Teacher Spotlight: Let's Play a Game

The winners get to grow as teachers. (Plus, I'm cleaning the link trap.)

Here is the game. I am going to post a two-minute clip of a teacher teaching math. (It’s this lesson in our curriculum.) Identify three moments of great teaching in that clip and you win. Identify a moment that no one has yet identified and you double win. I am going to very very subtly indicate three moments of great teaching as I see it, but I will not say why. Your move.

I’m going to spend the rest of this newsletter clearing out my link trap and adding commentary.  Most of those links will consider—I swear you will never guess this—generative AI.

Odds & Ends

¶ It feels like every article about Khanmigo, including this one about its rollout in nine Palm Beach County high schools, contains a paragraph disclosing that the implementation is funded by some wealthy benefactor or agency.

Implementing the Khanmigo program is being funded in partnership with the Stiles-Nicholson Foundation, a local nonprofit organization that supports education. The foundation is providing up to $2 million for the use of the platform through June of 2025.

That’s fine, of course, but it raises questions. Will the cost of generative AI converge to a price that excites teachers, schools, and districts? If these tools save significant teacher time, increase student learning, decrease teacher turnover, why wouldn’t districts pay full freight for them now? What fate awaits these chatbots once they are subjected to the purifying fire of the free market, district procurement processes, competitive bidding, etc? Maybe a good one! I only note that, for now, the excitement has been subsidized.

¶ Matt Barnum continues to do exciting work from his new post at the Wall Street Journal. His recent article on Khan Academy and Khanmigo is a great deal more critical than we usually see from the business press. With this quote and others, I continue to wonder, “Which kids are excited to talk to a chatbot?”

I don’t see a huge amount of my lowest-proficiency and lowest-motivation students gravitating towards it.

(It’s also worth noting that the Khanmigo rollout profiled in the article was subsidized by another $2 million grant.)

¶ The Markup has a breezy roundup of different ways that teachers are using generative AI, including perspectives from teachers outside the western world.

¶ EdSurge reports from a SXSW EDU panel of early childhood experts discussing generative AI and identifies a demand-side challenge in early education:

Still, with so many other pressing concerns in the field, few early childhood educators are hungry for AI, the panelists admitted.

“It's a little bit like, before the iPhone was created, asking people if they wanted all their songs in their pockets,” Stokes said. “It's hard to imagine what's possible until you create some very specific wins that move the needle.”

I don’t know if that’s right, though. If you told me, pre-iPhone or pre-iPod, that I could have zillions of songs in my pocket all at once, my answer would have been, “Cool gimme.” That was my answer, in fact, and the reason why I bought an iPod. “Teachers just don’t understand generative AI yet,” is how most people explain its lower-than-expected uptake among teachers. But another possibility is that teachers do get it, maybe think it’s neat, but don’t think it has a zillion-songs-in-your-pocket killer application for teaching.

¶ Accenture acquires early MOOC darling Udacity for what TechCrunch speculates is a ~90% discount on their 2015 valuation. Former edtech critic Audrey Watters falls off the wagon and takes a victory lap. Where are the other big three MOOC providers right now? EdX sold itself to 2U. I have no idea how to score that one. Coursera’s stock is down 70% from its IPO. I don’t know, folks. It’s exciting to think you found a technological cheat code for one of the oldest and hardest processes in organized human society. Every generation gets excited about a cheat code, and sometimes more than one, but at a certain point you should wonder if education isn’t harder than it looks.

¶ John Warner encourages colleges and universities to chill and not give into FOMO w/r/t generative AI. Among other reasons, he cites higher ed’s panic about MOOCs which led to the firing of prominent administrators for not converting their face-to-face courses into YouTube videos and BuzzFeed quizzes fast enough. Sober advice, IMO.

¶ Education Week reports on an AI chatbot designed for teacher-candidates:

In the scenario, the teacher-candidate sits down at their students’ lunch table and strikes up a conversation [with the chatbot]. The candidate picks among several choices for how to start the conversation. The kids respond, and the candidate goes back and forth with them a few times before the exchange comes to a close.

If you told me I had to use a chatbot to help teachers develop social competencies with students, I doubt I’d do better than this right here. But I’d also tell you that the difference between texting a chatbot and chatting with a group of real-life kids is comparable to moving from a heavy bag straight into the ring with Tyson in his prime.

The speed of those interactions, the ways students complete one another’s thoughts, complete your thoughts, rebut your thoughts before you’ve finished them, is utterly unlike a chatbot happily taking turns, waiting patiently for your response, etc. That’s before any questions of how well the chatbot reproduces the personality and culture of the students.

¶ I have been mostly content to let composition teachers [waves hands] figure this whole thing out. You know, to decide if they’re going back to handwritten blue books or if they’re going to insist on a replay of student writing history or something else entirely. I’m only going to wade into that discourse long enough to say, it seems like a rough situation when you have an increasing number of kids using AI to write essays that an increasing number of teachers are using AI to evaluate?

¶ See also: increasing numbers of researchers pasting text directly from LLMs into their papers which are then reviewed by people pasting text directly from LLMs into their reviews. We are using this technology to convert what used to be (occasionally) substantial negotiations of ideas into perfunctory social rituals like saying “bless you” after someone sneezes.

¶ If you are interested in someone whose outlook on generative AI is even more bleak than my own, may I recommend Ed Zitron and his latest article, “Have We Reached Peak AI?”

¶ Ethan Mollick pasting an image of a nuclear reactor into various LLMs and asking them what he should do if he accidentally pushed the meltdown button is, to me, very funny.

¶ I think Jeremy Roschelle probably has it right with this diagram, though what he’s describing falls far, far short of most of the financial and educational promises currently floating through the discourse.

¶ The SAT goes fully digital.

The test is also one hour shorter (down from three hours), has shorter reading passages and uses digital tools, like a highlighter, a graphing calculator and a bookmark to go back to skipped questions.

Shout out to my friends and colleagues at Desmos Studio PBC for being that graphing calculator.

¶ Talent Lab Solutions describes a FAFSA process that is in total disarray this year. Two notes here. One, extra filing requirements mean that fewer people will get aid, including and especially people who should receive it. It is hard not to see this as a design feature of the way we administer aid in the United States. Two, we fully fund every student’s public education from grades K-12 in the US. We do that because we believe (I think rightly) that educating people returns benefits to all of us, not just the people getting educated. So what’s special about grades 13-16? What principled reason is there to not fully fund public higher education for every student?