We Keep Asking What AI Should Do in Schools. We're Asking the Wrong Question.

There's a framework getting attention in education circles right now that draws a clean line: AI can be a flashlight, but it should never be the gavel. Let AI help with low-stakes, reversible things — a hint, some extra practice, adaptive drills. But keep it out of the high-stakes, irreversible calls — graduation eligibility, special education placement, college admissions. Those need a human who can actually override the machine, not a rubber-stamp review that just performs oversight without exercising it.

It's a good framework. But it's not just a policy theory anymore — we now have data that shows exactly why the line matters.

The Proof: What Happens When You Take the AI Away

A study tracking 26,000 students over 30 months found something that should stop every parent and educator in their tracks: students using AI saw their homework scores rise 18%, and they finished assignments about 30% faster. On paper, that looks like success. But the same students saw their closed-book exam scores fall 20% within six months — and their entrance exam scores dropped 18–24%, with the worst of it not even showing up until nearly two years of AI use had passed.

In other words: AI made these kids look like they were learning faster. It didn't make them actually learn. The homework grade became a worse and worse signal of real ability the more a student leaned on AI to produce it. That's the flashlight-vs-gavel problem in miniature — the moment you take away the tool and ask a student to perform under real stakes, the truth comes out.

And this isn't just a K-12 pattern. The exact same gap is showing up one level up, in college classrooms. Brookings Institution's 2026 global report on AI in education — an eighteen-month study spanning students, teachers, parents, and technologists across more than 50 countries — points to the assignment-to-exam grade gap as the clearest warning sign that a student is leaning on AI as a shortcut rather than using it to actually learn: strong grades on AI-assisted assignments, followed by weak performance the moment the exam takes the tool away. It's not just researchers noticing it, either — a national survey of more than 1,000 college faculty found that 95% believe generative AI will increase student overreliance and weaken critical thinking, and a large share say they're already seeing it happen in their own classrooms.

So the pattern holds at every stage: high schoolers, college students, and now the faculty grading them are all describing the same disconnect between the grade and the actual skill behind it.

Why This Actually Matters — Colleges and Employers Are Already Living It

This isn't a someday problem. It's already at the college gate. More than 1,800 professors across the University of California system have signed an open letter warning that incoming undergraduates are showing up without middle-school-level math skills — and they're pointing directly at AI reliance and shrinking use of standardized testing as contributing factors. Employers are seeing the downstream version of the same thing: basic quantitative reasoning — reading a chart correctly, working with fractions, avoiding the kind of graph-literacy mistakes that used to be rare — is now something hiring managers can no longer assume a diploma guarantees.

So the pattern is consistent at every level: high school, college, and the workforce are all reporting the same gap. Kids can produce work that looks competent. Fewer of them can demonstrate competence when the scaffolding is removed.

The Deeper Problem: We Asked Schools to Be Everything

Here's where I think the conversation needs to go next, because AI policy alone won't fix this. We've spent the last two decades asking schools to be academic institutions, food banks, mental health centers, and crisis response teams — all at once, with the same staff and the same six hours a day. When a system tries to do everything, it ends up doing nothing particularly well. Marva Collins proved the opposite approach decades ago: she took kids everyone else had written off and simply refused to lower the bar. "You can learn. You will learn. I am not giving up on you." Her students didn't need the system to solve every problem in their life first — they needed adults who held a high standard and taught them to meet it.

AI didn't create the mission-creep problem in schools. It's just the newest thing getting asked to compensate for it.

Where I Land — and Where Microschools Come In

I don't think AI or tech is the enemy here, any more than typing or basic computers were when I was in school. Those were tools for doing a job — not a replacement for the actual, hard work of learning and living. Tech, including AI, absolutely has a place in education. But it's a tool that builds on a foundation — it isn't the foundation itself.

That's why I believe tech has no business in a classroom before 6th grade. Grades K–5 should be tech-free, full stop — 100% focused on the unglamorous, foundational work: reading, math, grammar, the basics that everything else gets built on top of later. Once that foundation is solid, tech and AI become genuinely useful accelerants, not substitutes for a mind that never got built in the first place.

This is exactly the gap microschools are stepping into. Not by replacing parents or replacing schools' academic mission — but by rebuilding the infrastructure that's been quietly breaking for years, and by bringing parents back into the process instead of outsourcing everything to an overloaded system.

Parents need training and a seat at the table again, not a system that hands them a permission slip and calls it partnership.

Let's Put Education Back on the Throne

None of this — the AI data, the college alarm bells, the employer skills gaps — is really a technology story. It's a story about what we've allowed education to become, and what we're willing to do to fix it. This isn't being talked about enough outside of the education system - public or private sector. This is a national issue. This is a societal issue. For our kids' sake, it's time to make education a real priority again — not a dumping ground for every social problem we haven't solved elsewhere, and not a race to hand a 7-year-old a chatbot before they've learned to add fractions. Build the foundation first. Let the tools come after. And bring parents back into the room where it happens.

Sources

1. Eric Tucker, "A Hint is Not a Diploma: Consequential Educational Decision Making in the Multimodal AI Era" — the original flashlight-vs-gavel framework article (the one you uploaded)

2. 26,000-student study on AI and academic performance — covered by Psychology Today ("A Study of 26,000 Students Shows the AI Learning Trap") and The Decoder ("A 26,000-student study shows AI's hidden learning cost takes two full years to surface")

3. Brookings Institution, 2026 global report "A New Direction for Students in an AI World: Prosper, Prepare, Protect" — via USCA Academy's writeup ("Can AI Help Students Improve Their Grades? What 2026 Research Says")

4. AAC&U / Elon University national faculty survey (Jan 2026) — via Elon University's news release and AAC&U's own press release ("95% of college faculty fear student overreliance on AI")

5. UC system professors' open letter — via Dainik Jagran English ("Top Universities Raise Alarm as Students Struggle With Basic Math and Reading Skills")

6. Employer/graduate skills-gap piece — The Queen Zone ("12 basic Math skills employers say most new graduates completely lack")

7. Jennifer Ludy, "Schools Exist to Educate — Not to Be Everything to Everyone" (Nov 13, 2025) — the mission-creep argument and the Marva Collins quote