Why Banning AI Won't Fix What's Actually Broken
An education researcher argues AI shortcutting is rational behavior in grade-focused schools. Bans won't fix it - the incentive structure will.
Anshika Verma · Founder, GrowWise · Last updated
In an April 6, 2026 EdSource commentary, USC researcher Erika A. Patall argued that schools reward performance (grades) while claiming to value learning - so "students' use of AI to perform better while learning less is entirely rational behavior." She predicts bans and detection software will fail on their own, and recommends redesigning what schools reward: process-oriented feedback and genuine competence over output scores.
An Education Researcher Just Explained Why Banning AI Won't Work. She's Mostly Right.
Most responses to AI in schools fall into two camps: ban it, or build detection software to catch it. Erika A. Patall, writing in EdSource on April 6, 2026, made a case that neither camp is addressing the actual mechanism.
Her argument, compressed to one line: schools say they value learning, but they systematically reward performance. Grades measure output, not the process that produced it. Given that incentive structure, a student who uses AI to produce a better-looking result while learning less isn't cheating the system - they're responding to it correctly. "Entirely rational behavior" is her phrase.
Why bans alone don't touch the incentive
A ban removes access to a tool. It does not remove the reason a student reached for it. If the grade is still the thing that determines the reward - the GPA, the class rank, the college application line - then removing one shortcut just sends students looking for another, or back toward whatever produces the grade with least effort.
Patall's underlying claim about learning itself is the part worth sitting with: learning requires struggle and error-correction. AI, by design, removes friction instantly. That's a feature for output and a bug for skill formation. Research she cites shows AI use can reduce skill acquisition and undermine self-regulation - the capacity to notice your own confusion and work through it, which is arguably the core skill school is supposed to build in the first place.
Her recommendation, and where it gets hard to execute
Patall's fix is to replace grades as the organizing principle of school and build around three things: autonomy (pursuing goals that feel meaningful to the student), genuine competence (not performance theater), and connection to peers and teachers. Concretely: meaningful choice in assignments, feedback focused on process rather than final score, normalizing setbacks, and designing instruction from the student's actual perspective rather than the administrator's.
This is a good argument and a hard one to run at scale. A single teacher can give process-oriented feedback to 15 students. A teacher with 28 students across five sections, already working a 49-hour week with 53% of the profession reporting burnout, cannot manually produce process-level diagnosis for every student on every unit without help. The idea is right. The bottleneck is capacity, not conviction.
The missing piece Patall doesn't name - but her argument requires
If a school wants to reward process and competence instead of just output, someone has to be able to see the process. Right now, for most teachers, the process is invisible. They see a score. They don't see which specific misconception produced it, whether it's the same misconception behind three other students' wrong answers, or whether last week's reteach actually closed the gap before the next unit built on top of it.
That is exactly the layer teacher-facing instructional intelligence is built for: not replacing grades with something softer, but giving a teacher the process-level evidence Patall's framework requires - which student, which misconception, what to reteach, whether it stuck - at a scale one person could never produce by hand. It doesn't answer every question in her argument. It answers the one that makes the rest of it operationally possible.
See Teacher-Facing AI in Action.
Related reading
FAQ
What is Erika Patall's argument about AI and school grades?
That schools claim to value learning but structurally reward performance through grades, making it rational for students to use AI to improve output while learning less. Published in EdSource, April 6, 2026.
Does she think schools should ban AI?
No. She predicts bans and detection software will fail because they don't address the underlying incentive structure that makes AI shortcutting rational in the first place.
What does she recommend instead?
Redesigning what schools reward: supporting student autonomy, genuine competence, and connection to peers/teachers, through practices like process-oriented feedback, meaningful choice, and normalizing setbacks rather than just grading final output.
Why is this hard for schools to actually implement?
Process-oriented feedback requires seeing what happened during learning, not just the final score - something one teacher can't manually track across 25-30 students per class without a system that surfaces the evidence for them.
Sources
Frequently asked questions
- What is Erika Patall's argument about AI and school grades?
- That schools claim to value learning but structurally reward performance through grades, making it rational for students to use AI to improve output while learning less. Published in EdSource, April 6, 2026.
- Does she think schools should ban AI?
- No. She predicts bans and detection software will fail because they don't address the underlying incentive structure that makes AI shortcutting rational in the first place.
- What does she recommend instead?
- Redesigning what schools reward: supporting student autonomy, genuine competence, and connection to peers/teachers, through practices like process-oriented feedback, meaningful choice, and normalizing setbacks rather than just grading final output.
- Why is this hard for schools to actually implement?
- Process-oriented feedback requires seeing what happened during learning, not just the final score - something one teacher can't manually track across 25-30 students per class without a system that surfaces the evidence for them.
See teacher-facing AI in action
AI that helps teachers see misconceptions and decide what to reteach next, without student chatbots.
2026 AI-in-schools policy comparison · Five questions before signing an AI vendor