STOP TEACHING TO THE TEST

Why AI Should Finally Finish the Argument the Calculator Started

Somewhere in the mid-1970s, a math teacher stood in front of a classroom and banned the calculator. “You will not always have one of these with you,” she said, tapping the little plastic box like it was contraband. She was wrong. Not eventually wrong — immediately, structurally, permanently wrong. Forty years later, everyone in that classroom carries a calculator, a spreadsheet, a search engine, and a research assistant in their pocket at all times. It is called a phone. We didn’t just get the calculator back. We got Wolfram Alpha, Excel, and now a large language model that can explain the calculus behind the answer while it’s at it.

I open with this because we are about to make the exact same mistake again, at a much larger scale, and with much higher stakes. Except this time it isn’t a calculator. It’s the sum total of recallable human knowledge, delivered instantly, in natural language, for free. And most of our education systems are still optimized to reward the one skill that artifact has just made close to obsolete: memorizing facts and reciting them back on command.

Let me say the quiet part loudly: teaching to the test was never actually about learning. It was about producing a measurable, defensible number for an accountability system. Recall is easy to grade. Reasoning is not. So we built a twelve-year pipeline optimized for the thing that’s easy to grade — and we are now sending its graduates into a world where the thing that’s easy to grade is the thing a machine does better than any human alive.

The Critical Basics

Before we talk about fixing anything, two questions have to be answered honestly, because most of the debate about “AI in education” skips straight past them.

Question 1: What is school actually training people to do?

Not what the mission statement says. What the grading rubric actually rewards. In most systems, it rewards accurate retrieval under time pressure — the closed-book exam, the multiple-choice test, the fact recited on cue. That is a real skill. It is also, as of roughly 2022, a commodity skill. A fifteen-dollar-a-month subscription now retrieves facts faster and more comprehensively than the best-prepared student in the building.

Question 2: What does AI actually replace, and what does it leave exposed?

AI replaces retrieval. It does not replace judgment. It will hand you an answer, a summary, a first draft, a plausible-sounding citation — and it will do so with total confidence whether it is right or catastrophically wrong. The skill AI cannot substitute for is the one required to catch it when it’s wrong: the ability to interrogate a claim, trace its logic, spot the gap, and ask the next question. That is not a technology skill. That is philosophy. That is the Socratic Method. That is critical thinking, and it has been sitting in the corner of the curriculum, underfunded and untested, for decades.

J.Paris: In forty-odd years of walking factory floors and sitting in boardrooms, the people who move fastest under pressure are never the ones with the most memorized procedures. They’re the ones who can look at a system falling apart in real time, ask the right diagnostic question, and reason their way to root cause before the postmortem even starts. That is not a talent. It is a trained capability — and it is trainable in a classroom exactly the way an operational excellence program trains it on a shop floor.

Before You Defend the Old Model — A Caveat I Won’t Let You Skip

Here is where I depart from the easy version of this argument, the one making the rounds on social media that says “facts don’t matter anymore, just teach kids to think.” That version is seductive and it is wrong, and if you build a curriculum on it, it will fail.

You cannot think critically about a subject you know nothing about. Critical thinking is not a content-free muscle you can flex in a vacuum — it is a set of operations performed on knowledge you actually hold in your head: compare, contrast, infer, question, synthesize. A student with zero internalized history cannot apply the Socratic Method to history; there is nothing there to interrogate. A student with no internalized number sense cannot catch the AI when it quietly drops a decimal point. The calculator didn’t eliminate the need to know that 847 times 3 lands somewhere around 2,500 — it eliminated the need to grind out the long multiplication by hand. AI raises the stakes on background knowledge. It does not retire it.

So the real target is not memorization itself. It’s memorization as the terminal goal — the finish line — instead of memorization as a byproduct of genuine engagement with a subject you’re being taught to question, not just absorb.

The Rebuild: Four Steps

This is not a call to burn down the curriculum. It’s a call to reorder its priorities and rebuild the assessment model around them. Here is the sequence.

Step 1: Stop grading what the machine already does for free

Any assessment a student can fully complete by pasting the question into a chatbot is not testing the right thing. That doesn’t mean ban the tool — you couldn’t enforce that ban even if you wanted to, any more than the calculator ban held in 1985. It means redesign the assessment so the AI’s output is the starting material, not the finished product: give students the AI-generated answer and grade them on finding what’s wrong with it.

J.Paris: We do a version of this in engagements constantly. I don’t ask a client’s team to hand me a report and trust it. I ask them to defend it — walk me through the logic, show me where the data could be lying, tell me what would change their conclusion. The report is cheap. The defense is where the actual capability shows up. Grade the defense, not the report.

Step 2: Put the Socratic Method back at the center, not the margins

Most classrooms use questions to check whether a fact was retained. The Socratic Method uses questions to expose the limits of what you think you know — a completely different exercise. It is uncomfortable, it is slower, and it is exactly the discipline that separates a person who can use AI well from a person AI quietly makes worse. It is also, not coincidentally, the exact discipline behind writing a good prompt: knowing enough to ask a sharp, specific, well-aimed question instead of a vague one. Bring it back as a structured practice, not an occasional technique: every unit should include a session where the goal is not the right answer but the better question.

Step 3: Teach logic and argument structure as an explicit subject, not a hidden one

Formal and informal logic — how to spot a false dichotomy, a hasty generalization, a circular argument, a non sequitur — used to be assumed knowledge. It is not assumed anymore, and it shows. This is the single most practical, portable, AI-proof skill you can put in front of a sixteen-year-old, because it is the exact skill required to evaluate an AI-generated argument for holes. Teach it directly. Test it directly.

Step 4: Redesign assessment around reasoning shown, not answers produced

Grade the process, not just the product. Oral defenses, live problem-solving, error-hunting exercises, open-book exams designed around synthesis instead of recall — all of these measure something a language model cannot do on a student’s behalf: think, in real time, out loud, under scrutiny. This is harder to grade than a scantron sheet. It is also the only grading model that still means something.

J.Paris: Readiness — the ability to respond effectively to what you did not specifically prepare for — is the core of everything I’ve built a career around. An education system that only tests what a student memorized in advance is optimizing for the opposite of readiness. It is optimizing for a world that no longer exists.

What I Would Push Back On — From the Other Side

In fairness to the teachers still defending memorization, they are not entirely wrong either, and an honest argument has to say so. Foundational knowledge — times tables, historical timelines, basic scientific facts, vocabulary — builds the working memory scaffolding that makes higher-order thinking possible at all. A student who has to look up every fact has no bandwidth left to reason about the relationships between them; cognitive load theory backs this up plainly. The fix is not to strip foundational knowledge out of the curriculum. It’s to stop treating the recall of that knowledge as the finish line, and start treating it as the entry fee for the harder, more valuable work that comes after.

And here is the argument that should end the debate for good, because it comes from inside the AI itself: a large language model is only as good as the question you put to it. Prompting well is not a technical skill — it is a knowledge skill. You cannot ask an intelligent question about a subject you know nothing about, because you don’t yet know what you don’t know, what’s actually in dispute, or what a good answer would even look like. The student with no foundation doesn’t get better answers out of AI. They get a fluent-sounding response to a shallow question, and no way to tell the difference between insight and nonsense. Facts aren’t the enemy of good prompting. They’re the prerequisite for it.

Final Takeaways

• The calculator argument was won forty years ago. AI is that argument again, at the scale of all recallable knowledge — don’t lose it twice by defending memorization as a terminal goal.
• Background knowledge is not obsolete. It is the raw material critical thinking operates on. Cut it and you don’t get better thinkers, you get faster guessers.
• Redesign assessment around defense, not production: can the student explain why the answer is right, catch it when it’s wrong, and ask the next question?
• Put the Socratic Method and formal logic back into the core curriculum as taught, tested skills — not electives, not enrichment, not optional.
• The organizations — and the graduates — that win the next decade will be the ones built for readiness, not recall.

“The illiterate of the 21st century will not be those who cannot read and write, but those who cannot learn, unlearn, and relearn.”
— Alvin Toffler

We banned the calculator and lost anyway. The only question left is whether we’re willing to admit it before we do the same thing again — this time with the stakes an entire generation higher.

About the Author

Paris is an international expert in the field of Operational Excellence, organizational design, strategy design and deployment, and helping companies become high-performance organizations.  His vehicles for change include being the Founder of; the XONITEK Group of Companies; the Operational Excellence Society; and the Readiness Institute.

He is a sought-after speaker and lecturer and his book, “State of Readiness” has been endorsed by senior leaders at some of the most respected companies in the world.

Click here to learn more about Joseph Paris or connect with him on Linkedin.