the skill underneath has ever been built. And much of what it produces, an essay, a summary, an argument, has no single right answer to check it against. The calculator replaced a skill that was already there. AI can replace one that never got the chance to form.

Issue 9 argued that an answer is not understanding. This is the same point seen from earlier on. The understanding has to exist before a tool can safely take over part of the job.

I lost a map I never noticed I had

This kind of loss has been measured, though in driving rather than school. On April 14, 2020, Louisa Dahmani and Véronique Bohbot published a study of 50 regular drivers and how much of their lives they had spent following GPS. They found that "people with greater lifetime GPS experience have worse spatial memory during self-guided navigation." When they retested a small group about three years later, the people who had leaned on GPS more in the meantime had lost more.

I recognized myself in that immediately. For years my work had me driving constantly, and I let GPS route every trip. Somewhere along the way I noticed I could drive to a place several times and still not know the way. I cannot prove the GPS did that. But the more I leaned on it, the less I could find my way on my own. I never decided to give up my sense of direction. I think I gave it up one turn at a time, by doing what the voice said in the moment instead of building the map myself.

Nobody warned me this would happen, and I probably would not have listened even if they had. The trips all went fine. That is what makes this kind of loss hard to see. Nothing goes wrong on any given day. The cost shows up on the day the tool is not available and you find out what you can no longer do on your own.

That is erosion in an adult who had the skill and stopped using it. The worry with a child is one step earlier, and quieter, because there is not a before to compare against.

Not whether. Before or after.

Our first principle is Think Better: "Use AI to sharpen reasoning and ideas, not to replace effort or judgment." A tool that sharpens is a tool used after the skill exists. A tool that replaces is a tool used instead of building it. From the outside the two are identical. The essay is turned in either way.

So, the useful question for a parent or a teacher is not "did you use AI on this." It is "could you have done this without it."

What is in your hands

Ask for the first one by hand. For any new kind of work, a first lab report, a first argument essay, a first proof, ask that the first one be done without the tool. Not forever. Once. The first time through is when the map gets drawn.

Ask whether the answer can be checked. If the work has one right answer your child can confirm, the tool is behaving like a calculator. If it does not, your child is the only check there is, and that only works if they could have done the work themselves.

Close the tool and ask for one piece again. Not a full recitation, a small rebuild. Pick one step, one paragraph, one calculation, and ask them to produce it from scratch with the screen off. A child who built the skill can rebuild a piece of it. A child who was handed the result has nothing to rebuild from.

For teachers, the classroom version is sequence. Teach the skill, let students practice it unassisted, and introduce the tool after. The calculator made sense to most of us because we already knew what it was doing.

The tool can wait

A skill that never forms does not announce itself. It is simply not there on the day it is needed, the way my sense of direction was not there the first time the phone died.

So do not ask your student whether they used it. Ask what they could have done without it, and make sure the answer is not "nothing."

What the checking turned up

Five things about the evidence behind this piece, because you should be able to go look.

The NCTM recommendation above is quoted from the catalog record for the document, which summarizes all eight recommendations in a paragraph. I did not read the 30-page document itself. The wording is the record's summary of the third recommendation, not necessarily the Council's own sentence.

The GPS study's follow-up is small. It retested 13 of the original 50 drivers, and the authors say plainly that "the longitudinal sample was small." The main finding, from all 50, is a correlation. On its own it does not show that GPS caused anything, which is why the follow-up matters and why its size matters too. The authors read the follow-up as evidence that GPS use caused the decline, while cautioning against strong conclusions because spurious correlations are possible.

The study also asked whether people who use GPS more simply have a poor sense of direction to begin with. It reports that the heavier users did not rely on GPS for that reason. That is what the drivers said about themselves on a questionnaire, not a separate measurement.

The study is of young adults, ages 19 to 35, primarily undergraduate students in Montreal who drove at least four days a week. It says nothing about children, and it measures a skill eroding in people who had it. Applying it to a child who has not built the skill yet is this piece's argument, not the study's finding.

The account of how the calculator entered a child's life is a general description of the sequence, not a sourced finding about any particular curriculum.

Dahmani and Bohbot, Habitual use of GPS negatively impacts spatial memory during self-guided navigation, Scientific Reports, 2020: www.nature.com/articles/s41598-020-62877-0

National Council of Teachers of Mathematics, An Agenda for Action: Recommendations for School Mathematics of the 1980s, 1980: eric.ed.gov/?id=ED186265

When we get something wrong in a published issue, it will be listed here too.