The Wall & I
Some rambling thoughts on the state of AI in Education
I’ve realized I haven’t written in this space for quite some time.1 With July 1st now in the immediate rearview, I’d like to change that and to reinvigorate this space for the 2024-2025 school year. I’m certainly writing lots elsewhere (over at EA’s CTL Substack The Beacon mostly, but also lots of emails that are too long, and my ever present and mosaic Twitter feed).
What follows is a response to an inquiry from an educator over in a Slack channel of like-minded, interesting, and super thoughtful folks. I’m breaking contain and sharing to my personal Substack, because it captures some wide-ranging thoughts on the state of Generative AI in education. And I am frequently asked these questions, or questions like it. For some folks this will all be fairly obvious. For others, it might be really informative. But I spent lots of time crafting the message over in Slack and when I got to the end, I realized more folks outside of that space might benefit from it. What follows is a lightly edited version of my response. It’s a Slack response to a thoughtful question. It’s not an academic article. There will be some holes. And I am speaking from my personal stance, while referencing some of the collective work we’re doing at our school.
Question: What do we think of this (the release of Khanmigo’s teaching assistant)? I’m eager to hear Justin’s take.
My response:
So it's probably best to start from my personal perspective on AI in education, that being one best described as a pragmatic optimist. I tend to believe the impacts of generative AI on education will be profound and, in due time, a net positive—but that doesn't mean I think schools should rush head first into adoption. The barriers to entry for many educators on the AI front are often profound. It can be daunting to stare at a blank GPT-4 window (or Claude, or Gemini, etc.) and know what to do next. Something we espouse frequently here is a stance of playful iteration. The best way to learn these tools is to get your hands dirty and to play with them. Now when my 9, 7, and 4 year old boys ask me to get down on the floor and play with them and their Transformers, that's a daunting task for a 41 year old. I forget how to play. Asking professionals to play with a blank computer window that acts as a kind Wizard of Oz can feel doubly odd. But it's a useful metaphor in many ways—once we get over the aspects of ego and identity that can cloud this approach (which are real and I am not lamenting or diminishing these feelings many educators possess).
Khanmigo lowers the barrier to entry for many, and it can help folks begin to explore some of the ways they might offload some of the tedious aspects or tasks associated with teaching. There are some sound aspects in the platform that can help early-career teachers and even veteran teachers streamline things, expand on thinking, or to serve as a kind of eager and useful teaching assistant. And the stakes are really low. But again, the best way to experience the potential—and the drawbacks—is to try it out. At the very least the advent of this platform creates a more equitable point of entry for educators. I'm not necessarily advocating for folks to rush to adoption. Having a stance of critical consumption is a great place to start. But we all know teaching has in many ways gotten more demanding in the last decade plus. If I'm rushing from a department meeting, to a call with a parent, and I'm looking for an Exit Ticket to understand student learning and the effectiveness of a particular lesson for the class I need to teach four minutes later? Khanmigo can give me some pretty good options, really quickly and mostly effectively. Which leads to the next issue, and one that I don't think many are considering as deeply as they ought to. We can think of three domains for AI in education: AI for teachers, AI for students, and AI that sits at the intersection of teaching and learning (really the interplay between teacher and student). Khanmigo can do all of these three things, but it can only do the third one if your school supports Khanmigo Classrooms. At the very least, Khanmigo can act as a kind of force multiplier for a teacher. But doing this means letting go of some things folks might hold as central to their work as an educator. And I understand that, certainly more now than I did a year ago.
For me, personally, Khanmigo is a kind of AI 1.0 for educators. It can do lots of low-level tasks well. To use another metaphor: Khanmigo is kind of like getting a food subscription akin to Hello Fresh or Blue Apron. It's packaged in such a way to promote efficiency and ease. They might be really happy "cooking" 2-3 meals per week with the subscription. Or, it might inspire folks to explore more complex culinary pursuits where they hone their knife skills, start surfing the web for obscure ingredients, and start dropping turmeric and rare caviar samples in every other meal. Heck, it might even turn some people off because they like grocery shopping for their particular ingredients and developing something entirely on their own. All of these are reasonable perspectives. I do think it behooves us in education to at least have some degree of AI literacy. Ed Tech has already started to bulldoze the space. Google has turned on many AI features for students recently. It's not going away, it's only going to become more ubiquitous, and it will stay largely undetectable.
I'll close with this. We use Lee Shulman's work on "teacher expertise" (see image above) as a kind of foundational stance here in the CTL. AI can do many of the things on the right side of the circle (in dark) really well. It can't do any of the things on the left side (in orange) very well, yet. And these are the things that make education inherently worthwhile, in my opinion. They're human. And for me, that's what this is all about. If we can empower folks to lean more into the left side, I think it assuages many of their concerns, and arguably, makes us better. My personal belief is that AI can be an incredible positive for teaching and learning, when deployed thoughtfully. But it's also an undeniable disruptor for assessment practices, which sits at the core of so much concern for education. That says more about the state of education than it does about the state of AI. But these are my opinions. They might come off as strong, but I hold them pretty loosely. I've worked in fairly narrow contexts of independent education and I am speaking from my experience at three different institutions over the last two decades—others likely (hopefully?) have different experiences. If you want more on the AI front the work of MIT's Justin Reich and the University of Mississippi's Marc Watkins come to mind. So too does the work of Andy Matuschak and of course Ethan Mollick.
Justin's stuff is freely available in many places—I like this one on AI as arrival rather than adoption tech and his TeachLab podcast, where he highlights a recent keynote address.
Marc has a really great Substack, Rhetorica. You should read every post.
Andy Matuschak is a really interesting EdTech thinker that I sense has a really good perspective on the longterm future of AI in education. He's kind of a futurist in this sense. It's a lot of signal, but the signals are talking. His recent essay Exorcising the Primer is primarily for an Ed Tech audience, but really informative and exciting.
Ethan Mollick out of UPenn is kind of the leading voice here, but the aforementioned folks are all worth your time. You can start with his Substack, One Useful Thing but he has a ton of videos and a book out there.
There are more critical voices out there, but they trend towards Luddism, which turns me off. I learn from them, but they seem to ignore the reality of the world that our students will likely inherit. I haven't even begun to touch on the real (and negative) implications for things like the environment and working conditions of folks behind the scenes of the flashy tech. I wish Audrey Waters was still writing in the Ed Tech space.
We're doing quite a bit on the AI front. We don't have any answers, but we have explored some interesting questions.
I always pull blog titles from song titles or lyrics. It’s a thing. It’s fine. Just go with it. I named the blog ‘The Academic DJ’ after all. Today’s title comes from ‘Nation of Language’s’ The Wall & I. It felt appropriate given the fact that I was breaking the Slack wall, and also because of the LLM aspect. Plus, it’s a synth-heavy bop that has been on heavy rotation of late.