Episode 101: Thoughts on AI and Student Learning

Intersection of AI and learning

0:00Welcome to the Learning Scientist podcast, a podcast for teachers, students, and parents about evidence-based practice and learning. Hi, everyone, and welcome back to another episode of the Learning Scientist podcast. My name is Dr. Cindy Neeple. I'm a director of learning services at St. Louis University School of Medicine. And today I want to have a little conversation with you to chat about the intersection of AI and learning. Now I know that this is

0:35a hot topic and a lot of people have some pretty strong opinions. The discussion that I want to have with you today is going to be somewhat limited. I want to talk about some sort of framework of when AI is appropriate for learning and when it is not. And more specifically about how often we should really be using it. So as we get started here, what I want to say and what I want to really bring to the forefront is the fact that in order to understand whether

1:11and how AI is going to promote or hurt student learning really requires a deep understanding of student learning. Anyone who is just throwing things out there to say, oh yeah, just use AI, it'll be great, is fundamentally missing the piece that is required, which is knowing the mechanisms

1:42for how learning takes place. And there's really no way to have this conversation without that fundamental understanding. So if you are going to talks or being told to use resources in your classroom or using resources without that foundational knowledge, it's really time to take a step back and say, wait, I need to know how to know when AI might be good or bad for learning. So we're actually

Prior knowledge and working memory

2:19going to start here with a little conversation about student learning. Now, of course, we have the Learning Scientists blog and, you know, however many podcast episodes out there at this point that all are about strategies for student learning. So in the less than 20 minutes I have with you today, I'm certainly not going to get through everything. But there are a couple of fundamental things that you need to understand and recognize in order for this conversation to make sense. And so one of those

2:53things is the fact that knowing things is really important. It's not as simple as just being able to look up information, to have information readily at our fingertips. It's important to actually have the information in our heads, in our long-term memory. And now, why is that? The reason behind this has to do with how our working memory processes information. And so if we can process information by chunking it,

3:30instead of hearing small, discrete facts, if we can somehow chunk that information into one thing, then we can hold on to more and process more. Now, that's very abstract. So let me give you a couple examples. So one example of this that I've used many times is the idea of baseball. And if you listen to a little passage about baseball, and you know nothing about baseball, you might hear there's a runner on

4:05first, and it's the seventh inning, and the team's name is the Lynxes. But if you do know something about baseball, you might listen to the passage and say, oh, that's a passage about a double play. And all of that knowledge, that prior experience that allows you to recognize a double play, also means that you can hold a lot of information just knowing that one thing. Just knowing double play will help you get back

4:35to a lot of the other details that someone without that prior knowledge simply can't do. Another great example of this is a classic study in chess. And in this study, what they did was they looked at chess experts and chess novices. And when experts look at a board of chess, unsurprisingly, they don't just see individual pieces like I do. Instead, they see a play. They see what moves should come next. They see, I don't

5:10know, the queen's gambit, whatever that means. What they don't see is the individual pieces. So I instead see, I don't know, look, there's a pawn and something that looks like a castle and maybe something that looks like a horse. Because I am a novice when it comes to chess. And so the experts are able to see one move, one play, and that allows them to already know where many pieces are on the board. Okay? So when we have

5:40knowledge inside of our heads, it allows us to not only hold on to information a little bit better, but it helps us to be able to reason and problem solve because we can hold on to more information at any given time. So this is fundamental to understanding why AI might hurt learning. Because any technique that enriches our knowledge helps us with problem solving and critical thinking and

6:13helping us to acquire new knowledge rapidly and effectively, building a really dense knowledge network where all of these different concepts are richly united is one of the key fundamentals to successful learning. So when we think about AI and learning, that's one of the pieces. Then we have to think about how do we do that? How does one build a knowledge network that is dense and leads to better

6:48understanding and better problem solving? And goodness knows there are so many answers to that question. We have retrieval practice and concrete examples and elaboration and lots and lots of strategies that we talk about on our website, on the blog, and on this podcast. But I think what's maybe most important for us to think about in this time is that all of this building of prior knowledge requires

7:24effortful thought. So in the world of transfer of knowledge, what we know is that in order for someone to be able to recognize that, hey, something I learned before can be used here, and then to do that successfully, they need practice doing it, right? So we need to practice what we preach. We need whatever it is that you want to successfully be able to do later on. We want to give individuals

7:55practice doing that during learning. So if we want them to successfully retrieve and use information, then while they're learning, they should be retrieving and using information. The closer we can make learning activities to the activities where we actually want people to be using what they have learned, the more similar those things can be, the more successful learning can be. But there has to be

8:26this foundational knowledge there first. So to get that foundational knowledge in requires some amount of

Cognition and offloading effort

8:33mental activity. And so one of the phrases that I like to use when thinking about AI and learning is that the idea that whoever is doing the thinking is doing the learning. So if AI is the one doing the thinking, then AI is the one doing the learning. And we don't necessarily want that. There are times in our lives where it's perfectly fine to have AI bypass something for us. So if I am trying to figure out how to set up an Excel spreadsheet to do some

9:10complicated task, and I'm struggling to figure out how to do that, and I ask AI to help me out, I think that's okay. And I think that's okay because I'm not expecting to be able to do that task again more effectively in the future, the figuring out how to do it part. But if I ask AI to help me to develop some new materials, if I am constantly relying on some sort of app to tell me how to get around my city, to show me the maps to

9:49my city, if I am offloading my cognition to an external device, then I'm not getting better at it. That's really the key fundamental here, is if we want students to be able to do something, we need students to do the thing. If we want them to get really good at using AI to do the thing, fine. I mean, and maybe that is a reasonable learning objective in some spaces, but it's really

10:20critical for us to take this moment to stop and think about the learning objectives in our classroom. That is what it comes down to. What knowledge and skills do you want students to have? Because if what you're teaching them is how to essentially Google things, how to just search for information, then what they're getting better at is searching for information. If what you are doing is having them use AI to produce something, then they're getting better at using AI to produce

10:56something. They're not getting better at producing it themselves. So if it is a task that you don't care the rest of their lives, they can rely on AI to do the thing, okay, fine. But if it's something that realistically they need to be able to do without the support of AI at some point, then they need to practice doing it. And even more fundamentally and even more critically, if you want them to be able to take that information and ever be able to reason with it, to problem solve with it, they need that

11:33knowledge in their heads. They need not to just be looking it up because it's going to fundamentally qualitatively change the way that they think about material, the way they process it, and how much they're able to do with that, how creatively they can problem solve. And fundamentally, I think that the goal of education is usually those things. We don't want to just put facts in students' heads. We want

12:03them to be able to take that, to interact with the world, to solve the next generation's problems. And if that's the case, then we need students to do some mental wrestling. Learning is hard work. You are changing the neural structures in your head. That shouldn't be super easy to do. It should be a challenge. We should have some desirable difficulty there. And so what does that really mean for

Classroom restrictions and educator use

12:33educators? What does this application of this look like? Well, this means that what, as an educator, you should be doing with AI is, number one, probably restricting your students from using it most of the time because students don't understand human learning. Students have not gone to school for teaching or education or cognitive psychology. And what is intuitive about learning isn't always what is

13:07actually true of learning. We have some great blogs about that I'll put in the show notes. But that means that students cannot be relied upon to always make good choices about their own learning. We know that from years of research looking at students' study strategies. If left to their own devices, students take the path of least resistance. They cram instead of spacing. They reread instead of retrieving. And I

13:40could go on a whole tangent about why that is. I don't necessarily blame students for that. It's a metacognitive error that's happening. But the same metacognitive error is going to happen with AI. It feels like they are learning something like they are learning something, but they're not. So first, I would mostly restrict students' use of AI unless your learning objective is how to effectively prompt AI or how to

14:10use AI to do specific tasks. But otherwise, I would restrict it if your goal is learning. Second is that educators can use AI for some things, I think. Fundamentally, you are the master of your own domain. You understand your students and your material better than I ever could. And because of that, I would argue that you will likely create far better materials than AI will because it doesn't

14:47understand the material the way that you do. So trust yourself, you have a place in a world of AI, a very important place. There's a lot that you can do that AI cannot because AI doesn't understand your context. That said, if you can effectively prompt, review, edit, and work with AI, you can use it to try to create some materials that if you were creating from scratch you might really struggle with. So to do a

15:23first pass of a quiz or another retrieval activity, get some ideas perhaps for some activities, and then take that and change it and update it and make it yours. I think that's reasonable. With my own students who are medical students, they are pretty highly motivated students. We often talk about the fact that the materials that exist for them out in the world for their self-regulated learning are often,

15:54you know, charts and things for them to just memorize, a diagram or an infographic of anatomical parts. And what they can use AI for is to create blank versions of that. So they're creating their own retrieval mechanisms. And that's fine. They can go to AI and treat it as a study buddy or a mentor where they can say, here's my understanding of this concept. Tell me what I'm getting wrong and tell me if

16:25there are things that I should add to my understanding. There are some things that can be done to be useful in the absence of an alternative. But I think that we just have to be very, very careful that we're using it in a way that enhances student learning and doesn't bypass student learning. So again, it fundamentally comes down to the idea that if AI is doing the learning, then you are not. And so if AI is doing the thinking, AI is doing the learning, I hope that this is a somewhat

17:01useful conversation and framework for starting to think a little bit more about how and when to use AI in the classroom. As I said, there's a much larger conversation here about ethical use and plagiarism and more tech questions of what AI is really doing to our society. I'm not getting into all of that today. I am hoping really just to talk to the educators out there so that you have a reasonable

17:34idea of what AI can and cannot do for you in your classroom. So I hope this was an interesting conversation. Conversation? Look, I'm just sitting here talking at you. But I hope that this is useful for you. If you have any thoughts or ideas, please always feel free to get in touch with us. I'll put some stuff in the show notes about what I talked about today, some of the research studies and links to other blog posts that are related. I will look forward to chatting with you on the next episode

18:10of the Learning Scientist podcast. This episode is funded by listeners like you. To support our work and gain access to exclusive content, visit our Patreon page at www.patreon.com slash learningscientists. www.patreon.com slash