The Tortoise and the Hare Contemplate AI in Education (Part 1)

By Jeremy Roschelle , Kip Glazer , and Judi Fusco

Key Points:

• Artificial intelligence in education can seem like a race, yet a slow, intentional pace can be beneficial.
• Clarity about the goal for using AI in education is important and takes time.
• Research can guide effective use, but also takes time.

One morning, the Hare burst into the teacher lounge full of excitement and FOMO (“fear of missing out”) regarding the use of AI in education. Hare, like her good friend Tortoise, was one of the district’s educational technology coaches. Students and teachers in their district have been exploring AI in their cautiously moving district. The morning’s Hare-brained idea was to throw caution to the wind and step on the accelerator. Tortoise, however, did not see the adoption of AI in education as a race, and certainly not a race to be won anytime soon. AI in education is important, perhaps even transformational—but not in a day, thought the Tortoise. Let’s listen in to their arguments.?

Our tale is presented in two parts; the first half of the race is presented below.

A Race to Where?

Hare: Torty, even you have to agree with me on this one. AI is happening everywhere in our schools, our community, the world. We’ve got to move fast. Our students and teachers can’t be left behind in this transformation. Let’s go, go, go!?

Tortoise: Go where, my friend? What do you see as the greatest benefit of AI for us?

Hare: Well, our teachers could be much more quickly writing better lesson plans. And why should students in district-over-there have super great lesson plans that take less time, while teachers in our district are still using lesson plans out of a book or ones they painstakingly craft themselves? More, better, faster—that’s what we want for our district.

Tortoise: Hang on just a second, Hopster. Have you seen how many errors, like wrong facts and made-up references, show up in AI-generated materials? My legs are not long enough for a race to go wrong. I say more and faster is not always better.?

Hare: But teachers can check those facts, and AI is getting better everyday. They can switch LLMs and chatbots to better ones, and we can read the reviews on social media to see what teachers-over-there are using. If we work, work, work, then we can go faster, faster, faster. I can’t bear it—let’s go, please!

Tortoise: Sounds pretty hairy to me! I hear you agreeing that generative AI isn’t quite ready for prime time, and that our teachers would have to work harder right now to overcome weaknesses… and that’s not all ….

Hare: What's a little work? Who's counting? We can’t lose the race!

Tortoise: What happens when there’s more stuff being produced by teachers and students all the time? Who is going to look at all that stuff? Who is going to check and approve? Are we really going to save our teachers’ time?

On the benefits

Hare: Oh Torty, you can be such a downer! Don’t you see benefits?

Tortoise: Well, my speedy friend, I actually do see benefits and I’m not saying “no.”

Hare: Tell me!

Tortoise: For example, newer AI models are multimodal—inputting not just text, but also voice and images, and outputting in multiple media at once —and that has profound uses to make learning opportunities more accessible, like AI could be the biggest thing for Universal Design for Learning ever. And that’s a scientific theory that can win the long race to improve education, but it will take time to develop materials and check them.

Hare: That’s great—what else?

Tortoise: You know, speaking of the benefit of slow and gradual, did you know there has been research on AI in education for almost 50 years?

Hare: That’s amazing! How did those researchers go back to the future?

Tortoise: Ha! For example, researchers have been studying intelligent tutoring systems for a long time, and there’s good evidence that can help students to learn by providing more and better feedback as students work on problem solving.

Hare: Fantastic. You know, I was showing our teachers eight brand new tutoring products that just came out. Great to know the research says those are gonna work.

Tortoise: Not so fast, Hare. The research identified the kind of components that enable a tutor to work, and I’m not so sure these newer products have the right stuff. For example, do they have a model of what the student already knows? A learning progression from prerequisite to eventual concepts??

Hare: Geez, I didn’t think to ask. It's all so POWERFUL, who needs components?

Tortoise: Yes, new generative AI is powerful, but it also may give wrong answers and be missing things that are needed to enable students to learn. It wasn’t built for the purpose of teaching students, and it was not trained on data about the learning process.

Hare: (smirking) You mean finding data to support student learning by scanning public discussion forums on the internet isn’t all it is cracked up to be??

Tortoise: (retreats into shell) Save me, save me, save me….

Hare: I’m off to the races!

Tortoise: (pokes out) I’ll see you at a finish line where we’ve had time to work out safe, secure, and responsible uses of AI in our district.

Read Part 2

Learn more about Digital Promise’s vision for AI in education.

Jeremy Roschelle is co-executive director of Learning Sciences Research at Digital Promise.

Kip Glazer is Principal at Mountain View High School in Mountain View, California.

Judi Fusco is director of Emerging Technologies and Learning Sciences at Digital Promise.