In 1975, a high school math teacher in Ohio wrote to a national education journal warning that the handheld calculator would "destroy a generation's ability to think." Students would never learn arithmetic. Mental math would die. Standards would collapse.
He wasn't alone. Across the US and Europe, schools banned calculators from classrooms and exams through the late 1970s and into the 1980s. The logic felt airtight: if the machine does the work, the student doesn't learn.
We now know how that story ended.
Calculators didn't destroy mathematical thinking. They relocated it. Students stopped spending hours on long division by hand and started spending that time on what the calculator couldn't do — setting up the problem, interpreting the result, deciding whether the answer even made sense. Math education didn't get easier. It moved up the ladder of abstraction. By the 1990s, the SAT and most national curricula didn't just allow calculators — they assumed them.
The teachers who banned calculators weren't protecting rigor. They were protecting a definition of rigor that the world had already moved past.
We are standing in exactly the same place with AI.
The ban doesn't work — and it never did
Right now, universities across Europe and beyond are spending enormous energy trying to keep large language models out of student work. Detection software. Honor-code rewrites. Oral re-examinations. Blue-book exams making a comeback.
Here's the uncomfortable data: AI detection tools are unreliable, with documented false-positive rates that disproportionately flag non-native English writers — which, for international universities, means flagging exactly the students we should be protecting. Meanwhile, students use the tools anyway. Surveys across higher education consistently show majority adoption, regardless of policy.
So the ban produces the worst of both worlds: it doesn't stop usage, it just pushes it underground, where no one is teaching students to do it well.
That last part is the real cost.
The skill is not "avoiding AI." The skill is "using it well."
When we banned calculators, the students who suffered most weren't the ones who used them secretly. They were the ones who graduated never having learned to use them fluently — who hit university or the workforce and had to catch up on a tool everyone else already commanded.
AI fluency is now that tool.
A graduate entering any knowledge profession in 2026 will be expected to know how to prompt effectively, how to verify and challenge a model's output, how to spot where it's confidently wrong, how to use it to draft and then think past the draft. These are not lazy skills. They are demanding ones. They require judgment, domain knowledge, and metacognition — the exact capacities a university exists to build.
A student who outsources their thinking to ChatGPT and submits the result unread has not learned to use AI. They've learned to be replaced by it. Our job is to teach the difference.
What "teaching with AI" actually means
This is not a call to let students paste prompts and call it an education. It's the opposite. It's harder than banning, and it asks more of us as educators.
It means redesigning assignments so the thinking is visible — asking students to submit their prompts, critique the AI's answer, document where they overrode it. It means assessment that rewards judgment over production, because production is now cheap and judgment is not. It means teaching students that the interesting work begins after the first draft appears.
The calculator didn't end math homework. It changed what good math homework asks. AI is asking the same of us now.
The choice in front of us
Every generation of educators faces one of these moments — a technology that threatens an old definition of competence and demands a new one. The printing press. The calculator. The internet. Each time, the instinct was to ban, then to fear, then — eventually, always — to integrate and teach.
We can spend the next five years policing a tool students already hold in their pockets. Or we can do the harder, more honest thing: teach them to use it with skill, with skepticism, and with the judgment that no model has.
The students don't need us to protect them from AI.
They need us to teach them how to be more effective with it — not less.
That's the assignment now.
Found this useful? Forward it to one colleague who's still on the fence. And hit reply — I'd love to hear how your department is handling this.