Reimagining Computing Education in the Age of Generative AI

Speaker A: Sa.

Speaker B: Welcome to episode 270 of Silver Lining for Learning Reimagining Computing Education in the Age of Generative AI. I'm here with my co host Chris Dede from Harvard, Lydia Tao from the University of Toronto and Punya Mishra from Arizona State University. With us today are Mark Guzdiel, a professor, uh, in Computer Science and Engineering and, and the director of Programming in Computing and for the Arts and Sciences at the University of Michigan. He studies how people come to understand computing and how to make it more effective. Along with Mark is ah, Barbara Erickson. Barbara got her PhD in Human Centered Computing from Georgia Tech University, is now an associate professor at the University of Michigan. She applies research results from Ed Psych to help students students learn to program. She creates free and interactive ebooks with new types of practice problems, including some that leverage AI, generative AI. She's going to talk about some success stories as well as some challenges she's faced in using generative AI. I must say in looking at Mark's bio, I had to print out your Google Scholar listing of all those articles. I think I ran out of paper. Mark, um, authored a lot of stuff but the one that in particular I'm looking at is this article called Motivating Project Based Learning Sustaining the Doing, Supporting the learning cited by 7,600 people. Right. With your friend Elliot Soloway and you know uh, Joe K. And Anne Marie Palar and Blumenfeld and Marx and all these people. You know, I must have cited that at least 100 times myself. So I helped your guru Scholar.

Speaker A: Thank you.

Speaker B: Score out. You know, um, but you know, you never know what you get involved with. Right. That's going to be read by a lot of people, incited. Right. Um, so you know, we could read yourself about that article for an hour and how that motivates through project based forms of learning. Right. Maybe a follow up show. So um, Mark and Barbara came down to Bloomington and had a chance to chat with our students and with our other faculty in our program, um, and with myself. And I found out Mark's involved in creating new ways, new pathways for learning computer science. And uh, his role is not necessarily in the computer science program. It's over in Arts and Sciences. I thought Mark, you might start us off talking about that program, uh, and how long it's been running and the successes you've had and what you're going to be working on in the near future. Then Barbara might talk a bit about how she's helping students succeed through innovative pedagogical activities. Whether it's peer based learning or other things that she's in generative AI. Ah. And so forth in computer science. Maybe you're going to mark, give us the big picture framing and then Barbara can, can jump in and tell us some of the on the ground, you know, in the trenches, what's going on in reality. Right. And what's successful in reality, what's not, and maybe what she's looking to do in the near future. So instead of asking you a question, I'm asking both of you to give an explanation. Um, does that sound good?

Speaker C: Okay, that sounds great.

Speaker B: All right.

Speaker C: Um, and we've done this sort of tag team thing before. We, we call it playing jazz. Right. Lay down some notes for her to pick up and vice versa. So. Okay. Uh, a theme that connects both of us is we're really interested in having a broad participation and access to computing education as possible. And so we're interested in making more people more successful at computing education. Um, um, in 2020, uh, September 2020, I was asked to lead the task force, uh, across the University of Michigan, but mostly within the College of Literature, Science and the Arts, which is the majority of the University of Michigan. It's twice as large as engineering at the University of Michigan. Um, to ask the question, what is it that LSA students need around computing education? Since I just got there in 2018, I didn't know my way around. They partnered me up with, uh, Gus Everard. Gus refers to himself as a first generation computational cosmologist. He simulates the Big Bang and tries to create this universe in silicon. It's pretty amazing. It's fun stuff that he works on. And so the two of us led a task force to try to ask the questions, what does LSA want to do with computing? So they bought me out of my time in computer science. So I am 50% computer science, 50% LSA now. And the big idea that we came away with was that the three themes for the way that LSA uses computing different than what people in school information or computer science do is thinking about computing for discovery, scientific computing in both the natural and social sciences, computing for expression. Humanities has always studied how people communicate with one another, but that tends to be text, um, and paintings and art. Now we have to think about social media and Pixar and all the different ways that in TikTok, all the different ways that computing allows us to communicate with one another. And humanities scholars study that. And the third area is computing for, for justice or critical computing. Looking at the way the computing is impacting our world and saying maybe that isn't the way we want it to be. How could we design it to be better? So we started in fall 2022 with a course in each of the expression and justice programs. Um, we had a total of 25 students in fall 2022. This semester we just passed 700 students enrolled. Nobody is required to take any of these courses. Almost all of the courses have a significant programming component. Um, I just got finished teaching a course on uh, how generative AI works designed explicitly for arts and humanities students. We call it alien anatomy. How ChatGPT works. I teach it with a computational linguist, um, who actually does know how CHATGPT works. And I just know how to teach things. We use a lot of things from Snap, uh, uh, out of Germany, block based programming language. We use a lot of stuff, uh, out of other groups in Eastern Finland and at Cambridge. Find in general that the Europeans are thinking more broadly about helping people to understand computing at the conceptual level as opposed to teaching them job skills in Python collaborative books. Um, we have got a bunch of different lecturers working for us now. Um, yeah, I think that's probably good enough to get a sense of what PCAAS is about and I'll pass it on to Barbara.

Speaker D: So I'm in a school of information and I teach an intermediate Python programming course and I've always been in computer science. So coming to a school of information I wasn't sure what that meant. What do our students know and do? And it turns out our students are pretty much either going sort of data analysis route and who think they will maybe program or will just become consultants versus user experience route. Um, but I also because, uh, the computer science department at Michigan mostly teaches C, I get a lot of computer science students because they want to learn some Python and my course is set up for that. So I bring in people who are supposed to know a little bit about programming and I try to take them to sort of develop intermediate skills. But I have this wide range of people who really want to program and people who don't want to program. And so it's a complicated class to teach. Um, but what I try to do is reduce the cognitive load of programming. A lot of people teach programming by making people write code all the time from scratch, or at least they used to before AI. And that's too much cognitive load often. So I try to investigate what are ways that we can reduce the cognitive load which is often completion problems rather than whole problems. One of the things I studied for my Dissertation and my students continue to work on is called Parsons problems where, where you give people all the correct code but it's mixed up like a puzzle and you can have distractors to get them to pay attention to things you want them to learn like case matters. Like you have to have a colon and then they have to drag the blocks, select the correct blocks, drag them in the right order and then they also, instead of getting compiler errors they just get told this block is out of order, this needs to be replaced. So a lot less cognitive load. My research, my students research has shown yes, a lot less cognitive load. Students think they're a lot easier. So the idea is scaffolding to help them learn to program but without making them write everything from scratch, which I think now especially with AI and there's less of a uh, uh, intention of that we want them to necessarily write all code from scratch. Now I think Parsons problems are a great way to still make sure they're thinking about the code, they're thinking about how things go in order, they understand code. That's one of the things we've been working on. Traditional Parsons problems. You, you break one or more lines of code into a block and you put them in order both horizontally and vertically for indentation for Python because Python matter, indentation matters. I also had a student, Zahan Wu who worked on horizontal Parsons problems because sometimes just a single line of code is overwhelming. Like public static, Void Main and Java. And so just put these blocks together or for regular expressions or SQL statements. Um, I also use it for those. Over the years I've had good success with scaffolding students by introducing code, making them use puzzles for code before we ever tried to get them to write code or modify code scaffold that process. One of the things we found though of course expertise reversal effect is that the people who want to write code find the mixed up code problems annoying. Uh, because they just want to write code. We added a toggle feature. If we've asked you to solve a mixed up code problem, you can always choose to just solve the equivalent write code problem. Then conversely we thought well if people are struggling while they're trying to write code, we should help them. We would pop up a mixed up code problem if you're struggling while you're writing code. And I'm m often having them do this in lecture where I have 200 students and I can't individually help everybody who's having trouble writing code, but they can pop up this mixed up code problem. That is an answer they can get to a correct answer. Then with the advent of generative AI, we're like, well, one of the downsides of the way that we do the mixed up code problems, um, is block based feedback. With block based feedback, there has to be one correct answer. But we know there's more than one way to write code. Certainly what we've heard from students is if the way that that code is written doesn't match where I'm thinking, I don't find it as helpful. What we're doing now with LLMs is if we've asked you to write code and you ask for help, uh, we can take your incorrect code, go to an LLM, find the closest correct code, but, but serve it to you as a mixed up code puzzle with your incorrect code as the distractor blocks. So get them to help them still, you know, use generative AI, but don't just give them the correct solution without any engagement, mental engagement. They still have to think about it, they still have to finish it. And students, at least so far in our research, are telling us they much prefer that than just getting a correct answer from gen AI. So that's one way that we're using gen. Another is I've been doing research on peer instruction, which is, so one of the things I try to do is to, to get more active learning in lecture, not as much passive lecture. So I do a lot of peer instruction, which is ask a hard multiple choice question, have students answer individually, often with like little clickers, uh, and then discuss with peers, what did you get? What did I get? And then answer again, and then the instructor leads a discussion. So we created a free tool in our interactive ebooks on the Brinstone Academy platform called Peer plus to support peer instruction. And then we also added a text chat feature because we, we had a hypothesis that if we could assign people to groups to maximize the number of groups, that people with different answers, we would get better learning. Because we know that learning happens more when people have different answers because then they want to discuss if, if they have the same answer, they're like, oh, we must be right even if they're not right. So if we have people with different answers, then there's a discussion. We have found that to be true. Uh, we've done research with, um, Duke University and Virginia Tech and Berea College and ourselves with this tool, uh, and we do find better learning gains from the assigned groups with text chat, but the students prefer verbal chat over the text chat. Part of the problem may be no shared annotation space to Mark up the question. If we're sitting together at a laptop, we can both see the question. We can talk about. It's harder when you're not. We're continuing to explore that. We're trying to do. Uh, we're going to do an AB study where we assign the same discussion group to either discuss together verbally or in the text chat so we can do a direct comparison. In an AB study, the way that we had tested before was with different questions. Then another way we're using generative AI is I'm working with, um, another faculty member at SI who's doing ah, an AI system called Learning Clues that ingests instructional material and, and then uses reg to answer questions. And so he's ingested our ebooks, our lecture slides. And then we're testing now that feature of can we have it answer questions, student questions, but not just give away the answer to them and point to

Speaker C: the right references at the end.

Speaker D: Yeah. Not make things up. Point right back to where in the book that is or where in the lecture slides.

Speaker B: Someone listening in here might not know what SI is. It's a school of information, right?

Speaker D: That's correct.

Speaker B: Okay, so I've got a question related to what you're talking about with Duke and such, but they can wait. I think Punya's got a comment and then we're going to go to Lydia.

Speaker E: Thank you. Um, this is honestly more of a comment, but I would like to see what kind of discussion sort of emerges out of it. So Mark, I was really, uh, really connected with me the three themes for computing for when you talked about discovery, expression and justice. Critical, uh, computing. And it promptly flashed back to me, uh, something John Dewey had written back in 1919 I think, uh, where he talked about the four primary impulses of learning. And he labeled them as being inquiry, communication, construction and expression. And so that aligns so well because you know, you are using in, in the humanities and the social sciences, you are building things. So that's the construction part. But it's also about self expression. It is about communication. And I just found that really resonated with some of my more sort of recent sort of writing and thinking about it. And the other piece that I would love to hear your responses to that is if you think about, there have been these two strands in technology and education. So one has been sort of the intelligent tutoring system where you'll have this infinitely patient guy sitting there, Socratic dialogue, all that stuff. So I heard Barbara, you talk a little bit about sort of at the back end of what you're talking about, sort of. But there's this other, which is like Seymour Peppert and so on, which is so much more about discovery and building and, you know, the bicycle for the mind metaphor and all of those things. And it seems to me that generative AI, if we think about it as like a chatbot, which is what most people have been thinking about, that leans towards sort of the intelligent tutoring system, but with this new powers of coding, you know, that you can do things that you can build. I hope they can hear me, what I'm saying. Um, and maybe we just continue the discussion if that happens. Um, that idea of the children's machine might actually be much more accessible with generative AI tools than it was ever before. Um, Scratch is a good example of that. But I think that the powers of multimodality and so on have become so much more seamless in some ways, uh, problematic as well. I mean, we can talk about those issues as well. But I think that that's an interesting tension in the field with generative AI is sort of highlighting. It is both like a tutoring system. You know, that's the Khan Academy kind of a mode. Um, or the other extreme is sort of that you can use it to do creative work with it, to build stuff with in ways that maybe could not have been possible before. So I'd just love to hear, uh, your thoughts, uh, on that, assuming that you can hear what I'm saying, because your video is off and I won't

Speaker C: share it was chopping up and so we turned off the video and I think. I think I got. I think I got the gist of it.

Speaker E: Okay.

Speaker C: Okay. So in my talks, uh, and Kurt saw some of this at, ah, Indiana. And this sort of. I think I saw Chris used to do this. Um, I started out from historical perspective. The first arguments that everybody should take computing, that we owe it to everybody to learn programming, come from the 1960s, way before Silicon Valley, way before anybody was interested in building apps. Um, so Alan Perlis made the argument that programming changes the way that you understand your world around you before Seymour Papert ever did. Uh, George Forsyth argued in 1961 that all STEM students needed to learn programming as well as mathematics and natural language. Um, and then both C.P. snow and, um, Peter Nauer, um, made the argument that, uh, they realized that software was going to run the world. And unless people understood something about software, as CP Snow put it, um, a handful of people will be making decisions in secret which are going to affect Our lives in the deepest sense, which was amazingly prescient for 1961. So I think there's actually a range of answers beyond the two that you laid out. Ponya There is certainly constructing things. We can make things with computing and it's so cool and it's so fun and lots of people can use it, um, and we can use it as a tutor. It can teach us. But there's also like what computational scientists do with computing. They don't actually care about the product so much. They care about the analysis and being able to look at data and be able to visualize data and create models and simulate. They don't really. Most m of the computational scientists, I do make no abstractions at all, not even a function. They're simply using libraries in order to come up with new understandings of the world. Um, then the critical computing people, they're not using it as a tutor. They don't really care about building things, but they want to look at how is this stuff impacting our lives. I particularly resonate with them today when you think about the people who are making decisions about whether they want a data center in their neighborhood. What do they even know about data centers? I think that's important work that Barbara and I did a long time ago at Georgia Tech. We developed this way of teaching computing called media computation. We were getting students from. She was dealing with high school students, uh, AP Computer Science includes the picture lab that Barbara wrote about manipulating pixels. Um, teaching liberal arts students at Georgia Tech, but thinking about how it is really motivating for students to learn computing in order to understand the digital media in the world. So they manipulate pixels of a picture, samples of a sound, frames of a video. And people want to understand their world. And that's different than computing as a, as a, as a building blocks or computing as a tutorial.

Speaker B: Thank you.

Speaker E: No, that, that's, I mean, ah, I think the point I was making was a. The resonance with doe's original ideas, you know, so that was like one that really struck me as you were speaking. But I, I completely agree that there are some of the variety of different ways that we can think about it. But uh, that this technology by. I think the discussion, a lot of the discussion, the air in the room oxygen has been sucked out by the chat bottom and getting, giving it the right answer. And I think that some of the examples you're giving and what I'm trying to talk about is that we definitely need to move away from that. That's one component of a larger array of things that we could be thinking about. That was, I think, the point I

Speaker A: was trying to make.

Speaker B: So Lydia is going to jump in here and I just want to point out that she was up late last night, uh, studying Parsons problems and trying to solve them because she wants to get her green card in Canada, you know, so until she, until she passes that, that person's problem exam. Lydia,

Speaker F: thank you for the joke. Kurt, you always make the best joke. Uh, thank you Mark and Barbara, for sharing. I really, really appreciate these really concrete, like, pedagogical ideas and scaffolds that you intentionally build into your teaching to really protect students cognitive engagement. Um, so really, really appreciate that. Um, I have my question, like kind of twofold, one kind of build upon Punya's question is, I really, uh, eager to learn about your take on the, this notion of Vibe coding. Because on the one hand people saying like, oh, you know, this is democrat, uh, democratizing, um, computing and anyone can build fast. And this notion of like, I don't know, solo founder, all of that. On the other hand, people saying like, well, like, people don't really understand the code. This is not, it's not safe. There's lots of safety guardrails. You need to be able to examine the code. So we really like, wonder your take on that. Um, my second question again is kind of, um, built upon this. And because of Vibe coding, a lot of people are like, oh gosh, you don't need programmers anymore and that will be the first job to disappear. Um, but I don't think that is true. And, um, from your deep expertise in computer science, I'm just wondering, what do you think is still worth learning in computer science? Science. And what's your response to people when they say, well, like, you know, AI can do this. We don't need people to learn programming anymore.

Speaker D: So what I'm finding talking to people that are doing coding with AI is that it's a multiplier for people that have experience. So people that have experience are much more productive. It doesn't, it's not a silver bullet. They still find problems and things, but they are able to be more productive and they can catch the errors because they know what they're doing. But if you multiply zero by anything, you don't get any, Anything. Right. So if people have no background, um, they can sort of like the way the block space made it easier for people to get started, to get hooked. I think Vibe coding is a way for people to create things that they want to do for their own personal Use right. Doesn't need to be secure. Not a lot of people are using it. If I can prompt well enough to figure out how to make something I want to make, then just like doing block based programming. I don't know if you're familiar with App Inventor, but it was a way that people could make apps. Uh, they weren't great apps, but they would work right. And so people could make things that they found personally meaningful and useful, which is great. So I think Vibe coding will have a place. Uh, but I would agree that I would not want anyone to use a system that was Vibe coded without somebody who knows what they're doing looking over that code and understanding it.

Speaker C: I. It's interesting to look at scientists and artists take on Vibe coding because that's where I live now. The blurts and sciences. Um, scientists use generative AI to produce code. But then this is really amazingly common, really surprising. They build a second system and something that they know and then compare the results. Because way more than computer scientists, scientists really care about the output from the code that the generative AI built. Um, and the idea of building two systems, one to validate the other. Computer scientists just don't do that. Scientists do that pretty regularly. The artists, sure, they have found that they can use Dall E, for example, to create something, um, but then it isn't exactly what they want. And so they start changing their prompt and to do the prompt engineering to get exactly what the artists want. It's about as much effort as they do processing now to learn processing, the programming language and build a sketch in that. Um, there are people who are using prompt engineering as a medium, as an artistic medium, but as a possible medium because it still takes a great deal of effort if, as an artist you want to realize your vision. So I think that particularly outside of computer science, Vibe coding is an enabler, but it isn't a replacement. Because what scientists are trying to do is advance knowledge. And so they actually have to understand the code to advance knowledge. And what artists are trying to do is represent their vision. And actually AI can sometimes get in the way of that.

Speaker F: I really like this. You said, uh, AI is a multiplier. When you multiply zero, you get nothing. And that really stays with me.

Speaker E: Yeah, If I can just jump in a couple of things. One is sort of the amplifier of expertise. And I think there is sort of an equity issue here because if you're thinking of a learner, by definition is a learner because they don't know stuff. And that to evaluate the output of AI, you need judgment and expertise. And so there is this fundamental tension, I think that's there. And I'll just speak for myself. Uh, Mark, when you talked about, you know, the way artists use it and the way, and I have that both parts of my life going on, so I do a lot of sort of artistic quote unquote with my definition of it, uh, stuff on typography and so on. And AI becomes incredibly interesting as a medium for me because there is a sense of control, yet lack of control. And how do I get it to do what I think is aesthetically pleasing? It's fascinating, it's just very interesting. At the same time, I dropped this link in the beginning. Uh, if you have a minute, go take a look at it. It's. I sort of used generative, uh, AI cloud code to analyze six years of the transcripts of that's 2.6 million words that have been spoken over. Silver lining for learning, for themes that I could do things that I could never do. But here is the best part. I kept building two models to test one against the other, which is why that resonated so well. So I would do the coding of the transcript, but I can't trust Claude to do it. Right. And so I would create a new chat or I would take part of the data and give it to chat GPT and say now you do the coding and let's look at the coding together. And so through that multiplicity of effort, I have a little more confidence in what I am now representing in those websites in those web pages. Right. And so both of those pieces really struck home because I live in both of those worlds at the same time, you know, or in parallel universes. And that resonated very well. So thank you for that.

Speaker D: Yeah. And I think, you know, just like the blocks help people get started with coding, that that vibe coding or just using prompts or agentic coding may be an easier on ramp for many people. Um, which is fine. You know, it's great that more people can do some coding and do something they find useful. I've always liked people to do something they find useful or interesting, uh, which sometimes in computer science classes isn't a goal of anybody. They do just boring stuff.

Speaker C: M Here's a hypothesis that I don't know if anybody's ever tested. Um, my guess is that students who use high level blocks to get something started versus to build the same thing using Vibe coding. I bet you the students with the blocks are actually learning more.

Speaker E: Mhm.

Speaker D: Sounds like a research question.

Speaker C: That somebody should, because the vibe coding can easily make the code completely invisible. And it's really hard to learn from things that are invisible. Um, but the blocks, you actually had to do something with it. You had to interact with the blocks, you had to assemble the blocks, you decide the sequencing, decide the composition.

Speaker D: Well, you could say that's like persons problems, right? We have evidence from that person's problems do help people learn common algorithms. You know, they have to think about the algorithm, what they're doing, you know, and so there's still, uh, there's, there is a tight correlation between performance on Parsons problems and performance on write code problems on exams. So if you know how to write code, you can put the Parsons problems in order. It is of course easier because it's not generating the whole search space of I have to write this code from nothing. And so instead I have just these things to look at and decide what order they go in. So it's easier. But I can still tell whether people actually know what they're doing at all from their Parsons problems on exams.

Speaker B: Olivia, do you have a follow up or a comment on what they had, uh, said there?

Speaker F: Oh, uh, I'm, I'm good, I'm good. Like Chris me waiting so patiently. Let's go to Chris.

Speaker A: No, no, no worries. Um, Mark, I resonate with your, uh, beginning talks with the historical framework. I've been recently quoting Shakespeare.

Speaker D: Oh, wow.

Speaker A: His plays. The moon is a thief, she steals her pale fire from the sun. And that's an interesting metaphor for human knowledge versus AI knowledge because human knowledge is like sunlight and AI knowledge is reflected out of human knowledge like moonlight. Moonlight's great, but it doesn't have enough energy for photosynthesis. It doesn't drive the water cycle. You know, we'd be in big trouble if our major source of light was from the moon because it isn't generated. And now, um, we're in a position where all of this AI slop is getting, uh, put onto the Internet and then built into the training sets. So we're actually diluting sunlight with moonlight, which is a really, it doesn't have a happy ending when you do that as a destructive cycle. And so it seems to me that, that what we have to do is increase human beings ability to create new knowledge so that the sunlight gets stronger and then the AI gets better because its reflection of that stronger knowledge gets better. So I really like the theme that you're working with and I think some of it gets into something else that I, I bring up in my talks. Which is doing things better versus doing better things. If, if you look at the whole history of not just computer science or not even just digital technologies, but if you look at the whole history of technology, um, there's nothing wrong with using technology to do things better, to take something that you're already doing and do it faster or more efficiently or whatever. But the big win with new technologies is always in doing things that were never possible without the technology. And that's where the creativity comes in. And I think, I think that the people who are doing better things are perhaps more likely to come out of the arts, you know, the liberal disciplines, um, than they are out of computer science, which is really, in a way oriented towards doing things better. Not as a criticism, but just as a descriptive statement. So I think one of the interesting things to chart is whether not only what, how well people are learning to, um, make computers do their bidding, but what kinds of bidding they're getting computers to do. Yeah, that's cool.

Speaker C: Can I tell a brief story? Okay, so, um, Chris, I'm sure you know Alan K. Um, so Alan has been also quoting Shakespeare in his talks. Um, his is from Puck from Midsummer's Night Dream, where he says, what fools these mortals be? And he uses this to say how AI is not actually intelligent, it is not actually human like, but we humans are so easily fooled. I mean, go back to Eliza. Eliza wasn't very much technically, but it fooled an awful lot of people to believe that it was technical. Okay, so the story I told in the chat, Chris was saying, JCR licklider. Okay, I want to tell my JCR licklider story. So in 1961, uh, there was this event at MIT, Sloan School Computers in the World of the Future. And both, uh, C.P. snow and Alan Perlis made the argument that everybody ought to learn computing. And Perlis made the argument actually so connects to what you were just saying, Chris. He was saying, um, that the. He was at Carnegie Tech at the time before Carnegie Mellon University and the business, uh, school was starting to run economic simulations. So before that you couldn't run economics as an experimental science. Let's devalue the dollar tomorrow just to see what happens. Um, but when you're running a model and a simulation, you can so explicitly we can now do things we couldn't do before. He was foreshadowing computational science. So in this book that Martin Greenberger wrote, that is all of the lectures and all the discussant notes, Peter Elias, then the head of electrical engineering at, uh, mit, explicitly says, I mean, it's amazingly foreshadowing. You know, Perlis, why do you think people should learn to program? Don't you think somebody's going to figure out how to interact with these computers in natural language? So only a few people would ever have to learn how to program? He literally said this in 1961. So J.C.R. licklider was the other discussant and Perlis responded as well. Licklider's response was so brilliant, he says, okay, you know, I'm going to bet that early Apes thought all of this language stuff was just so much work. Why go to the bother of learning language? He says, but today we have poetry. And so his response to it to Elias explicitly was, maybe with computers we are going to say poetry that has never been thought before, which is just lovely in 1961. So again, it's about doing things that we never did before. Um, Perlis pushes back on Peter Elias and says, no, no, you got me wrong. The point isn't learning the programming. The point is being able to build computational models out of processes and then simulate them. That that is the unique power of computing that completely changes computational science and engineering. So I love that. What Licklider and Perlis responded to Peter Elias with was essentially computing for expression and computing for discovery.

Speaker A: Well, both of you have explicitly made the point that I know I make, Lydia makes all of us make, uh, against the tide, which is that AI is not intelligent and, and it's fool's gold to think that we're going to turn our affairs over to this, um, quote, intelligent, unquote partner.

Speaker D: Well, I agree it's not. I agree it's not intelligent and I'm cautious about that. But I do think we'll find just like, you know, the MOOC hike hype, right? Oh, these, we have these massive online courses. We're not going to need colleges anymore. Like, no, that wasn't true. But they have found a niche, right? There are places they're useful. I think we will find uses for AI, that there will be things. For example, I'm excited about using our peer instruction stuff for people who can't come to lecture or people that are taking online large courses where interactivity is. One of the complaints they have is they don't interact with anybody in that situation. One of the things we're trying is if you're answering a peer instruction question and you're by yourself, yourself, nobody's there, you don't have anybody to interact with. Can we use an LLM to interact with you, to talk to you about the Question and try to initially, you know, try to get more information from you about your justification, try to lead you a little bit, so act a little bit as that uh, more intelligent other as far as, you know, trying to lead you to uh, think about these other things. And we're about to start experimenting with that. But I'm excited about that, that as you were saying, make things better. Because we didn't have a technology for helping people who were asynchronous, working on their own to do anything more interactive, that scaled. But now maybe we do.

Speaker A: You know, I've helped uh, a program at Chatbot, uh, that aids students in reflection. So they have an experience, in this case it's an experience playing a game about ethical decision making. And then they've made some decisions in the course of the game and afterwards the Chatbot helps them to reflect on it. And I think it speaks to the strengths and the weaknesses of the Chatbot because the strength is that it can be a mirror and you can use it, Eliza like to get people to really step back and look at themselves and think about issues and it can bring up new issues as you, that you may have missed in your initial thinking and so on. But uh, as an instructor for ethics, AI is problematic because it doesn't understand so many of the things that are what ethics is all about. So it is a matter of uh, like every technology, understanding what it's good at, but not thinking that if you have a hammer, everything looks like a nail.

Speaker C: Is it worth a trillion dollar valuation to be able to hit that nail? That's the real uh, for all of us, our economic question today.

Speaker B: So I've got a story to share, but Lydia has a question that's more important. Lydia, jump in here.

Speaker F: You can go ahead, Kurt. You've been waiting the whole time. I believe there's time in the end.

Speaker B: Well, we're telling Shakespeare stories and you know, on and on, but no one, we've had 270 shows, right. And no one's told an Elliot Soloway story. And wasn't he your advisor at least something like that, Right, Mark?

Speaker C: Oh, he's amazing. And he's not stopped. He was my advisor and his office is down the hall from me in the computer science building.

Speaker E: No way.

Speaker C: Popcorn. Um, and he's still going strong. Actually, um, uh, just yesterday was the hooding of PhD students at the University of Michigan. And Barbara and I sat with Chris Quintana, my academic younger brother. Uh, both of us were students of Elliot's and we shared a picture With Elliot, who sent us a picture of um, himself with his new grandson. Ah in St. Louis. Uh, his daughter Emma just had a baby and that was lovely. Um, yeah, and he's, he's going really strong. Uh, you know the work that Barbara and I did on media computation, you know there's, there's, there's an intellectual connection back to the work that I did with Elliot on media text where we were building a multimedia composition tool and Elliot used to refer to it as

Speaker B: the Brownie Kodak Instamatic.

Speaker C: Right, the Brownie and stomatic of multimedia of how do you get started? And you know I really think that's where, that's where PCAST lives in a place that computer science kind of ignores these days. Um, no professional photographer is going to use a Brownie Instamatic today, but it still is a pretty easy thing to get started with. An artist might for particular effect but it doesn't do production quality stuff. And this is the problem we have. You know I teach these courses for computing for artists and we use Snap and I give them a bunch of high level block code and they build so much. They build image filters, sound filters, video games, uh, drawing programs, interactive webpage, all of this because I can give them high level blocks. We then have a follow on course or we do it in Python and we cover less than half the topics because well, this is harder to code in Python but computer science, they're not going to do anything less than Python. C uh, one of my pet peeves, the most common programming language on any campus today, I'm going to bet in terms of number of people who use it is R. Because all the social scientists use it all the time. All the natural scientists use it for doing their modeling and simulation statistics uses it all the time. And I can't find a computer science department that teaches the computer science of R. They're just totally disconnected from what the rest of the university is doing. PCAs we do teach classes in R because that's what Arts and sciences uses. Um, so I think the computer science has sort of become too focused on the production on the Silicon Valley, on the professional practice. And that's what that Brownie Instamatic idea I think is really good at. Give you something simple so you can get started, get the ideas.

Speaker D: When we had first graduated undergrad and we were working for Bell Communications research or research, um, one of the interesting things then is that you didn't expect to stay a programmer. You started out as a programmer, but that was a low level job. And eventually, if you were any good, you were expected to move up to become sort of a manager, designer, whatever. You didn't stay programmer. And then over time, programming became, uh, an uh, end in itself. But originally it wasn't an end. It was just. That's the low level job you got to do so that you can move up a level. Uh, and I think with AI, we might see that more change that you're taking on more of the designer. But how do we train people to do that without teaching them how to program or without them understanding how to write code as much? I think they still need to learn how to read code, reason about code, test code, uh, break things down into smaller pieces. I think what we need to teach and the percent of how much we spend time on things needs to shift a lot. But I think that's interesting. Back to the future where programming or your job with computers wasn't all about programming, it was about doing the programming. So you knew stuff, but then you'd move up a level.

Speaker B: So, Mark, I, uh, have to ask, what year did you start graduate school with Elliot? Was it in the 1989. Okay, so I'm sorry, that's not true.

Speaker C: We moved there in 586. No, that's for our masters. He asked for my PhD. Elliot got. I got there in January, Elliot got there in October. And I was taking classes with, uh, uh, Pat Baggott. And um, uh, let's see, Carl Berger was there. Um, and uh, uh, this was before Joe got there.

Speaker B: And Bob Cosmo was there too. Robert Cosmo.

Speaker C: Bob Cosmo was the one who convinced me to start the PhD in the first place. The joint PhD in education computer science. But it was actually Yasmin Kafayi. Yasmin came over with Elliot as sort of his lab manager from Yale, and she was taking classes in the School of Ed. And I met Yasmine first and she introduced me to Elliot, and then Elliot became my advisor.

Speaker B: So what year was that, that they all came over?

Speaker C: I think 89. I think that by 1990 I was working with.

Speaker B: So my story is Aera. Aera was in san francisco in 89. It was also in 92.

Speaker A: Okay.

Speaker B: Um, and so Tom Reynolds and I were talking to Bob Kosma. He was trying to help us out in some of our research on computer assisted writing or creating computer prompts for student critical and creative thinking, and a keystroke mapping system, all this stuff. Great guy. We should probably have him on a show sometime, maybe have the clerk and Cosmo debate here. Um, but you know, we should have all. You know. Anyways, um, we should have Elliot on too. But, um, so one of my first ARAS is 89. My first one was 87. But just after John Cilly Brown wrote the article on situated cognition. So the big hype on. On that, they had a special symposium on that. I remember. But they had one special event where Lydia's now partner there at Toronto, Marlene Scardimalia was at. So Marlene, um, John Bransford from Vanderbilt, Roy P. From Stanford. Ah. And all these people were showcasing their research. And at the time, I think around that time, Apple Computer was funding these sideshow talks with a lot of technology in them for a brief moment, maybe a few years. That might have been one of the years. So all these people give brilliant speeches. And Tom Reynolds and I are sitting in the back just like work, like kids in the candy store. You know, we had read all their papers. You know, we were just finishing our dissertations and maybe I just finished we both in 89. Um, and, um, there's this guy in the back of the room and standing next to me, and he's just smiling the whole time. Kind of short and stocky guy. And, um, Tom looks at me, he goes, who's this guy? And I said, he's the janitor. Gonna clean up after everybody. They're done. He looks like the janitor. And so they all gave their speeches, Roy T. And everyone else. And this guy walks up the stage and he's the last speaker, and it's Eliot Soloway, and he's got his slides, and he's throwing his slides all over the place. He wasn't the janitor. He was the guy who's gonna squeaky teenage boys tell us about the state of technology integration and wherever around the world. Uh, anyways, he's brilliant guy, and I was just like carried away, but I thought. I seriously thought he was the janitor. That's my story for today.

Speaker A: So as long as we're talking about six degrees of connection, um, in the School of Information at the University of Michigan, of course, is Barry Fishman. Yes.

Speaker C: Just saw them yesterday.

Speaker A: Barry and I co authored a, uh, chapter in the Handbook of research on teaching for AERA in 2015. And these chapters are like books. I mean, it's. It's not a chapter chapter. It's like a book chapter. And it was. We had a great time together, Barry and I. And I. I want to bring up that because of the School of Information, I think Michigan was a leader in. In conceptualizing there should be a school of information. And what that would be about is a, uh, kind of an interdisciplinary nexus. And I'd love to hear your thoughts about the school of information there and the role that it's played.

Speaker D: Oh, that's definitely. You know, we have, it's a wide ranging place. We have economists and librarians and archivists and whole bunch of people doing social media and HCI and education. And it is quite a varied place.

Speaker C: Melting pot.

Speaker D: Melting pot with lots of different things going on. Um, I, unfortunately, I don't think we are as forefront as we used to be in some ways. Like, you know, I was excited about going there because I was working on interactive ebooks and I thought, hey, you know, place that does library science and stuff. This seems like a natural home for thinking about interactive ebooks. But, um, you know, that's. They, they haven't hired many other people to sort of work on these sort of things. And then Barry left to go to a school of Ed. So I was working with Barry with one student, but then he left to go to the school of Ed, back to the school of Ed. So he's left now the school of information.

Speaker A: Oh, has he? Okay. Yeah, I didn't realize he'd gone back again to the school that he has.

Speaker C: You all should have Barry on your show because he's doing something really interesting. I actually have if this will work. Uh, no, that doesn't really work. Um, I have his sticker on my, on my iPad. Oh, let's see. Let me try that. Uh, turn off the blur. Video effects. Uh, how do I turn off video effects?

Speaker B: Barry was a master's student in our program when I arrived here at Indiana. And uh, Indiana is one of the two or three for School of Informatics. Michigan number one.

Speaker A: There it is.

Speaker C: So that's, that's Barry's new program. Leaps, the learning, uh, equity and problem solving for the public good. Okay, so here's the story. Barry wants to create a new degree program to help students learn about community education, being learning sciences leaders, but in their community. And he gets permission to offer his major and freshmen admit. But he's told, but you can't do it in Ann Arbor. There's no more. There's no rooms. We don't have enough. We can't accept any more freshmen. So it turns out that Mary Grove College in Detroit, uh, has gone under. But the University of Michigan kind of owns a piece of it. And so he was allowed to start his program there. So students are admitted into leaps and they live for their first.

Speaker D: I should explain Mary Grove College became school, uh, a, uh, high school.

Speaker C: That's part of it.

Speaker D: Okay.

Speaker C: Yeah, those are different buildings.

Speaker D: Okay.

Speaker C: So Barry has a dorm. I mean this is actually his program is incredibly expensive. He's built the dorm, he's built the classroom space, he's had to build kitchens for the cafeteria.

Speaker A: Right.

Speaker C: So he's had to build all this out. Um, and students spend their first year in Detroit taking learning sciences classes. But in addition they work with community ed folks. Um, they can't get all their classes there so they have to spend two days a week bussing back and forth to Ann Arbor to take classes in Ann Arbor, where. But it's a really, you know, way out there thinking kind of a program to say, okay, we're going to start out with kids not living Ann Arbor, instead living in Detroit. There are still, I think it's five nuns living there. Marygrove College was a Catholic college and the nuns have retired and still live on the property and interact with the students. Um, yeah, in addition, as Barbara saying, they also have a preschool through 12. So they basically have kind of a lab school they refer to as a community school. Uh, and the Marcel Family School of Education, that's where they have their students do pre service teaching, student teaching and they do research in those classrooms as well. So yeah, Barry's doing something pretty innovative.

Speaker B: Uh, there's also Rebecca Quintana and Chris Quintana. We have another husband and wife team on sometime here from Michigan. Uh, I think Lydia has a follow up question.

Speaker F: Yeah, thank you for sharing this history and uh, what exciting work and going back to your earlier ah, notion of programming not for the sake of programming but programming for a purpose, to serve a community that's just really, really beautiful. Um, my follow up question is about you know, um, learning to program. I mean like arguably right now in this age that's like saturated with AI, it's, it's been more important to democratize computing education for everybody to learn like what they're grappling with. And I really, really appreciate uh, your lifetime work on um, um, democratizing computing education. So from your perspective as a computer scientist, as a computing educator, um, if someone want to say like I've never touched programming all my life but I want to get started today, whether that's ah, a young learner, whether it's as a teacher, whether it's as a parent, whoever, maybe someone who retired, um, where do they get started? And I, and in terms of starting to learn how to program,

Speaker D: well there's so one of the things I did was create an online course, Python for Everybody, or Python for Creative Expression I think is the name of it on Coursera, um, based on the course that I teach. And so I think that's a nice. If you wanted to learn Python, that's a nice, uh, place to start. Um, there's one of the things we're finding is that a lot of people are turning to online things these days. Uh, if they, you know, if you have. I used to run summer camps and did a lot of teacher workshops. I was a research scientist at Georgia Tech and my job was tried to get more kids interested in computing, more secondary teachers teaching computing. So that was a really fun job to have. And I used to joke that I got parents to let me experience experiment on their children, to pay me to experiment on their children because I would try out all these ideas. What gets kids interested, what keeps them interested, what gets teachers. Because a lot of teachers don't have any background in programming. So how do I take business teachers who've been teaching Word, Excel, PowerPoint and ease them into programming? And there definitely what I found is I tried teaching them Java. Yeah, way too hard. Tried teaching them Python easier. But, you know, and then we tried Alice, which was a 3D drag and drop thing. Thing. Three, uh, D is kind of hard. And then we went to Scratch and they could all do scratch. So for, for, uh, a lot of people, I say, you know, start with something like scratch, drag and drop, snap. You know, if you just want to learn something. I haven't done as much Vibe coding, but, you know, maybe that's something that would be useful for people that just want to get started trying something.

Speaker C: I think my answer would be it depends what you want to do learn programming for. Um, you know, if you wanted to learn programming as a way of understanding science or math, man, I would go with Mathematica or R. The amount of stuff that exists for Mathematica is just stunning. And there's, ah, R is the language of social sciences, the computational lines of social sciences.

Speaker A: That's.

Speaker C: And there's. So there's so much code available to look at if you just want to make apps. Oh, I've got a great idea for an app. I'd love to get an app that goes into the, into the, into the App Store. Um, I would go with Vibe coding for sure. If you wanted to go with art and you want to interact with other artists, you have to learn processing. Uh, all the computational artists tell me that is the language. That's their lingua franca. That's what they talk to one another in. Um, but on the other hand, if you just want to make art, oh, Snap is so much better. Snap is so much easier than processing and the tools that are being created. Well, in particular, I think, uh, y' all may know Glenn Bull at the University of Virginia. He's still making some pretty amazing Snap things. He has a music language called Tunescope that he has implemented all in Snap. It is essentially, uh, everything you could do with a digital audio workbench, but you can program it and with blocks. Ken Kahn, uh, at Cambridge, uh, is doing this absolutely stunning stuff. With AI and snaps, you can actually build all kinds of neural net toys and things and actually play with the pieces and understand how they work. So I really think the answer to Lydia's question is, so what do you want to do with programming? It's like, that's really saying, I want to learn English. Okay, yeah, but are you going on a trip or you're going to try to read Shakespeare? Uh, you know, it's. It's different kinds of skills, different kinds of starting places.

Speaker F: That's really amazing. Thank you for introducing, uh, us to all of the different languages and tools. What about for teacher? If a teacher say, well, all my kids are using AI these days. I want to understand programming, like how AI works. I want to do a deep dive. Where should they start?

Speaker D: Which language is it?

Speaker E: Python?

Speaker D: There are teachers out there. I went to, uh, the Dairyland Conference in Wisconsin, gave a talk, uh, and there are teachers out there playing with AI. So really, um, I don't know that I would love for him to get his generative AI course, uh, online so that teachers could take that. I think that would be an excellent course. Or to get more teachers taking out. There is some work on K12AI stuff. I'm not as up to speed with it, but there are teachers trying it out and helping each other. But I do think that's a place that we need some more focus.

Speaker C: I'll answer the question with a bit of a research story that we're in the middle of. So I don't have any papers to point to. Um, the state of Michigan decided that in 2027, 2028, all high schools in the state must offer a course in computer science. Uh, the Michigan Department of Education was really worried about this because there's no budget involved, there's no funding for training or for hiring new teachers. Um, and then, uh, uh, Cheryl Blanking, Cheryl Wilson, the cs, uh, coordinator for Michigan Department of Education, uh, she got to influence the bill. And she made two really important parts to work with it. First of all, the definition of computer science is really broad and 2 a course can double count. So what she's done this last year is to hire two expert English language arts teachers, Ela, two social studies, two science and two math. And she hired them for a year to build lessons that meet computer science standards but that could fit into their classes. And I grabbed a team of four undergraduates to make things for the teachers. So the teacher would say I'd like to use Sage Modeler, but I'm really having a hard time. I want to do something with photosynthesis. Well, I'd have a student build them some examples. I um, want to use CODAP and I want to do probability stimulation while the student built a simulation of a basketball shooter that would generate numbers for codap. So students building things for the teachers to make it easier for the teachers. Um, and then in the end each team produced two different lesson plans. And my team of students then analyze the lesson plans. So in some sense this is a best case scenario, right? We've got expert teachers, they've got support. I spent two days with them in East Lansing, two full days where I'm answering all their questions, going around, I'm building things for them. I built a gerrymandering simulation that got built in the one of the social studies lesson plans. But here's the issue. Most of them have nothing that we would recognize as programming. Most of them, um, just use things like Google Sheets. Um, most of them, um, one of my students said they're really good at explaining the math, but they're really bad explaining the computer science. And it's like, well, yeah, they're not computer science teachers so they don't know how to teach that part. Uh, so I think that the question is not just where should the teachers start, but what will the teachers actually adopt? What's the trajectory? And maybe you get them started with simpler things like code app and Google Sheets and Sage Modeler and think about what comes next based on their domain and where they want to go to. I mean this is something that we're still figuring out. We're going to do this again next year. I think Cheryl's really interested in us having a try again. Um, and I think that it's an ongoing research question to say what helps teachers to be able to embrace computer science within their courses, in their discipline

Speaker D: now for like, I've been involved for years with advanced placement computer science, so I was on the development committee and then one of the first ebooks I started working on was the what's now called CS Awesome. So it's a free ebook for advanced placement Computer Science A and it is college board endorsed curriculum and lots of teachers use it. So uh, you know, and I put all my stuff in there, you know, we've got Parsons problems in there and all that. So that's an exciting thing for me to see. Something that I kind of joked was my guilty pleasure project when I was supposed to be doing my PhD work or my other work, I would spend time on that one just because I cared about it. One of the things I'm trying to do right now is create a new mooc, a new online course for teachers who uh, want to learn to teach AP Computer Science A. The problem is there's a uh, AP Computer Science uh, CS Principles. This is what it's called. It doesn't require any particular language. You can do it in blocks, you can do it in Python, you can do it whatever you feel like. But that's a big gap from that to Object Oriented programming in Java, which is what CSA is. And so I'm trying to create a MOOC to help teachers and or students who want to go learn the A material which is full CS1 object oriented programming in Java make that transition. So my first MOOC was about my course really aiming at those teachers that have taught CS principles so are familiar with a few ideas in programming but maybe aren't even comfortable with text programming at all. And so I really wanted them to take that online course first and then I'm as a, and maybe teach Python for a little bit and then I'm trying now to transition them to Java, uh, but again doing the kind of thing I do with lots of mixed up code problems and things. I worked a lot with teachers. I worked with over 500 teachers when I was uh, working in Georgia. So I did a lot of teacher professional development and one of the best things I learned over time was to harness the teachers themselves. Right. I'm not a high school teacher so I started working with the high school teachers who were great at making materials and helping them make things and go to conferences. Uh, but in general, yeah, ah, they need a lot of materials is one of the things I learned. They want lesson plans and pacing guides and a whole bunch of stuff that is a lot of work. But I'm trying to add to that right now by doing some AI assisted stuff for learning object Oriented programming in Java.

Speaker B: Well, we've run out of time unfortunately. Here today. Um, I was going to ask, where do you hang out? What conferences would people find you at if they want to be involved in all of this? Is there one or two that you could quickly mention before Lydia is going to introduce the next show?

Speaker D: Uh, this summer we're going to ITSEA in Madrid and icer it is C

Speaker C: is the Innovation in Technology and Computer Science Education. It's the European CS Education Conference and then we're involved in icer, the International Computing Education Research Conference. We're also often at the sigc, the Specialist Group Computer Science Education Technical Symposium. Um, I get to AERA every couple of years. I haven't been to ISLs in eight ages but would love to go back. Uh, I really miss hanging out in that community.

Speaker D: And sometimes my students publish at KAI and also AI and Ed Learning at Scale. Uh, but less often I haven't gone to some of those as much.

Speaker B: This has been wonderful. And your audio turned after the first few minutes has been fabulous. No problems at all. And video too. So it's turned out really great show. We've covered so many topics, so much history, yet so much that's relevant today. So this is an episode that I'm sure many people will be sharing, uh, watching, commenting on. So thank. And we got recommendations for people like Barry Fishman and others to bring for follow up shows. So thank you for that with this popcorn bag. Um, um, so I think uh, it's just a big thank you, uh, Mark, uh, and Barbara, just it's been wonderful to have you with us. Um, but Lydia needs to introduce the next show. Oh sure, yeah.

Speaker D: Thank you.

Speaker F: Thank you Mark and Barbara. Ah, really great show today. Next week, uh, we'll be learning about building democratic schools all around the world from Linda, Nathan, Jonathan and Gus. So stay tuned. Uh, see you next week.

Speaker C: Sam.