This post was originally published on this site.
Peering into the crystal ball to see how artificial intelligence will change education can be a tricky exercise. While some previous unknowns are taking shape, such as how much teachers will embrace the time-saving power of the technology, other aspects remain foggy. Among them: How will the technology change the way schools teach K-12 fundamentals like math? And, is a predictive technology even a good fit for math instruction?
One person in a position to bring some clarity to these questions is Bob Hughes, the director of education for the Gates Foundation, which is investing more than $1 billion in math education, with a big focus on AI. An integral part of Hughes’s job is keeping his fingers on the pulse of innovations in the K-12 sector that might improve educational outcomes for kids, and AI is driving a lot of those innovations right now. (Editorial Projects in Education, the publisher of Education Week, receives support from the foundation for its coverage of math instruction. The media organization retains sole editorial control over its articles.)
From AI-powered teaching assistants and tutors to programs that can analyze teachers’ lectures and provide feedback on their delivery, Hughes says there are many emerging products that are poised to change how schools teach math and students learn it.
Here’s a look at the developments he’s keeping a close eye on, and how the K-12 education sector is adapting.
This conversation has been edited for length and clarity.
A criticism of AI tools is that they’re often not accurate. Will that make math teachers hesitant to adopt them?
We’re still in the really early days, right? We’re two years into ChatGPT. Math is an incredibly hard thing to get right in AI. The early [chatbots] had lots of hallucinations. They were not necessarily accurate. There were a lot of those early stories comparing ChatGPT to a 7-year-old—can it to do the fundamentals in math?
We’ve made real progress in math in both large and small [language] models so that [chatbots] are increasingly accurate. [They’re] doing very well on international and national benchmarks of accuracy. We’re entering a new stage where we can say with greater confidence that the math that is being done [by AI chatbots] is accurate.
Once those accuracy concerns are addressed, what are the next steps?
Accuracy is just the beginning. Then you have to be thinking through what are the use-cases for AI? How do I as a teacher integrate and use these tools to create learning experiences that are engaging and motivating for students and, ultimately, lead to not only greater proficiency, but persistence in math?
Because one of the challenges we find is that students get the fundamentals in math, and then suddenly they hit a wall when they get to algebra. It becomes a highly emotional subject. In the worst cases, they start to internalize their own sense of intelligence based on whether they’re good at math or not.
How does that AI enable a young person to have more opportunity to do math that is relevant and meaningful to them? And ultimately, how do we build and promote academic achievement in ways that we haven’t been able to scale yet? And we’re really at the beginning stages of that work.
How can AI be used to motivate and assess students in more meaningful ways?
A second purpose is to really figure out how we motivate and engage young people in things that they might not be excited about. So, giving opportunities to think about real-world problems [and focus on] things you care about. If I’m a baseball fan, [personalizing math questions to ask] what do fractions and decimals look like in a baseball context?
And then beyond that, there’s an opportunity for diagnostics and assessment. We’ve been so tied to the multiple-choice test. We’re in early days, but what would assessment look like if we started to give open-ended questions that we could actually grade them in ways that were comparable? What would an assessment look like if it were group work where we could track the individual contributions of every student and then build on that?
We’re at the beginning, and the changes are going continue to be fast and furious.
What do students need to learn in school given the capabilities of artificial intelligence?
I don’t think there’s an easy answer to that. People want to say, ‘oh, you’re not going to need fundamental skills because you’ll just write a prompt [for an AI tool] and isn’t that great?’ Yeah, that’s true. But if you don’t have number sets, if you can’t recognize patterns, what’s the prompt you’re going to write?
I was talking to an AI expert a couple weeks ago, and they were saying, “We’re really excited [students] are learning coding in school, but the reality is a lot of their skills will be outdated the first day they start those programs.” So, what we really need are kids who understand, deeply, math and the ability to flex as we learn more and as the technology evolves and as the skills necessary to master where we are increase in complexity.
So, I don’t think there’s a simple answer there, but hey, this is education. When is there a simple answer?