Tears welled in her eyes, “After a year in her class, how could she assume I’d cheat?”
On the last day of classes last year, my daughter submitted her final assessment for 11th grade IB math: a paper including research questions, samples and a method for testing her hypothesis using calculus and geometric formulas.
The next day her teacher sent her a cryptic message, “Your Interim Assessment was flagged for AI usage. Please make sure to fix and resubmit by Sunday.”
Fix and resubmit. That was all.
My daughter panicked. She wrote back asking what exactly what to fix because she had not used AI to write her paper. Her teacher finally sent a screen shot from the scan indicating 24% chance of AI usage, still without comments or specific evidence. More baffling, the AI detector objected to her most unique aspect of the paper, her personal reflection on why she chose to explore the relationship between crochet patterns (a hobby she’s had since 6th grade) with math and geometry.
Teachers are increasingly using AI detectors to scan students’ work for authenticity. As a former adjunct professor, I sympathize with the urge. Students, like the rest of us, are figuring out AI’s magical potency. But without reviewing the work or considering a student’s capabilities, relying on AI’s judgement feels is not sound judgement. And it’s inconsistent with NYC Public Schools’ guidance. “Educators must critically evaluate all AI-generated output [by AI detectors] for accuracy, appropriateness, and potential bias. AI responses should never be accepted at face value” NYC Public Schools’ [interim] Guidance on Artificial Intelligence (AI) March 2026.
And here lies part of the problem. School systems are scrambling to write policies that may never get ahead of what is already happening. And policies rolled out this urgently without sufficient implementation jeopardize their utility.
The teacher’s message did not include detailed feedback on language, concepts, logic or formulas to “fix” a paper that serves as the end of year assessment. For my daughter, who produced her school’s highest score on the Algebra Regents exam in 8th grade and takes school seriously, this accusation was devastating. But this isn’t just about her. We can’t let AI become the new school bully.
I’m not naive. Students turn to AI for help, as do teachers. It saves time writing and grading. But the current rules are vague. We don’t know what software is allowable, or how any AI software is allowed to be used. And even if the AI detection tools that teachers use (like Turnitin or Copyleaks) were more reliable, there’s still a mis-match between the software promises and how they’re used. As with any AI tool, human judgement must be part of the decision making equation. And right now sound judgement that proceeds cautiously and with care in the fraught education environment needs better guidance.
Besides its impact on genuine learning, the importance of trust between teachers and students is under threat and cannot be ignored. If a child cannot trust that a teacher has their best interest in mind, how might that perception influence a student’s level of effort and resilience? Eroding teacher-student trust not only contributes to a negative learning experience, it also harms social stability in the classroom and contributes to moral ambiguity. Particularly when detection tools, notorious for producing false positives, can automatically be deployed on a scale that could include all students for all assignments and all classes in all schools, the breakdown of trust contributes to a sense of injustice like assumed guilt, while strengthening perverse incentives to avoid AI detection, whether it’s been used or not.
I’m not writing off AI in education. But I worry that whatever benefits it may offer are being overshadowed by its potential long-term harm like declining human reason and empathy. And after the public uproar following NYCPS preliminary guidance in March, this school year doesn’t seem to offer much protection against misuse of AI: “AI may assist educators, but it does not replace professional judgement.”
We can’t blame teachers entirely for sloppy use of AI detectors. Decision makers who spend public money in support of policies must also be held accountable. The policy makers, elected officials and administrators must think through the secondary and tertiary harms of AI in education, including indiscriminate usage of AI detection tools for vetting student work.
The next afternoon, after her teacher’s original message to “fix” her assessment, my daughter added photos of her work samples and made a few minor edits. In the end, it was accepted without penalty, but only two days of tears and frantic phone calls. Has she figured out how to work around AI detectors that are notoriously imperfect? Maybe. But the collateral damage of mistrust is already done.
I don’t doubt that AI detectors could get better. Models could be trained on each student’s prior work in a subject. To be more credible, it could use the course curriculum, grading rubric and output of peers for the same course and assignment. In fact, the companies creating these AI detection tools are probably at work now, trying to perfect their products to make them harder to resist. And yes, there are exciting potential benefits.
But as humans, we must do better than default to machines. We must consider the value of trusting each other and our own judgement when we ought to know better.
100% written by a human
