The other day I had a friend, someone who knows my work, and who even works in education-adjacent spaces ask, “Have you written anything about that ChatGPT stuff?”
Have I? It seems like that’s all I’ve been writing about, to the point where I purposefully go looking for any other topic for this space these days. When I consider what I want to write about, I sometimes imagine a reader who literally reads everything I post and then try to judge if they will feel like I’ve been repeating myself. For sure, there’s areas that I return to over and over, but I want that hypothetical reader to feel like they stand a decent chance of encountering an idea they hadn’t heard from me before.
This also happens to be a good way to keep the work interesting for me. Writing is a tool for figuring out what I think and believe, and to endlessly go over old territory is fundamentally uninteresting.
But, in talking to this friend, and from a few other conversations I’ve had, it’s clear that even as ubiquitous as talk about the impact of generative artificial intelligence (AI) on education may seem to me, I think the reality is that most folks are too pressed to attend to what seems to be a rapidly evolving space.
Going back through my archives, I have written a lot about the challenges of assigning, teaching and assessing writing in a world where ChatGPT (and other apps of that ilk) exist, but one of the things I realized in this archive review is that I think it is a mistake to see this as a rapidly evolving space.
Yes, on the one hand, OpenAI seems to be regularly rolling out new capabilities for its technology.
But the core problem people in education must grapple with hasn’t changed at all: How do we help students learn? In this case, I’m specifically focused on helping students learn to write. ChatGPT doesn’t alter the problem in the slightest. It may be part of the solution to this problem, but we are very very far from having sufficient evidence to make that determination.
Jane Rosenzweig, Director of the Harvard Writing Center, frames it clearly with her question, “To what problem is ChatGPT the solution?” When we look at the “problem” of learning to write—which is not the same thing as producing written artifacts—it becomes very difficult to see if ChatGPT has any utility, except at the margins.
But of course, the technology is here. Students are using it and instructors are getting hugely frustrated knowing that students are outsourcing the work to the technology, making the whole exercise seem kind of pointless. Just because I think there is a solid case that ChatGPT (and its ilk) have little to no productive role when it comes to learning to write doesn’t mean it can be ignored.
But if we’re going to think about it, we have to do it in the context of the problem we’re trying to engage. I’ll say it again: How do we help students learn?
For the sake of my friend and anyone else who thinks additional perspective might be helpful, I’ve compiled some of what I’ve been writing on this topic since the arrival of ChatGPT in November of 2022. There’s more where this came from. I guess I really have been writing about it a lot.
How to Think About Generative AI
My most important recommendation is to not get wrapped up in the hype that surrounds any new technology, particularly one as unproven at generative AI and large language models. To that goal, I’ve written a couple of pieces about not falling for the hype or falling prey to FOMO.
Remember that almost everyone who is productively using generative AI now had no experience with the technology prior to November, 2022. The notion that the sooner one trains to use the technology the better just doesn’t follow.
“Resisting the Hype Cycle in Education”
“ChatGPT Both Is and Is Not Like a Calculator.”
How to Think About Assignments in a Generative AI World
One of my mantras about learning is that it is done through experiences. The chief threat of ChatGPT to the kind of writing students often do in school is that its output can be substituted for the experiences we’ve been asking students to do.
My view is that if this is the case, we should examine those assignments primarily from the point of view of what experiences we’re asking students to have, and whether or not those experiences prior to ChatGPT were truly associated with learning as opposed to what I call “academic cosplay” activities which allow us to hold on to an illusion of learning for the purposes of schooling. As I argue in the first piece linked below, we must make students write, but if we make them write, we have to provide experiences which are linked to learning.
“We Must Still Make Students Write”
“If ChatGPT Can Do It, It’s Not Worth Doing”
“How About We Put Learning at the Center?”
How to Think About Assessment in a Generative AI World
The problem of students turning to ChatGPT to complete work we would rather see them engage with without the intervention of the technology cannot and will not be solved by assignment design alone. For sure, authentic assignments that students want to engage with help, but they are not in any way sufficient.
This grouping tries to show that when it comes to meaningful assessment, the problem has not changed, and we should’ve been focused on a more meaningful array of criteria all along.
“ChatGPT and Writing Assessment, an Old Problem Made New”
“On AI and ‘Meaningful’ Feedback”
Why We Shouldn’t Let ChatGPT Teach
I’m a bit of a fanatic on this front, but it simply makes no sense to me to have something that cannot read, think, feel, observe or communicate with intention respond to student writing as though that response has meaning. To believe this is to embrace a delusion that is not rooted in genuine educational values and that privileges schooling over learning.
“Reject Automated Grading of Student Writing”
Signs of Progress in Teaching Writing in a Generative AI World
While the frustration and despair some are experiencing is real and understandable, I think it’s also important to celebrate the way that adapting to this new reality has unlocked improved ways of thinking about and responding to student writing.
“Making Progress Against ChatGPT”
“ChatGPT Can’t Kill Anything Worth Preserving”
I hope this helps. If anyone wants to talk about their own challenges and approaches, I’m easy to find. Don’t hesitate to reach out.