Two of Your Predictions for 2024 and Why They Are Probably Not Correct
Two predictions from prominent edtech figures, and a teacher survey that says, “Hold on now.”
I am a sucker for every part of the new year. Yes, it’s just another day, but I track down every year-in-review article I can find and roll around in them like a puppy in a pile of laundry. I maintain my own list of goals every year and use the new year to take stock of my hits (I ran a bunch of miles consecutively and didn’t throw up during, after, or before) and misses (my goal of a perfect Wordle streak in 2023 ended mid-January).
I am also an eager observer of the edtech predictions market so I wanted to share and comment on two predictions for 2024 that will be relevant to many of you, and then share a teacher survey that calls both into question.
Sal Khan: Generative AI will perform 90% of teachers’ administrative tasks.
First, Sal Khan predicted that generative AI tools will eliminate 90% of a teacher’s administrative work. I have been critical in the past of the edu-futurist’s tendency to make predictions that are airy enough to capture the imagination of the business class but which are also totally unfalsifiable. So credit to Khan for a prediction that is at least this specific.
For a little context on Khan’s prediction, in Merrimack College’s survey of the teaching profession last year, they found that teachers spend less than half of their working hours teaching students and 31% of teachers wish they could spend less time on general administrative work.
Those same teachers report spending 3 hours per week in the category of “general administrative work.” Reclaiming those hours for tasks teachers enjoy more (and which use more of their unique value) could result in a significant improvement to a teacher’s job satisfaction and I wish generative AI all the best here.
Jomayra Herrera: AI tutors will become great.
Jomayra Herrera is a partner at Reach Capital, an education venture capital investment firm, and predicts that in 2024:
We will see a big evolution in the Al tutor landscape. Al tutors will go from being better web search experiences to actually being a great tutor, which means they personalize learning experiences to students. They are customized to the content that students are actually learning in the classroom. They also can predict misconceptions that students will have and actually reflect what great tutors do.
We should wish generative AI tutors all the best here as well. Given the grim results of large-scale assessments we have seen since the start of the pandemic, we should take support wherever we can find it, especially if that support scales digitally.
Me: None of this will happen in 2024.
I argued much earlier in the hype cycle that, over the next five years, generative AI will product modest quality-of-life improvements for teachers and students and that it will alter national trends in student learning and teacher retention in ways you will need an electron microscope to detect.
This prediction looks sturdier every time we see new surveys of teacher usage. Generative AI shows no sign today of making the kind of impact Herrera and Khan and so many others predict—not because I say so, but because teachers say so. Teachers report they aren’t using generative AI and don’t plan to use it.
The results of this Education Week survey would ideally provoke a lot of self-reflection among edtech operators, investors, philanthropists, and boosters. One could probably make a glass-half-full interpretation here. (“31% of teachers are using GAI a little or some!”) But after a year of the most breathless marketing of any consumer product in my lifetime:
only 2% of teachers use generative AI a lot,
22% of teachers have no plans to use it this school year,
37% have no plans to use it ever.
My question for anyone making the kinds of predictions that Khan and Herrera are making is, what is going to change in the next 12 months in order to fulfill those predictions? More speed? Greater accuracy? Fewer hallucinations? More marketing? More teacher training? Is it, as
has suggested, that too many teachers are stuck using GPT-3.5 instead of GPT-4 and haven’t experienced the true power of generative AI?None of those explanations are particularly convincing to me. I don’t think teachers are waiting for technical benchmarks to change, for example. Show me a fifth-grade teacher who is saying, “Okay GPT-4 is smart enough to match the 89th percentile of SAT test takers. No, thank you, but once it clears the 95th percentile, I’m all in. Given these survey results, it seems more likely to me that generative AI chatbots are the wrong thing entirely, that teachers have sized up generative AI chatbots and decided they have the wrong shape for the work of teaching.
These teachers’ most common reason for not using generative AI is that it doesn’t rate for them.
If any one thing can be said about all the practitioners in a field as vast as teaching, it is that they are ruthlessly pragmatic with their time. They can afford to be nothing less. If teachers thought generative AI had a realistic shot of saving them time, we would have seen different survey results here.
For those with ears to hear, teachers are telling you, “this isn’t it.” Teachers are telling you that the value of generative AI chatbots that is so plain to you is not plain to them.
Certainly, I would like my predictions to be right instead of wrong, but that isn’t why I use this platform to occasionally harangue edtech operators, investors, philanthropists, and boosters. Rather, I would like more people to make more tools that meet more of the real, rather than imagined, needs of teachers and learners.
Last year would have been a great year for edtech operators, investors, philanthropists, and boosters to try to understand the real needs of teachers, to understand a few more of the many facets of the student learning experience, to understand how teachers actually spend their preparation time if they have any, to understand the kinds of administrative work they assign to a teaching assistant if they have one. Last year would have been a great year for them to understand the profession they seek to transform, but this year is also good and it is just beginning.
Odds & Ends
If you’d like to learn what specific administrative tasks teachers would like to shove off their plate, and whether they believe generative AI can take on that work, check this thread on Twitter. (Special shout out to the person asking if generative AI can watch the mandatory blood borne pathogen training video on his behalf.)
If Sal Khan’s prediction proves out, it will be in some huge part because of the efforts of Adeel Khan and his team at MagicSchool.ai, a “teacher copilot” tool which announced last week they have registered one million users. They’re also now throttling free usage more tightly and directing teachers to a $12 per month subscription tier. I think any one of several numbers, including a) their conversion rate from free to paid usage, b) their monthly active usage, or c) their weekly user retention rate, would do a lot to help us understand the state of teacher usage of generative AI. Much more, anyway, than knowing that one million people clicked “Sign up with Google” once. I wonder why Khan Academy and MagicSchool haven’t disclosed those figures.
Natasha Singer from the New York Times is back on the AI in education beat with an article titled, “Will Chatbots Teach Your Children?” My answer there is, “yes,” although the questions we’re actually bouncing around the playground here are “which ones?” and “how often?” and “how much?” and “to what effect?” Singer’s article recapitulates much of Audrey Watters’ history of teaching machines. and I think she applies some necessary skepticism to the claims of the creators of these modern iterations.
Speaking of predictions, Inside Higher Ed asked a bunch of experts if they thought their predictions about AI from last year were proven correct a year later. Basically everybody thinks they nailed it!
Julia Freeland Fisher of the Christensen Institute describes 3 education innovations to watch in 2024. This one right here gets an 👀 from me: “AI-powered tools to multiply, rather than replace, human conversations.”
You shouldn’t need me to recommend Jill Barshay’s work to you at this point but you should read her recent careful analysis of the tutoring space.
Appreciated this breakdown. The sanguine expectations of people not in classrooms reminds me of the four-square meme "What my friends think I do; What my parents think I do; What society thinks I do; What I actually do." What I actually do is open up a world to students, help motivate them, grab "the book" for a reluctant reader and put it in their hands, make students feel seen/heard, create a safe space, help them through hard conversations (for example in Montana, addressing anti-Indiginous bias). I know AI can help me do my job, and I know there is inefficiency in education, but the idea that schooling is on the cusp of a breakthrough... I'll believe it once school bandwidth and equitable access to tech becomes a reality. I see a lot of room for the "Matthew effect" in children's education as AI takes off.
Recently I had dinner with my former assistant teacher who had moved to New Hampshire. She had decided to become a substitute teacher because the district she now lives in is in desperate need. So she has been substituting in a school walking distance from her house that was declared a Blue Ribbon School. The school has an overflow of funds. Hanging on the wall of each classroom is a giant TV type screen. She was working with first grade. The sub notes told her to turn on the TV thing for it to read the days book choice to the children. So she did. Barely any of the children watched while the TV turned the pages and read the book. They squirmed and talked and poked each other but few engaged. Later in the day she READ a book to the children. She is a master of voices when she reads and the children were completely enthralled. They threw question at her about the reading. Since the story was placed in the South one asked if she could read in a Southern accent so she did. Not one child wiggled, rolled, poked, or talked to a friend. It was the TEACHER who is a master at reading a story. This is not limited to reading. I have had teachers tell me their children are not interested in "calendar" time until they have watched me present the "calendar" to the class and see the magic of engagement. All the AI in the world cannot do this.