AI surveillance and the deprofessionalization of teaching

Recently, Michael Petrilli floated an idea that, on the surface, sounds like a dream for data-driven researchers: What if we asked a large group of teachers to wear AI-powered recording devices, then turned everything they said in the classroom into searchable transcripts? Imagine the possibilities: we'd know exactly which instructional materials teachers were using, how consistently, and even which methods seemed to correlate with better outcomes.

As someone who's spent years in both education policy and research, I appreciate the allure of this proposal — and also feel a deep unease about the direction we're heading. Posts like Petrilli's only deepen a feeling I've carried for a while now: we've lost our way when it comes to understanding and improving teaching.

The myth of the 'unchanged classroom'

People love to say that education never changes, that classrooms today look exactly as they did a hundred years ago. I've never bought that narrative. Yes, if you squint, you'll see a teacher at the front and students in rows. But the resemblance is shallow, and fixating on aesthetics ignores real, often profound shifts — some positive, some not — that have occurred in who teaches, what is taught, and, just as importantly, how we value the profession.

When I look back, I think not just of the industrial-age classroom, but even further, to the one-room schoolhouse. There, the teacher was a pillar of the community, someone townspeople respected deeply, trusted with their children's learning and character development. That social contract, as old-fashioned as it may sound, rested on respect and professional autonomy. It's a far cry from today's tendency to look for 'the answer' in increasingly granular research or technological intervention — while underpaying, overburdening, and frequently second-guessing teachers.

Chasing effect sizes while undervaluing teachers

Much of our research energy these days is devoted to the search for very specific, actionable tweaks: instructional techniques, curricula, professional development models. That approach isn't inherently wrong; evidence-based practice matters. But I sometimes feel like we're asking the wrong questions — or perhaps, asking them only because they're the ones left to us after we've abdicated responsibility for the bigger picture.

We have a system that underpays teachers and, in many cases, has streamlined teacher preparation to the point of minimalism (alternative cert programs, 'fast tracks' from industry, or military-to-classroom pipelines). When results inevitably fall short, our response is not to rethink the underlying support or status of the profession — but instead to redouble research efforts on the small stuff, hunting through the weeds for ideas that might buy us another half a standard deviation in outcomes.

Imagine if we took the same approach with brain surgeons: slashing their training time, paying them modestly, offering 3-month military-to-operating-room pipelines — then funding studies on whether a new ergonomic scalpel would yield slightly fewer errors. None of us would find this reasonable.

Yet this is the norm in education: we demand superhuman results with subpar support, and try to engineer our way out of the results with data and algorithms.

AI, surveillance, and the illusion of control

Petrilli's proposal is just the latest — and most high-tech — version of this paradigm. Record everything a teacher says, analyze it with machine learning, then optimize practice. It's appealing in the way that all big data solutions are: it promises control and clarity in an impenetrable, complex system. And to be fair, there have been promising advances: AI can code classroom interactions, flag open-ended questions, and maybe even differentiate between more and less effective questioning styles. Similar successes exist in parsing digital footprints in online learning environments.

And it's no coincidence that such surveillance proposals target teachers first. In a profession we've systematically stripped of autonomy and respect, it's easier to frame constant monitoring as 'research' or 'improvement' rather than what it often is: a continuation of the very deprofessionalization we're supposedly trying to solve.

But what gets lost in this obsession with measurement is the essential nature of teaching. It's profoundly relational, messy, and rooted in trust. Good teaching isn't just a collection of reproducible moves — it's the artful combination of expertise, context, intuition, and care. Reliance on surveillance and data doesn't just risk creating a chilling effect in classrooms; it actively diminishes the craft by suggesting that what matters most can be captured, coded, and compared.

I believe in research. I want more of it — more thoughtful, rigorous engagement with what works. But I can't pretend that ever-more-intrusive forms of data collection will solve the real problems facing educators or students.