The Coming AI Teaching Assistant Boom (And Cheating)
Matt Barnum (who is, thank God, now at Chalkbeat) just made three predictions about AI in education, and one of them makes my head hurt.
Students will keep cheating? Yeah, that seems likely. AI will not become a super-tutor? No, of course not, particularly given that support for that concept always rests on the old Benjamin Bloom essay about super tutoring that A) involves human and B) is kind of bunk.
But Barnum also makes this painful prediction: AI will become a ubiquitous teaching assistant.
Lord, but I want this not to be true. I want the woman he uses as an example, a high school English teacher who now uses AI to crank out college recommendation letters-- I want her to be an outlier. I want her to be shamed in her teachers’ lounge. I want teachers who use AI to extrude lesson plans to be embarrassed about it and/or to be teaching at a school where nobody ever looks at the lesson plan-- including the teacher whose name is on it. And if a teacher is using their AI “teaching assistant” to grade essays, I want to encourage that teacher to leave the profession immediately.
And yet, in my pained heart, I know those “teachers” are out there. I have no trouble imagining “teachers” I’ve known who would be delighted about the chance to outsource some of the thinking and effort to some computer program. They’re the same ones who used to use Google to find lesson plans. Beyond that, I know how tremendously pressed for time teachers are, so absolutely crunched that the prospect of getting even a half an hour of their day back would be incredibly tempting.
This is all cheating.
You know how I know it’s cheating? Because all of these scenarios assume that the people on the receiving end of this AI slop will act as if they have received the work of an actual human.
I’m betting nobody is opening their AI recommendation letter with “I am having ChatGPT write this letter of recommendation for Pat McStudent.” No, the letter (just like that ChatGPT paper about Hamlet that Pat submitted) is meant to be taken as the work of the human whose name is attached.
I will predict that AI will kill the letter of recommendation as dead as the follicles on Dear Leader’s dome. Admissions officers will shrug and say, “Well, all of these are from some LLM. There’s no real point in doing this.” And they will look for some other way to find out if a real human who knows the applicant wants to speak up for them. Maybe a phone call.
It will take students roughly five minutes to figure out if their essays are being scored by a machine instead of their human teacher. What will it do to writing instruction and student growth when students realize that they are writing for an audience of zero humans? I don’t know-- but I expect we’re going to find out, and I also expect it won’t be anything good. If no human is going to bother to read your work, why would you put any human effort into writing it?
Some folks boosting (or contributing to) the coming AI teaching assistant boom talk as if teachers will offload some cognitive labor to the machine, and that will be the end of it. It won’t be the end of it. The substitution of AI for human will affect and alter the results. It changes the process, the whole process, from inputs to outputs to reactions to the changed process. The notion that you can just swap out teacher judgment at this one point in the process and nothing else will be altered is naive and foolish. It’s like figuring you can swap out mushrooms for burgers at your barbecue, or replace bolts with molded tofu in an automotive assembly line.
To grapple honestly with all this, a teacher would need to stand in front of her class and announce, “I am not going to grade this assignment. I’m going to have ChatGPT do it instead. What do you think?” Or saying, “I can send a letter to the college for you, but I’m going to have ChatGPT write it. Is that okay with you?” And in a nation of three million teachers, there may well be some that are doing so. I hope there are. I hope there are many, soon to be more. Because it’s going to take a lot more honesty and soul searching and a whole lot less cheating to get through the advent of AI in education.
Honest question: I teach science and my writing is... I think I will call it "unpolished" to save face. When I write something that needs to read as smooth and professional I frequently write a draft that I send to the LLM to polish up. Do you think that the AI phrasing will overpower my actual thoughts when the letter is read?
A wise Yogi said: predictions are difficult especially when they are about the future.
Or something like that.
Education is by definition conservative and is too often reactionary.
Anyone making broad statements about education and especially about the future of education is ignoring the classroom, the students in it. There is no value in this. Education has no value when reduced to abstraction and prognostication.
Teachers cheat.
Students cheat.
This is nothing new.
Technology, be it books or chalkboards or AI have always challenged education to adapt. Unlike mist fields there is no existential threat to the teaching profession as it remains immune to real productivity gains and in this security the fear of technology is merely an expression of fear of much needed change and improvement.