Resist the AI guidance you are being given

Things go off the rails with Chicago Public Schools and the AI for Education organization

Look at those students learning together. This graphic is taken from the report I’m about to savage.

Well, I thought this week I might share some thoughts on the differences between deterministic versus probabilistic problem-solving, but you’re just gonna have to wait for that barn burner of an essay, because something interesting is happening when it comes to AI in education.

The resistance to the nonsense is growing.

A few days ago, my friend Jane Rosenzweig, director of the Harvard Writing Center and author of the great Writing Hacks newsletter, pointed me to this “AI Guidebook” created by Chicago Public Schools with help from an organization called AI for Education. As I sit here typing this on August 21^st, it’s not entirely clear that CPS has pushed this out yet, but AI for Education has posted it on its website and it looks legitimate, so I’m treating it as such.

Yesterday, as I read this document, my initial irritation rose to genuine outrage. Only one thing to do – take to Twitter! And so I unleashed a mini-tweet storm wherein I took issue with some of the most egregiously misguided statements that CPS and AI for Education are promulgating as a matter of policy.

As I type this, that thread has 69,000 views and counting, and has been shared by people worldwide with huge followings (including the fearless Timnit Gebru, which brought me special satisfaction). As much as anything I’ve ever tweeted can be said to have “gone viral,” this has gone viral. Now look, as I wrote just last week, Twitter is not real life – but between this yesterday and the broad pickup of Cognitive Resonance’s guidebook last week, Education Hazards of Generative AI, it’s clear to me that there’s widespread yearning for a counternarrative to the AI hype that is currently dominating the “education discourse.”  

I have mixed feelings about my role in this.  When I launched Cognitive Resonance way back in, um, April, my aspiration was to stay in the “pragmatic center” regarding AI – and that’s still true. What’s more, I was not planning to focus my attention exclusively or even primarily on the use of AI in education. But old habits die hard. With a nod to the Hegelian dialect, if the dominant thesis in education right now is that teachers need to integrate AI into their pedagogy because “the future”…I am prepared to be the antithesis.

Here’s just one reason why. I concluded my thread on the AI Guidebook with the subtle observation that, “The side-by-side cliches of teaching with or without generative AI…are the purest example I've seen yet of the stupidity running rampant in education right now.” I stand by that statement. To the very limited credit of CPS and AI for Education, they at least have the courage of their convictions to state exactly what they hope to see changed in elementary, middle, and high schools as a result of this new technology.

So what is their vision? Let’s take a look:

All of these tasks fall somewhere on the spectrum between “irrelevant” to “actually quite bad if teachers were to do them.” For some reason, ed-tech AI enthusiasts are convinced students will fall in love with math word problems if they are “customized” to student interests; Dan Meyer has called nonsense on this and rightly so. In science, if students actually learn the habitats, diets and survival adaptations of animals that’s a great outcome, but generating an AI image from that isn’t doing any pedagogical work. The suggested AI-fueled activities in literacy and social science are just straight up bad, displacing activities where students might actually think and learn new content with pointless exercises that won’t. (This is one reason why teacher Larry Ferlazzo rightly wonders whether any actual teachers were involved in crafting this guidebook.)

So to summarize: Instead of teachers teaching math, generative AI tutors should do it instead (absolutely not). Instead of conducting actual scientific experiments themselves, students should have generative AI do it virtually (what?). Instead of activating their own background knowledge to interpret what they read, students should have generative AI explore it for them (no no no). To top this off, using AI image generators to explore cultural heritage is just about the single biggest landmine I can imagine a middle-school teacher stepping on in a classroom – good luck and godspeed to anyone following this terrible suggestion.

This is so bad!

I don’t even know where to start. Do we believe students will “receive instant feedback” on their drafts, or might they, say, just have AI produce those first drafts for them? Do we have any reason to believe chatbots can help students understand complex mathematical ideas, given they struggle to do simple calculations and explain basic concepts? What does it even mean for teachers to “facilitate unsupervised use of GenAI by students” in science…isn’t that supervised use? Also, why exactly are we discouraging teachers and students from spending weeks on research and experiments in science class? Can anyone make sense of the word salad suggested for social science? Am I losing my mind here?

So. What to do about this nonsense. Personally, I’m trying to remain calm, knowing that central district offices love to pump out these sort of documents that many if not most teachers will remain blissfully unaware of. But we can’t count on that, and ideas do matter – complacency is not an effective strategy. Instead, here’s my three-pronged sketch of an agenda to resist the AI hype in education:

1.      Speak up. Don’t be afraid to voice critical opinions on AI! It’s really ok to say, “you know I am not convinced I need to integrate this technology into my practice.” Ask AI enthusiasts to provide evidence of impact on student learning (they will not have it). Ask colleagues if they think it important to model critical thinking for students – and if so, whether adults are employing critical thinking in deciding whether to use AI in the classroom.

2.      Push back. There’s a reason the first major resource from Cognitive Resonance is a guide to the many ways in which AI can cause educational harms (find it here). My mental model was to provide educators with something tangible they can point to when saying, “look, I have concerns, and there’s research-supported reasons that I think we should all have concerns.” Seek out other voices in education that are raising alarms about the misuse of this technology – there’s more than you might think, and our numbers are growing.

3.      Embrace the human. Formal education is unique to humans. Our superpower as a species is that we learn from one another and across generations, through teaching that imparts knowledge. Resisting AI hype need not be simply reactive; we can offer a positive story about the process of teaching and learning as fundamentally grounded in human-to-human connections. At a time when people are becoming more suspicious about the pernicious effects technology is having on social cohesion, educators can offer a different vision of the future, one that celebrates the special relationship between teachers and students.

Ok, got all that off my chest. Next week – deterministic versus probabilistic! You’ll be riveted!

---

Quick addendum: As I was hitting send on my essay, veteran educator Peter “Real Talk” Greene was publishing his analysis of the same guidance document. In his view, “it's an exemplar of the kind of wrong-headedness that is evident in so many educational leaders' thoughts about AI.” Link below.

Chicago Schools' Terrible Awful No Good Very Bad AI Guide