S4xE18: AI Has Not Made Learning Obsolete
Episode Summary
In this episode, we talk about how AI has not made learning obsolete with Dr. Neil Brown, Senior Research Fellow at King’s College London in the UK. We start with Neil explaining how learning works by using prior knowledge to organize and make sense of new knowledge more easily. Without chunking that prior knowledge, it is much harder to understand and use more advanced knowledge, even when AI makes all knowledge available. We then discuss how we should use AI in learning by focusing more on the learning journey than the product, rethinking feedback, encouraging productive struggle, and doing more research.
You can also download this episode directly.
Episode Notes
King’s College London in the UK
Neil C. C. Brown, Felienne F. J. Hermans, and Lauren E. Margulieux. 2023. 10 Things Software Developers Should Learn about Learning. Commun. ACM 67, 1 (January 2024), 78–87. https://doi.org/10.1145/3584859
Sverrir Thorgeirsson, Theo B. Weidmann, and Zhendong Su. 2026. Computer Science Achievement and Writing Skills Predict Vibe Coding Proficiency. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI ‘26). Association for Computing Machinery, New York, NY, USA, Article 947, 1–17. https://doi.org/10.1145/3772318.3791666
Irene Hou, Owen Man, Kate Hamilton, Srishty Muthusekaran, Jeffin Johnykutty, Leili Zadeh, and Stephen MacNeil. 2025. ‘All Roads Lead to ChatGPT’: How Generative AI is Eroding Social Interactions and Student Learning Communities. In Proceedings of the 30th ACM Conference on Innovation and Technology in Computer Science Education V. 1 (ITiCSE 2025). Association for Computing Machinery, New York, NY, USA, 79–85. https://doi.org/10.1145/3724363.3729024
S4xE14: GenAI’s Impact on Student Help Seeking and More
Transcript
Kristin (00:02) Hello and welcome to the CS-Ed Podcast, a podcast where we talk about teaching computer science with computer science educators. I am your host, Kristin Stephens-Martinez, an Associate Professor of the Practice at Duke University. Joining me today is Dr. Neil Brown, Senior Research Fellow at King’s College London in the UK. Neil, thank you so much for coming on the podcast.
Neil (00:22) Thank you for inviting me.
Kristin (00:23) So, Neil, first I’d like to ask what you get up to outside of work, just to normalize that we are all whole people.
Neil (00:29) So I’m an enthusiastic footballer and a reluctant gardener. So on the football side, I organize a couple of games a week. And last year I managed to organize a couple of games at some CSed conferences, which was great. And then on the gardening side, basically I lived in apartments for many, many years and then sort of bought a house a couple of years ago, you know, the dream, have a house, and the house comes with a garden. And it turns out. It’s a lot of work.
Kristin (01:01) I keep a small strawberry patch in my backyard. So, is this a flower garden or is this like a vegetable fruit garden?
Neil (01:11) It was kind of just more of a jungle when we moved in because it had been the house had sat empty and it was completely overgrown and even two years in, I’m still working on brambles and trying to like get on top of everything. So it’s not even kind of really got a good shape yet. It’s just, just clearing, and it keeps growing back. It’s unfair.
Kristin (01:30) [laugh] That’s what plants do after all. The point is the journey, especially when it comes to the garden, I think.
Neil (01:37) It certainly gets me outside and off the screen.
Kristin (01:38) Yes, exactly!
Neil (01:39) So, I view that as a positive.
Kristin (01:42) That’s the point. That’s the point to get outside. But speaking of that, we’re gonna talk about completely being probably inside. Though I guess you could go outside. And to set the stage, I read an article that you wrote with Lauren Margulieux and Felienne. I’m gonna forget Felienne’s last name.
Neil (02:02) Felienne Hermans.
Kristin (02:04) Felienne Hermans, thank you. And we’ll definitely have a link to it in the show notes. Called Ten Ten Things that software engineers should know about learning. And I’ve read the sixth point, which was the internet did not make learning obsolete. And I loved it so much that I was like, Neil, you have to come on the podcast and talk about it more, because we’re going to talk about how the internet and AI have not made learning obsolete. But let’s start with the basics. What is learning to help us better understand why the internet and AI have not made learning obsolete?
Neil (02:35) Okay, so I mean, there’s obviously all kinds of components of learning. Today I’m gonna mainly talk about the cognitive side. So, cognitive science says that we have a fixed working memory capacity. You might have heard of this kind of five plus or minus two items that you can hold in memory, that kind of rule. And if you like computing analogies, it’s a bit like kind of registers in your computer. There’s a fixed number of, like, storage locations that you’ve got there, and it can’t ever be expanded over your lifetime.
And so knowing that, it’s then really odd in a way that we even can learn because if your capacity for like how much you can keep, you know, in your working memory is fixed, then how do you ever get better at anything? And so the way it works is that, broadly speaking, like this is kind of obviously abstracting away some of the detail and finer points, but those items that you hold in what’s called working memory are kind of pointers to things that you hold in long-term memory.
And the way that we actually sort of learn or get better at things is that we increase the size and complexity of the things that we hold in long-term memory so that the thing we’re pointing to becomes kind of larger and larger. So I’m gonna give an example to try and sort of explain this.
Kristin (03:59) Okay.
Neil (04:00) Imagine that you want to read an academic paper. And if you give an academic paper to like, you know, a novice reader, like a four-year-old or something, five-year-old, then they’re going to start by trying to read letter by letter because that’s the kind of thing they’re technically called chunks. This is actually the technical term that you’ve got in long-term memory. They know what each letter is, perhaps, or they’re learning that. So they read letter by letter and then try and assemble it into words.
If you give it to someone more like, you know, say sort of a 12-year-old or something, then they’re instead generally reading by words. They know words like “the”, or “method” or whatever appears. And so they can then kind of go word by word. So they’re working at a kind of higher level of abstraction. They’ve got these larger things, like a word is larger than a letter to hold it in your memory.
But then if you sort of level up again and you become like an academic reading these papers, you start to have kind of, even more abstract knowledge about what it is. So you maybe start reading a paragraph, and you’re like, this is describing the method or this is describing the related work or this is about how they obtained ethical permission from their, you know, participants. So we have these kind of more abstract and larger items that we can store in long-term memory. So even though we’re actually got this fixed working memory capacity, the level that we’re working at gets higher and higher. And that’s how we have sort of increased capacity.
Kristin (05:21) Yeah, I’m just like, I am almost giggling at how you’re trying really hard to use like, what a computer scientist would understand to then try and help them understand how learning works. Cause I feel like for me generally, I balk at that because I’m kind of like metaphors work until they no longer work, and the breakdown potentially makes it much harder to get someone to understand what you’re talking about. But I definitely… We’ll work with our audience here, which are mainly computer scientists who also know how some ideas of how learning works.
So I think given that, I suspect that the next question is how do you create those chunks? They don’t just like form out of the ether or in the matter of your brain, like as we said, abstraction is great, like computer science, like that’s kind of like one of our big jams. But how do we get there? How do we get to those chunks?
Neil (06:18) Yeah, and it is a tricky item. Part of it is obviously, you know, practice engaging in whatever it is. In this case, if we’re talking about reading, obviously just reading a lot of things and having that practice there. For other things, it might be more like a bit of doing so if we’re talking more about programming, for example, you know, you probably want to be writing programs and reading programs and just generally kind of engaging with programming in order to form the chunks. Of course, everybody forms their own chunks. You can’t sort of instill a particular, you know, bit of knowledge or abstraction into someone’s mind directly. All of it comes from, you know, them sort of building their own mental models and kind of finding the chunks that are useful to them.
Of course, there’s an element in sort of formal education of trying to structure it so that that’s encouraged. You know, an obvious thing that we instinctively do with reading is that we start with simple books with short words, you know, and short sentences and not very complicated concepts. And we just intuitively know that, you know, we should be doing that. But in education in general, you know, that’s how we structure our courses. You know, we start with things that are simple and then we increase the level of difficulty of what we’re dealing with so that people have a chance to engage at the sort of shallowest level, but then actually gradually learn to get to a kind of deeper level.
Kristin (07:41) I think, isn’t it also, though, because, like, through the practice, we’re just kind of like seeing it and familiarizing it and kind of being able to wrap our brain around it a lot better. But I think also part of formal education is about providing some structures and providing language for some of those structures. So like the one that always comes to mind for me is like the accumulator pattern. You can teach students the for loop, the if statement, the Boolean, and like these variables and data structure or whatever.
But you can also then explain and teach them the idea of the accumulator pattern, of like you have a for loop and then you have an if statement, and you iterate through the thing. But also providing the language gives them a name for that chunk, which makes the chunk more likely to like exist as part of their mental model. Does that make sense?
Neil (08:25) Yeah, and there’s this idea of what’s called semantic waves, which is that you go from concrete to abstract to concrete and repeatedly switch. And that’s basically because we find abstract alone quite hard to get a grip on. So we like concrete examples. That’s why I gave you the reading example a minute ago when I knew that my explanation was a bit too abstract. But then we want to kind of do some concrete examples to sort of get a grip on it. But then we also need to abstract to say, these things are the same because people don’t always appreciate that.
You’ve done these five programming exercises, and you use the accumulator pattern in all of them. Here’s what it’s called, and here’s the like more abstract knowledge. And then you can kind of go back to, okay, now here’s a new instance. Now you know to kind of apply the accumulator pattern. So you’re forever going from concrete example that people can get to grips with, to abstract concepts that kind of explains and unifies and applies a more sort of general or higher level rule, that then goes back to a concrete example so that they can actually try it out.
Kristin (09:26) Yeah. So, it’s chunking happens through this process of concrete to abstract, but also providing that like these are how these things are connected and these are how things these things can be organized to make it easier for them to get it together.
What you reminded me more of like math education, watching my six and eight-year-olds like do their math homework. I’m like, oh, I can see why it has these like concrete dots and then it goes to this rectangle representing a 10. And the poor eight-year-old was so like bored of this work. He’s like, why do I have to write it this way? I already know how to do like write the numbers. And I’m like, I know, but like you have to draw the dots in the rectangle because that’s the way the teacher’s gonna like check your work. Yeah. I understood the method behind the madness, but it was one of those moments where I’m like, I know you’re past this part of the chunking phase, but like the teacher needs to mark your work and this is just, you just have to do it.
Neil (10:22) Yeah, there is this phenomenon called the expert blind spot, which is where people, you know, who become proficient, you know, they basically, what’s the best way to explain this? So once they become proficient, they can kind of forget what it’s like when they weren’t proficient. So that they are unable to kind of teach people effectively anymore, because they are past that and they just know how to do the next stage. And it then becomes, you forget that all students don’t know that, you know, int is a type or, know, they don’t know that, you know, language is case-sensitive or things like that. It’s kind of easy to just forget what, what you know, in a way, you know, forget that, you know, what, know.
And so teachers often have to kind of explicitly try and overcome that to be very attentive to what could students be, you know, have as a misconception here. How could they be getting it wrong? Even though obviously in my mind it’s firmly embedded and it’s so obvious I don’t even think about it anymore.
Kristin (11:23) Yes. Another part is you’ve lost the context of what it’s like to be a novice that you cannot understand their experience enough to help them understand what they need to understand. And that’s what pedagogical knowledge is, and having that knowledge so that you don’t have that blind spot, even though you don’t experience that stuff anymore.
Neil (11:40) Yeah, and of course, that’s, that’s why peer teaching often works so well, because you’ve got these people who’ve just got out of that zone, and they can more easily help up the people who are kind of immediately behind them, sometimes, than the professor who’s kind of, you know, way past that.
Kristin (11:54) Yes, yes. And that’s why we do peer instructions and that kind of thing. All right. So why do you think AI is not gonna make learning obsolete? Like make it concrete for people.
Neil (12:05) So essentially, you know, the internet and AI both kind of have this thing that all the knowledge of the entire world is basically there, right? It’s, it’s in the computer in front of me, if I just kind of ask in the right way. So you think, okay, well, if, all the knowledge of the world is there, you know, why does it matter who I am? Could I just be kind of like a dumb operator, you know, who’s just sort of sitting there, kind of, you know, asking these questions, but cognitively, the thing is that unless it’s kind of - I’ve got these chunks in my mind, I have to operate at a really basic level.
So let’s imagine I started asking the AI something about biology, which is a topic I know very little about. Then all of a sudden, it’s going to be using these, you know, terms that I sort of very vaguely know, like cell and membrane and, you know, all these other things that I can’t even think of ones I don’t know. But I’m going to then struggle a lot to kind of understand anything about this domain because I just don’t have any of that knowledge in my mind.
So I will struggle to kind of ask useful questions of it or spot, you know, okay, well, tell me a bit more about that because it’s just so overwhelming. So this fact of how these chunks are embedded in our mind means that the use that we can make of the kind of AI or the internet is limited by what we’ve already got in our minds. And in actual fact, there was a paper just out at CHI this year by Thor Gisson et al. We can put the link, I think, in the show notes.
Kristin (13:47) Yes, we will.
Neil (13:48) And if I’m summarizing this reasonably, they were looking at vibe coding performance. So this is, you know, people coding with AI to see how proficient they were. And they found that it was predicted by existing computer science achievement and also by writing skills, which kind of fits with this idea that it matters what you know already.
And there’s a kind of rich get richer effect that people who already know a lot can go very fast with it. But actually people that don’t know much already are going to struggle, you know, more. So it really matters what you’ve already got in your brain. You’re not just kind of a simple operator.
Kristin (14:16) So this had me think in two different directions. I’m gonna go down the first one that’s a little bit earlier in what you said. And if we go back to the reading example, have you ever had a moment where you read a sentence and you’re like, I know the definition of every word in that sentence, but I don’t understand what the sentence is trying to say? And it’s, and it’s a good sentence. It’s not like a, this like novice writer wrote this thing, and it makes no sense. It’s like, the sentence does have meaning, and that is important, but I could not grasp it. Do you think that that’s kind of similar to this idea of like you haven’t chunked things enough to be able to have that kind of reading comprehension?
Neil (14:58) Yeah, and I think especially early academics often experience that when you’re reading academic papers. And that’s again, an expert blind spot thing where we forget it’s a learned skill. But when you sort of give PhD students a paper, you’re sort of like, you know, why does it take you, you know, so long to read it? And of course, the answer is that they’re struggling with the way that the academic papers are written, and they kind of have to, you know, grapple with the sentences at a slightly lower level than when you’ve been reading these things for 20 years. And you can kind of skim a bit because you’re like, you know, that’s not very important. That’s not very important. This is a critical detail. And it’s all bound up with, like, reviewing papers and things like that, that you get more proficient because you know what to look for.
And I’ve definitely had that feeling with some academic papers before where you’re just like, okay, I’m going to have to reread this like three times to try and get what the meaning is here. And that’s kind of magnified if you’re less experienced.
Kristin (15:50) The super dense sentences where you’re like, wow, you packed in a lot of meaning into this sentence because there was a word limit. And you’re trying to go like, okay, what are you trying to say? Yeah.
So the second thing that you said of that more goes with the CHI paper, of vibe coding performance. Do you think, though, even though that they found that vibe coders were better if they had more experience. And basically that meant they had more, like, knowledge in their brain from the chunking and everything. That there is still, though, ways to learn with AI that supports kind of the end goal without forcing people to write a lot of code by hand. Because like, I feel like some people could take the results of that paper and go, like, this is why we have to first make the students write all the code by hand.
Do you think there’s an alternative option though?
Neil (16:47) I suspect so. I mean, I think we’re still early in obviously understanding how to make good uses of this technology.
Kristin (16:54) 100%. We’re so early. Literally the data cannot exist because not enough time has passed to figure out what are the other practices. I just personally find it annoying when people like either say we must all be using AI or we must not, no students must be using AI for whatever reason. And I’m like, where’s the nuance? So what are your thoughts? Where’s the nuance in your thinking about this?
Neil (17:20) Yeah, and so there’s a study that we linked to in the paper that you mentioned earlier, the 10 things paper, which suggested this was about internet searching, but we can sort of say, okay, maybe that’s kind of similar, you know, asking the AI internet searching. And they suggested that essentially, if students immediately resort to searching the internet, they don’t have as kind of, as sort of a stronger learning experience, as if they have to think about it first before, know, sort of struggle with it bit cognitively before kind of, you know, saying, okay, I need to look this up. Because some of that struggle and attempted reasoning seems to kind of, you know, improve your cognition about this particular subject that you’re trying to think about.
And so I think one of my worries is that if it’s all too easy to just, you know, delegate to the AI again, that will miss some of that kind of cognitive struggle of like, okay, you know, I’ll attempt it first. I mean, that’s something that teachers have done for a long time, right? They say, you know, don’t immediately ask the teacher first think for yourself, then ask a neighbor, then ask a teacher. You know, it’s quite a common teaching pattern, isn’t it? And I think the challenge is, you know, when the AI prompt is just sitting there ready for you to use, will we have the discipline to still try ourselves first and then try the AI after that?
Kristin (18:43) Yeah, I think like the metaphor that I have used is like your brain is not a muscle, but it is useful in this context, I think, where the reason why you need to want to think first is that your brain is lazy, just like you are lazy. And so if you force it to go through this effort of trying to think of something or trying to figure something out before looking it up and basically like preventing it from like exerting any effort, it’s gonna try and find a more efficient way to do that thing that you kind of are forcing it to do because it’s lazy. And that actually literally is what learning is. Your brain trying to be more efficient so it can be more lazy at things. Like that’s why you chunk, why your brain is chunking, because it wants to be lazy and it’s learning these things so it doesn’t have to exert as much effort next time. And I don’t know if that’s helpful for my students, but I feel like some students are like, now I get why this matters.
Neil (19:48) Yeah.
Kristin (19:49) But the cost. The cost to getting help of whatever kind has like dropped to the floor with AI. And so it’s like the ease of it now has shifted from in the past where like still searching the internet was a little bit costly because it took time to figure out what’s the keywords and all of that. But now it’s the cost is even lower. So rather than before, it was kind of like, I think about this on my own because I don’t want to go through the effort of the internet search, versus now it’s a metacognitive question of like, do I want to think about this on my own first to help myself learn it? Or do I just ask the AI for the thing? And that’s hard. Metacognition is hard. And especially if you’re potentially an undergraduate college student who the prefrontal cortex isn’t fully online enough to go, pause, is that a good idea?
Neil (20:32) Sure, yeah, and I think that is the challenge. I mean, I feel like the value of sort of formal education as a whole is to try and sort of instill like structures and practices that guard us from our worst impulses when it comes to learning. Because I don’t think of students as any particular magic category here, it’s just people. Students are just people that happen to be learning. And people ultimately, as you say, they’re lazy and know, like they’ve all the knowledge of the world is there. And yet we still have universities to help us learn it. And that’s because we need these kind of frameworks and like, you social practices and, you know, assignment deadlines and things to motivate us. And so we need these kind of frameworks to guide us to the right thing. And I think that’s still going to be true with AI, you know, how do you use AI effectively? What should you not be doing with AI? It’s like, you know, don’t cram for the exam. There’s cognitive reasons why it doesn’t work. And there’s also cognitive reasons why I’m pretty sure that asking the AI immediately every single time is not going to produce as good results as a kind of more mixed use of it, where you think for yourself for a while, ask the AI, and kind of have a mix of these practices.
Kristin (21:49) It also, it reminds me how, like, you know, a hundred or two hundred years ago, like exercise was built into your day because literally there was no other option. Like you had to walk to get to someplace. Or like you were a laborer. You had to like pick up all the things that are heavy and do all of the things. And like there was no such need for exercise outside of that because l your everyday life was that. And now we have to choose to do exercise because our everyday life does not necessarily have built in those kinds of things. And now we’re like reached a new level of humanity, whether or not this is good for us, where we have to choose whether or not to exercise our cognitive functions and in some ways also our social ones, because of like the ease of asking the AI for help rather than like your friend or your dorm roommate who’s like potentially even in the same room as you. Yeah. That is, I think, one of my fears.
Neil (22:42) Yeah, and I think that’s another, you know, I think worry that I have a bit is that AI is a relatively anti-social technology. I don’t say that with necessarily like terrible negative judgment, because there are advantages to that, you know, the fact that people can ask it really stupid questions, and it doesn’t judge them is almost certainly a positive. But you get this weird effect. And I think, you know, we mentioned, we discussed earlier about a paper from Temple University about the social practices with AI, where essentially people will ask a friend and that friend will just turn to AI, ask the AI and give them back the AI answer and also feel like, well, why did you bother asking me?
Kristin (23:24) Yeah, that was the January episode.
Neil (23:26) So it could kind of destroy these kind of social learning practices and make it make social interactions feel unnecessary in a strange way that definitely wasn’t the case, you even with kind of maybe internet searching.
Kristin (23:38) Yeah. I think like to add nuance to that, I think also though, even asking your peers stupid, quote-unquote stupid questions. Like I agree in some ways that AI being there is good for that because it lowers the barrier to a student to potentially ask that question. But it also, I think, robs the students the opportunity of practicing putting themselves out there.
I think the main message that I want this episode to have is about this idea that learning is not obsolete. And this is why. Do you think that there are practices or something that a teacher can do or they can teach to their students that might help us in this moment? Because I think the practices right now, some of them aren’t going to be obsolete very quickly. Others are potentially gonna work out, but we have to kind of in a sense wait to get the data to find out if they are. Like what are your thoughts around any of that?
Neil (24:40) Yeah, I mean, this is all kind of speculation, but I would think maybe as I said, maybe trying to avoid reaching for the AI too often or too immediately. And also, of course, there’s quite a different use pattern between, you know, asking the AI to do the thing for you where you say, okay, you know, I need to write this program, you know, I need to, you know, sum up these numbers, write me the program that sums up these numbers versus, you know, I don’t understand what I might need for this, could you tell me the kind of concepts that I might need? So perhaps teaching how to ask those kind of more abstract questions, even though it feels less direct, might be more beneficial for learning rather than just, you know, give me the answer, give me the code that I need to kind of learn how to ask it for the sort of more conceptual questions or for feedback rather than just do the thing for me.
Kristin (25:37) I think, so, this reminds me of, I think Mark Guzdial said it, there’s a difference between offloading and outsourcing your cognition. And the dialogue that I often see is more about out offloading. Like you offload your cognition to the AI. And I liked his distinction though, is like, no, it’s outsourcing because the cognition is not happening at all in your head, it’s happening on this other thing. While offloading, like you can call, you know, writing notes technically offloading.
But that doesn’t necessarily mean that the cognition is somewhere else. The cognition is still in your own head. So I kind of like that distinction. I think the hard part, though, for students is that they don’t know what is reasonable to outsource versus, like, it’s a tool for offloading. Does that make sense? And I don’t know the way to help the student because often they have competing interests in that moment because we are all lazy and we are all trying to be efficient with our time. And it’s easy for them to rationalize. I can learn this later. I will just have the AI do it now. And I will learn it later when I need to, if I ever need to. And I don’t know what to say to that.
Neil (26:58) My colleague, Michael Kölling, often talks about, you know, the whole of formal education is to some extent, I don’t think you use the word trick, but something like a trick that, you know, students think that we want them to write this program and submit this program that, you know, does this particular task so that we can then grade it. But of course, we don’t want that program. We could write that program ourselves. We’ve got hundreds of programs from last year that all do that thing perfectly well. We’re not actually interested in the program in the slightest.
What we’re interested in is the process that you took to come up with it and the learning that took place along the way. So, you know, students are always thinking of, you know, what artifact is due by what deadline, you know, I need this essay, or I need this program or whatever. And actually none of the product, like, you know, the actual things they produce, is of any interest to us. It’s all about the learning journey along the way.
And we need to somehow also kind of come up with ways to use AI that respects that same kind of misdirection sounds like we’re doing it deliberately, but you know what I mean? Like we kind of, say we want this, but really it’s because we know to do that, you need to do some other thing. And the reason why AI is causing so much consternation in education circles is because it broke this. So now you can just say to the AI, give me this program in the same way that students could always have cheated by copying someone else. But now it’s even easier. And then none of the processes happened.
So we got the artifact that we said we wanted, but of course it wasn’t what we were actually looking for. We wanted them to learn along the way, and it’s how we can kind of use AI, I think in ways that respect that, that still take the student on a learning journey. Hopefully AI-assisted, you know, rather than kind of just, as you say, just immediately kind of, you know, handing over to the AI.
Kristin (28:47) And I think that it’s fascinating how, since the the kind of the cost of producing the product has gotten basically down to zero. And so the students are in some ways rightly saying, “Why do I need to do this given that there’s something that could do it at no cost?” While we’re like, well, also they’ve been kind of trained that the product is the point because unfortunately, that like we got so focused on our structures, we forgot the method behind why we have those structures. And I can’t help but wonder how well we can get students to kind of once again more believe like the point was the journey the whole time rather than the product. While also hoping that in the heat of the moment they make the right choice, even though we all know that we are humans and I don’t even make the right choices every single time.
Neil (29:45) Yeah, and I think part of this, course, will come back to kind of assessment practices and how do we assess students’ knowledge and so on? I mean, there’s something that I find kind of interesting that, you know, when AI came along, you know, not all technological advances necessarily improve our sort of day-to-day lives, you know. Some do, like, you know, the washing machine made everyone’s life better, right, because you don’t have to do all this laborious washing, and that’s all great. But there’s other cases where maybe it eliminated whole job classes or made them more boring.
So academics, when they hear about AI, they’re often like, great, I can get the AI to do the grading, which is the really boring part of my job that I find very laborious. And I can focus on content delivery, which is the bit I find really exciting and talking to the students and so on. And then when you ask the students, you may well find that actually they’re like, well, all the knowledge is on YouTube anyway. And I like just watching the videos in my own time and skipping the lectures and lying in bed watching YouTube.
But I spoke to someone at another university recently where they started putting in AI grading. And actually, the students were quite resistant because they felt kind of cheated. They were like, well, I expected a real person to look at my program and kind of give me feedback. And I think to some extent, there’s a like, there is a social impulse there. I want another person to judge me and tell me how well I’m doing. I think that’s quite a human impulse when you’re learning that you look to a teacher or a parent or whatever to say, you did well, or you didn’t do well, or what did you do wrong? And somehow just the knowledge that it’s a machine kind of removes a part of that, even if maybe you got the same words coming out of one or the other. It’s important, I think, as a person.
And so there’s kind of this, you know, the tension between like, okay, maybe what the students really want from us is the feedback and the grading part more than the content delivery part. But that may not be the part that perhaps everyone was really invested in doing as a lecturer. And how do we navigate those kinds of issues? And what does feedback look like? Some people have been doing things like one-to-one interviews, but it’s hard to scale that up, right? You’ve got five minutes a student, but you’ve got 400 students on your course. How do you make that work? And even five minutes is pretty short, you know, to do what you want to do. So I think that’s something that we’re going to have to have a think about. And I’m not sure what the right answers are there.
Kristin (32:07) Yeah, I think that there’s a lot of tensions here where like, I think the students rightly are wishing for a relationship with the faculty. And like one way that they kind of envision getting that is through the like someone cared about the thing that I produced. And they took the time, and they gave me the feedback on that thing. Like they’re that’s one thing that they’re kind of imagining is how to build that relationship.
While, us faculty, like grading jail is real. We don’t like being there. And we more see the relationship building happening when they, we see the students’ faces as they light up with like, now I get it. But that is admittedly very much a one-way street, except for the few students who obviously do raise their hand and ask a question. Then you might develop a relationship with, like, those specific students, but it’s still in a very much different context compared to even the grading relationship, which is much more one-on-one in essence than it is the one-to-many when you’re doing a class.
And I the but the and the other tension is that, like I in some ways I think value large classes because to me, if that many students want to learn this content, I want to teach the like as big of the class as possible because I want to make sure that all the students get this content if that’s what they want. But there’s only one of me. And like, how do you manage that? And I don’t, yeah, I don’t that yeah, I don’t have a good answer to that one. I wish I did.
I think also though. I’ve taught data science so long I feel like I imagine everything as a normal distribution, even though most of the distributions I deal with don’t actually have, aren’t normally distributed. But like if we think of the bell curve, there’s probably only like the top ten percent that really want that relationship. And like the last ninety percent are like, I’m just here to try and make sure that I learn this content well enough in some way, shape, or form. But I’m not necessarily interested in developing a relationship with the faculty.
Neil (33:58) Yeah, it’s like, I think every teacher at some point has received the question, you know, you try to engage them by saying, don’t you think this is really interesting? And they ask that sort of killer question, “Is it going to be on the exam?” Like, you know, “Are there marks for this?” And essentially with the implication that if not, I don’t care whatsoever. And it’s quite soul-destroying, isn’t it? But I’m sure we’ve, we’ve all had that a lot.
Kristin (34:20) Yeah. And that’s probably like eighty percent of the students, if we’re honest. They’re like they’re there because like I want their their motivations are so different. Yeah.
Neil (34:28) And that’s this misdirection again, essentially. They come because they want a degree. And so that’s all they want. They want the degree so they can get the job to do whatever they want to do. And we kind of trick them into educating them along the way in order to give them that degree. People get so fixated on the kind of literal outcome. I need the certificate. And then we have to kind of structure things so that we actually produce what we really want, which is a learning experience, you know, and to come out with them having learned and be more confident and so on.
Kristin (35:03) It’s another like product versus the journey situation.
Neil (35:07) Exactly. And be the consumer wants the product somehow. And, know, we’re there to engineer the journey. I think, I think that’s the value, like I say, of formal education. And that’s something that will still be there even under AI. I mean, know, the MOOCs is a another kind of comparator that, you know, people thought, okay, maybe MOOCs are going to completely replace universities, because everyone will just self-study, you know, self-certify, and kind of that will be enough. And obviously, that didn’t really pan out for various reasons. And I think the same is true of AI. I don’t expect AI to somehow remove universities and formal education in general, but I think universities will have to adapt a lot on the sort of lower level as to what does assessment look like? What does teaching look like? You know, how do we structure these whole things to account for that? But people will still need some kind of extra learning structures and learning assistance, no matter how good the AI gets. That’s my speculation.
Kristin (36:09) I definitely agree. I think it’s like in anything, it’s very much like a what environment will get me to learn the thing that I want to learn or do the thing that I want to do. And we do that environment curation all the time. And in some ways, university and going to school is part of that environment curation to get yourself to do what you want it to do or get your future you to do the things you want future you to do for the good of future you, even if like current you is like pass me, why’d you do that to me?
Neil (36:39) It’s like the exercise example you mentioned earlier. In theory, we can all just exercise by ourselves. We don’t need any equipment. You can just do star jumps and press-ups and whatever. You can do your whole fitness stuff completely yourself at any time you want to in your own home. And then we can all be super fit and healthy. But in practice, I need a time in the week where I must go, and there needs to be other people as well, you know, some like football or exercise class or something, so that it actually makes me go. So I need that environment and that structure to make me do it, even though in theory, you know, I could do it at home anytime, but that’s not the case. And I think it’s exactly the same for learning. I could learn everything in the world right now, if I just Google for the right thing, but it’s not going to stop me just, you know, getting off this call and playing a game or, you know, just like, you know, scrolling through YouTube or something.
Kristin (37:35) Yeah, it’s like, why can’t I will myself to do what I want to do? Oh right, because I’m human and lazy.
Neil (37:40) Indeed. [laugh] We just save us from ourselves, think that’s it.
Kristin (37:45) Ha ha, yes. All right. Let’s go with too long didn’t listen. What would you say is the most important thing that you want our listeners to get out of our conversation?
Neil (37:54) I think it’s that having all this knowledge available via the internet or AI doesn’t mean that you can skip learning it. It doesn’t mean that we won’t need formal education in order to help us learn it, but it might well change the day-to-day practices that we engage in in terms of content and in terms of assessment and probably in ways that we don’t fully understand yet.
Kristin (38:15) Yes. And it’s now it’s all a research question. And unfortunately, research takes time. All right. Well, thank you so much for joining us, Neil.
Neil (38:24) Okay, thank you for inviting me.
Kristin (38:18) And thank you for listening. I am sure you know someone who would find this episode helpful. Please consider sharing it with them. That is a great way to support the podcast. And if you’re looking for other ways to help us out, join us on Patreon at patreon.com/csedpodcast. Particular shoutout to patrons Glenn Downing and Michael Shindler for helping to keep the podcast ad-free and supporting production. For past episodes, transcripts, and links, visit csedpodcast.org and don’t forget to subscribe so you never miss an episode! And otherwise, this was the CS-Ed Podcast. I am your host, Kristin Stephens-Martinez, and our producer is Chris Martinez. And remember, teaching computer science is more than just knowing computer science. I hope you found something useful for your teaching today.