Okay, so how can AI be used to improve education?

I’m pessimistic about the AI hype but very optimistic about AI as a tool. Is AI going to “replace” teachers or somehow do our thinking for us? Not at all likely. But will AI prove to be a powerful new tool for expanding humans’ capacities and capabilities? I suspect so.

Part of the problem here is that the term “AI” is being used to liberally. There’s a big difference between ChatGPT and the machine learning analysis techniques it uses. Do I trust ChatGPT to give me reliable, accurate answers to complex questions? Not at all. Is there value in the machine learning analysis techniques that ChatGPT uses? Absolutely. Notice the difference there - when it comes to the machine learning techniques used by ChatGPT and other technologies, as Eric Clapton once sang, it’s in the way that you use it.

A good example of a solid plan for using machine learning in education has been advanced by Jarvela et al. (2023). They argued for Hybrid-Human AI (i.e., machine learning) models where AI is used to do the things humans are bad at: organizing and analyzing huge volumes of data.

See, Jarvela and colleagues study how students collaborate and, more specifically, how they adjust their work when things go awry (i.e., socially shared regulation of learning, for those of you interested). That’s really tough to observe - you might see a group member suddenly exclaim they figured something out or another person suddenly start yelling at a team member, but all the thinking the preceded those actions is mostly hidden. How can we figure out what caused those actions, so we can intervene to improve their outcomes?

Well, there’s been real improvement in eye tracking, emotion-detection (e.g., tracking facial expressions, heart rate, etc.), and other types of fine-grained data collection tools that can detect human thinking, but the problem is they produce a TON of data. Those data are very difficult for humans to understand and analyze. On the other hand, AI is well-equipped to analyze those data, with human supervision.

So, there’s the Hybrid part. Humans theorize about what data matters, they gather that data, have AI analyze it, and then the humans go, “Okay, this part is useful. This part is not. Let’s analyze again focusing on these things…etc.” The AI is used as a (powerful) tool to help humans do the work they are ill-equipped for (e.g., analyzing large volumes of dense data) so humans can complete and expand the work they are well-equipped for (e.g., theorizing, understanding human interactions, intervening).

Now, what Jarvela et al. call “AI” I would call “machine learning analysis techniques” - it’s not like they are using ChatGPT in their work. And that’s the difference. They aren’t asking AI to think for them, instead, they are using machine learning analysis techniques, in very narrow and theory-guided ways, to think broader and better, themselves. And that’s the promise of AI in education.