Every few weeks, a smart person publishes a version of the same essay. Schools banned the pocket calculator, the argument goes. Teachers worried students would never learn arithmetic. Then we came to our senses, let the machines in, and math education got better. Generative AI is the same story one more time. The panic is generational. The tools will win. Get out of the way.
The analogy is doing an enormous amount of work in the debate over AI in classrooms, and almost none of it is honest. The calculator and the large language model are not the same kind of tool. They differ in the one thing that matters for learning: what the student still has to do after the tool does its part. Until that distinction is taken seriously, the calculator analogy is not an argument for embracing AI. It is a way of not thinking about it.
What the Calculator Actually Did
The pocket calculator arrived in classrooms carrying real anxiety. The National Council of Teachers of Mathematics recommended that calculators be integrated into mathematics instruction at every grade level as far back as the mid-1970s, and by 1986 the organization had formalized the position that "calculators do not threaten the development of basic mathematical skills" (NCTM, 1986). The College Board permitted graphing calculators on the AP Calculus exam beginning with the 1994–95 school year. None of this was reckless. It was a slow, evidence-led accommodation.
The reason it worked is the reason the analogy fails. A calculator executes arithmetic. It does the long division so the student does not have to. But the student still has to know which operation to perform, in what order, on what quantities, and whether the result is plausible. The calculator automates a procedure the student has either already mastered or is actively mastering. It removes the tedium. It does not remove the judgment, because the judgment lives in setting up the problem, not in executing the final computation.
This is the textbook case for cognitive offloading. Risko and Gilbert, surveying the literature on the subject, argue that offloading part of a task to an external tool reliably improves performance on the parts of the task we keep, precisely because attention is freed for them (Risko & Gilbert, 2016). The calculator earns its place in the classroom because the part it takes, computation, is not the part the lesson is trying to teach. The part the lesson is trying to teach, mathematical reasoning, stays with the student.
What AI Takes Instead
Here is where the analogy breaks. When a student hands a writing prompt to a language model and pastes back the result, the tool is not taking the tedious part of the assignment. It is taking the assignment. The reasoning, the selection of evidence, the sequencing of arguments, the voice, the act of deciding what to say next, all of it moves out of the student's head and into the model. What remains for the student is not a higher-order task freed by offloaded tedium. It is paste.
The defender of the analogy will say this is just the calculator in an earlier, scarier costume. Eventually we will teach students to use AI the way we taught them to use calculators, as a tool for the higher-order work. But this assumes the higher-order work is sitting somewhere else, waiting for the student to pick it up once the machine handles the low part. In writing, the low part and the high part are the same part. There is no way to outsource sentence-level drafting without also outsourcing the thinking that produces it, because for most writers the thinking happens in the course of producing the sentences. Sparrow, Liu, and Wegner's finding that people remember where to find information rather than the information itself anticipated exactly this drift toward transactive memory (Sparrow et al., 2011). When the model holds the first draft, it also holds the first thoughts.
This is why the calculator analogy lands as reassuring rather than rigorous. It asks us to believe that AI, like the calculator, will free students for deeper work. But the deeper work in a writing assignment is the work the AI just did. There is nothing left to be freed for.
Verification Is Not Symmetric
There is a second difference, and it is the one that should settle the argument on its own. A calculator's output is deterministic and checkable. Seven times nine is sixty-three on the machine and in the head, and a student who has learned the times tables can tell, in under a second, whether the calculator's answer is sane. The verification is cheap because the student retains the skill being outsourced.
A language model's output is none of those things. It is probabilistic. It is fluent in direct proportion to how wrong it is, which is the most dangerous property a teaching tool can have. A student cannot glance at a generated paragraph and know whether the reasoning holds, because the very capacity that would let them judge it, the capacity to construct the argument in the first place, is the capacity they never built. The calculator offloaded a skill the student had. AI asks the student to supervise a skill the student has never acquired. These are not the same relationship to a tool, and a classroom that treats them as the same will produce students who cannot tell the difference either.
Why the Analogy Persists Anyway
The analogy survives because it is useful, not because it is true. It reframes a genuinely hard question, what is the point of a writing assignment in a world where machines write, as an easy one, have we had this panic before. It borrows the moral authority of a debate we already won. We were right about calculators. Therefore we will be right about this. The conclusion is smuggled past the premise.
It also flatters the person making it. To invoke the calculator is to position yourself on the enlightened side of history, the side that came around last time. It costs nothing. The people who resisted calculators look foolish in retrospect, and the analogy asks us to identify with the people who did not. This is a story about ourselves, not an argument about the tool.
The honest version of the calculator position is narrower than its defenders admit. The calculator succeeded because it took a discrete, checkable, already-learned procedure off the student's hands and left the actual mathematics intact. The question for AI is whether anything comparable is possible, and the answer is not obviously yes. A writing assignment in which the AI produces the draft and the student "edits" it is not the same assignment, because editing text you did not generate exercises a different and lesser faculty than generating it. That may still be worth teaching. But it is a different thing, and we should say so instead of pretending the assignment survived intact.
So What
None of this is an argument for banning AI in schools. The calculator story does not end with the machine forbidden forever, and it should not. It ends with the profession deciding what the tool is for, what it cannot replace, and how to teach the part that remains the student's job. That decision took the better part of two decades for arithmetic, a procedure far simpler than reasoning.
The reason to reject the calculator analogy is not to reach the opposite conclusion. It is to reach any conclusion at all. As long as we are telling ourselves that this is just the same movie, we are excused from the work of figuring out which parts of an assignment the tool may take and which parts, when taken, empty the assignment of its purpose. The calculator analogy is not an argument. It is a permission slip to stop thinking. The classrooms that take AI seriously will be the ones that stop using it.
Sources
- National Council of Teachers of Mathematics. (1986). The Use of Calculators in the Mathematics Classroom. NCTM position statement. https://www.nctm.org/Standards-and-Positions/Position-Statements/The-Use-of-Calculators-in-the-Mathematics-Classroom/
- Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688. https://doi.org/10.1016/j.tics.2016.07.002
- Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 333(6043), 776–778. https://doi.org/10.1126/science.1207765
- The College Board. (1993). Advanced Placement Calculus: Graphing calculator policy. See AP Calculus exam materials and calculator policies. https://apstudents.collegeboard.org/ap/courses-ap-calculus-ab