What the f*** are teachers for?
What leveraging AI to do teachers' work does to education
Last night I turned to my wife at half past eleven (which I do every night, just as she is drifting into a sense of security that the day is over and she can sleep) and I said:
“Do you think teachers should be able to use AI to grade student papers?”
And she said, tiredly, “No, honey.”
“Why not?”
She snorted. “Well, what the fuck are teachers for?”
And that’s it. Tens of thousands of words and countless acts of thought leadership, and that’s it. That’s the whole damn argument.
The case against grading with genAI
A couple of days ago, the brilliant Leon Furze wrote a piece that went viral about the sharp, discriminatory consequences of using genAI to grade student work.
He showed through real-world examples and experiments with current GPT models that an AI-generated assessment judgement on a student essay will produce vastly different grades based solely on the name typed at the top. GPT-4o gave “Ash” a 95/100 while “Fei-fei” scored just 78/100 for the exact same paper.
Fiddling with prompt data only produces surface-level “improvements”, as does upgrading to a larger/more advanced GPT model. (When I say “surface-level”, I mean it’s not that the judgements become less biased, but that they appear less biased. The bias is still there, it’s just harder to detect.)
And this is huge, and important, and horrifying, and shows the under-the-hood reason why teachers should not use genAI to grade student work.
I just don’t think we even need to look under the hood to answer that question.
Because what we’re really asking is exactly what my exasperated wife1 said to shut me up last night: what are teachers for? This is at the heart of the entire educational technology impasse. When we disagree about whether or not a tech tool is beneficial to learning, what we really care about is whether that tool is helping or displacing human teaching.
What are teachers for?
Here’s the rub (I hope Gert Biesta would be proud): education is learning with teachers. Otherwise, it’s just learning. And if it’s just learning, we don’t need schools. We all know people are perfectly capable of learning by doing, learning by watching, learning by failing, learning by looking up, learning by living. A person (young, old, whatever) can get feedback in all sorts of ways.
But if they’re engaged in education, they should be getting it from a teacher.
Not a chatbot.
Not even partly from a chatbot.
Teaching is an act of relation between humans; a sharing and evaluation and creation of knowledge which enables the student/s to make sense and make use of being human in the world. It can be digitally simulated in the same way as skateboarding can be simulated. With graphics, haptics and audio effects, a video game can make me feel for a moment that I am grinding and sliding, but I am not skateboarding because there is no skateboard.
I don’t know how else to explain it, because it really is that simple.
If we want students to play at getting an education without actually getting one, then let’s go ahead and give them Tony Hawk’s ProGrader. But if we want to actually offer real education, we need to strengthen our teachers, not dilute them.
Of course, the question what are teachers for? is not rhetorical. There are lots and lots of things teachers are doing, or being told to do, or trying to avoid, which are probably in the not this category. And those tasks are ripe for either AI-offloading or, far better, designing out of the education process altogether. It’s just that assessing student work isn’t, and will never be, one of them.
I was recently inspired by Tom Mahoney’s case for opening up a space for dialogue about what teachers should be doing, what teaching is, what education is for. But this, of course, is not just one person’s project. How and where can we do this? Who needs to listen, and who needs to speak?
And how can we amplify the value of teaching in a climate that seems set on suppressing it?
My wife wants to make clear that she was not exasperated at me, but at the relentless AI hype and use-case-fever. Which is very generous of her to say, but it really was 11:30 pm and she had absolutely finished all her jobs.