Pedagogy of the Depressed
An illustrated guide to a smattering of online AI courses for educators
Hello! Are you an educator who’s interested in using AI but unsure of which online “AI for educators” course to take? Not to worry, Cognitive Resonance has you covered! We’ve completed the offerings from OpenAI (in partnership with Common Sense Media), Google, and AI for Education, all of which take an eerily similar approach to explaining what AI is and why you need to be using it in your classroom.
We are now pleased to share with you this illustrated guide to what these courses cover—consider this an early holiday present!
1. First, AI will be defined using anthropomorphized and highly debatable claims about its current capabilities
Everything underlined in red is, um, either not true or else highly contested among cognitive scientists.
OpenAI
AI for Education
2. You will then be urged to personalize your instruction, even though the AI-in-education course you are taking involves nothing remotely resembling personalization
Relish the irony.
OpenAI
AI for Education
(Note: Teaching to specific “learning styles” has been empirically studied and discredited as an effective pedagogical strategy.)
3. Proceed to the obligatory prompt engineering acronyms and mnemonics
These all reduce to “be as explicit as possible when prompting these tools.”
OpenAI
AI for Education
4. Now you’re ready to test out examples of “creative” prompts that should lead you to do some serious pedagogical soul searching
What follows is from my very first attempts to actually use the suggested prompts offered in these courses—I promise these aren’t cherry picked:
OpenAI
Here are screenshots from the “interactive multimedia presentation”—i.e., a PowerPoint deck—that ChatGPT4 provided me after using this prompt:
(Note: No interactive maps or animations were included)
(Note: there were no explanations or resources to access upon completion)
I used this prompt with Google Gemini for a 5th grade math standard on orders of operations. Here’s what it produced:
Oops! (Note: LeBron did not score 38 points in the second half.)
Double oops! (Note: Mahomes did not pass for 500 yards, and multiple steps in this evaluation are mistaken.)
AI for Education
In the section of their course titled “Creative Uses of AI in the Classroom,” AI for Education CEO Amanda Bickerstaff suggests teachers use prompts such as “What if dogs could talk?” to promote interactive storytelling, and “What is a popular holiday in Mexico?” to use AI to promote cultural exploration.
I’m awash in the creativity.
5. Teachers, the hallucination problem is yours to solve!
LLMs often “hallucinate,” that is, they often produce responses that are not what we humans would consider true. What should we do about this? Well educators, I’m sorry to report that it’s on you to check and verify all LLM output and fix all their errors. Or maybe have your students do it.
OpenAI
The use of “seemingly” here just slays me. Everything these models produce is “seemingly” factual, that’s the whole challenge.
I genuinely laughed out loud at the observation that LLMs may invent quotes and produce responses that are “utterly meaningless.” Google, points for honesty!
AI for Education
This explanations of why LLMs “hallucinate” is not remotely true. For a wonky but thorough exploration of what’s really happening, I recommed this overview from Colin Fraser. The hallucination problem cannot be fixed by making training data less biased or more complete; in fact, as far as we know right now, there is no way for anyone to reliably eliminate hallucinations.
And yet AI for Education then proposes that students should be trained to spot these mistakes:
I do appreciate the acknowledgment that reputable websites and textbooks are more reliable than AI. Maybe teachers and students should just use those resources in the first place?
6. Let’s conclude with some vague hand waving toward ethical issues
Look, it’s not easy to navigate the ethical considerations surrounding the use of AI in education. I wrestle with these issues myself, given that I’m using these tools in the process of helping to demystify them. So I want to grant these organizations some grace and leeway as we all try to figure out what’s acceptable and what’s not. But that does entail digging into some thorny issues, not just breezing past them or telling educators they have to figure it out themselves.
See if you can guess what approach these courses take…
OpenAI
The burden again falls on the users—teachers—to mitigate the inherent flaws of the tool.
All fair points, but what if “keeping in mind” these biases lead you to conclude it’s unethical to use these tools at all? That’s the possibility that is never acknowledged in these courses.
AI for Educators
So to use AI responsibly, teachers, you just have to evaluate and verify all of its output, provide critical feedback and oversight to others using it, revise all results to reflect your unique needs, and remember that in the end, you the teacher bear the ultimate responsibility for the output. You you you you YOU.
Cool.
The Pedagogy of the Depressed
The companies promoting AI in education through these courses gave themselves no easy task. As we learned en masse during the pandemic, online education in the form of asynchrous activity is pretty lame. People don’t want to learn this way. It’s almost as if hundreds of thousands of years of human cultural evolution have primed us to want to learn from each other, rather than by passively clicking through digital presentations online.
But let’s leave that aside. What’s more striking to me is how fundamentally boring all this is, even on its own terms. Despite all the hype about how AI might revolutionize education from top to bottom, these courses mostly offer pedagogical examples of modified lesson plans and tweaked scoring rubrics. That’s not entirely useless, of course, but I find it…depressing…that the opportunity to use these tools to really explore what is happening in our minds—and how this differs from what LLMs are doing—is just blown right by on the way to prompting acronyms and algebra mistakes.
It doesn’t have to be this way. I swear it’s possible to create online educational experiences related to AI that are informative and fun—here are four examples. Frankly if I had the resources—hint hint donors—these are the sort of things I’d love to be making myself. And for all my criticisms of generative AI, the introduction of this technology creates an opportunity for us to collectively think about what makes humans special, to examine deep questions about how our minds work, what constitutes thinking and consciousness, and so on. We could be entering an era of truly transformative intellectual advancement of understanding ourselves.
But to get there, we’re going to have to do better than this pedagogy of the depressed.
This is fantastic. It came at a great time as I prepare to present on Friday. This is a great resource that I will link for the audience.
This is invaluable! And I love your title... I'll have to do some soul-searching on whether I've been guilty of any of these sins...