How are students using AI right now in their learning and work? To kick off the new season, we sit down with students Enaya, Chong, Falak, Rudy, and Tanvi to hear their perspectives on how AI is shaping the student experience today.
Chapters coming soon!
Credits
Audio editing done by the University of Toronto's Arts and Science Digital Teaching & Learning Studio, located in the Sidney Smith Building at the St. George Campus.
Mario: Welcome to Season 3 of In the Loop! It's good to be back.
Diane: It is! We've had a lot of fun digging into topics that we hope you find interesting and useful, starting with one that is on everyone’s mind, you can't escape it: AI.
Mario: Since AI roared to the forefront of the world’s attention, it has disrupted many aspects of our lives, including teaching and learning, which is my entire life. In this episode, we’ll take a look at how students are using AI right now in their learning and their work.
Diane: It’s incredible what AI can do to support learning, and our students are taking great advantage of that, but they also shared some pitfalls to avoid.
Rudy: Doing one of my assignments, I realized I had no idea what was actually going on.
Rudy: I realized that I can't proceed like this. It's not beneficial to me or my learning if I have no idea what's going on.
Chong: Even though the code works, I still had to go back and understand it.
Falak: You should have the will to learn and only then you can learn
Mario: Before we dig in, let’s meet our guests, undergrads at various points in their computer science degree at the University of Toronto, plus one recent graduate.
Enaya: My name is Enaya. I'm going into my second year at UofT studying CS and English as a double major. I am currently doing an internship at a real estate startup this summer and I'm happy to be here.
Chong: Hello, I'm Chong. I am in computer science and bioinformatics, and I'm heading into my third year. This summer I am working with the Temerty Faculty of Medicine, and I'm doing a studentship with them to develop. Implementation of the machine learning model that my lab has previously developed.
Falak: Hi I'm Falak. I'm going to my fourth year at University of Toronto. I'm specialising in computer science with a focus in AI and minor stats, and I'm currently interning at Accenture as a tech consulting intern.
Rudy: Hi my name is Rudraksh but I go by Rudy all the time. I'm currently completing my ASIP 16 month internship program at UofT and I'll be heading into my fourth year in the Arts and Science Department where I'm specializing in computer science. Currently, I'm interning at Amazon Web Services, which is pretty cool.
Tanvi: Hi. My name is Tanvi, and I recently graduated from U of T with the degree in computer science and I currently work at IBM, a software developer.
Diane: We asked how they're using AI in their learning and course work. No surprise, they're using it to understand concepts. AI can be really good at explaining things.
Tanvi: When I was a student, I used genAI for like understanding basic concepts more like a learning guide. So like if there was a certain topic that I'm not too familiar with or I'm confused about, then I would kind of use it to clarify my understanding
Falak: I think a course I remember very properly is CSC369. I think it was Operating Systems, probably one of the hardest courses I took, and I remember there were like so many concepts like addresses and so many different back end works, so many hardware stuff. And I think what I definitely do is. After all my lectures, I go back home. I try to learn the concepts via AI, get a bit of a background about the concept itself.
Mario: Rudy got specific about what he asks for when he's grappling with difficult concepts.
Rudy: There's often times where I encounter concepts that are difficult for me and it's extremely helpful in breaking down those complex topics into more digestible bites that I can consume with tailored explanations, analogies, visual diagrams, all that good stuff. I can have it generate prerequisite diagrams, or road maps for things that I don't quite understand. When I'm lost, I can be like, what do I need to understand 1st and have it generate the things and have a back and forth with the AI to understand which things I already know, where I'm lacking, how to get to where I want to be.
Diane: So he’s getting concepts broken down and the pieces put into a prerequisite roadmap that can then drive the conversation as he fills in gaps in understanding. And genAI is giving him explanations of the pieces, and also other layers like diagrams, examples, and analogies. I love analogies, Mario. They’re so powerful because they can connect something you’re learning to something you already know, and give you a concise way to remember the new thing.
Mario: Rudy described a specific time when he used genAI to get a roadmap and concrete examples that helped him grasp a tricky concept.
Rudy: And to give an example, for example, when I was struggling to understand auto regressive generation in transformers, I couldn't quite understand what made it auto regressive. So I had it create a couple of prerequisite road maps and walk me through individual examples of how- individual token by token examples of how auto regressive transformers work until the concept just clicked in my head.
Mario: Moving from understanding to doing, students shared a range of ways they use AI to help with their course work. Of course, it’s important to find out what uses of AI are allowed in each specific course.
Diane: I wonder if that will settle into something consistent across courses. The University of Auckland’s Faculty of Science has a policy saying that students must be allowed to use AI for all “unsupervised” work – everything done outside of a classroom or exam room.
Mario: We’ll have to see. Meanwhile, the most likely things to be permitted are to do with understanding what the assignment is asking you to do.
Tanvi: Another thing I used it for was to simplify instructions. So sometimes they're assignments that have too many technical terms or some things are just super confusing, so it was really helpful to simplify what- to explain what needs to be done. Because part of doing anything requires a stronger understanding of what you need to do, and if you have the "what" part clear, it just makes doing the task a lot more easier.
Diane: This is a place where I’d insert some caution. In some courses, learning how to understand the very technical requirements might be one of the learning objectives. In that case, using AI for this may undercut your learning, and may even be disallowed. I also think it’s a spot where AI – at least right now – is more likely to make a mistake.
Mario: In computer science, another part of understanding the assignment is often getting familiar with starter code. AI can help with that.
Falak: When you have a starter code, they're like so many tonnes of files and like so much code already in there, and instead of, you know, going through every line one at the same time like in the very beginning, I tried to just put in the code to AI and like, get a crux of what is going on.
Diane: Once they understand a task, some students use AI to help them get started.
Enaya: So I use genAI a lot for when I'm like, learning how to code or like writing software or doing just projects for my own self. Typically for like scaffolding. Sometimes I don't know how to approach a certain problem or I want to execute something but I don't know where to start and this has happened both in like project based courses but also like in personal projects. So I typically view genAI kind of as like a little agent that I can send into the world of the Internet to kind of show me like, what is the environment like for perhaps the specific task or the specific technologies and it can give me recommendations which I can then use as kind of a blueprint for me to then go in and start the task.
Mario: Enaya gave us a specific example of how genAI scaffolded a task for her during an internship.
Enaya: This summer I was working a lot on front end development and I have never done that before. I've typically just done like back end or like little scripts. So I had no idea how to deploy a website, so stuff like technologies for front end, for deployment, for containerization. All of that. I had no idea how any of that worked. So I asked AI to kind of give me "how would you deploy a website if you were doing it and you didn't know anything?" and it gave me a lot of options.
Diane: GenAI can also do a good job of giving feedback on code you’ve written.
Enaya: If I need to refactor I can use genAI to show me what kind of needs to be changed or what issues might be potentially in the future. Just kind of like as a review of my code.
Mario: We also heard about students using genAI to generate practice problems.
Chong: Another thing I did I remember succinctly was in CSC209, I was confused with the piping and how the command line bash works. So before the final, I just asked to generate like tons of questions for me to practice, and I did them until I got it.
Diane: And students are using those AI-generated problems to help prepare for both tests and interviews.
Tanvi: So when I was studying for interviews and when you have to do a lot of leet code, I would try to do the leet code questions on my own. But there are times when like I'm failing a certain solution and I didn't want to have to read through other people's solution on that website, it's too much, right? So I would give AI my solution and the question, and ask it where it where this is going wrong or like how I could improve this or like make the code run in a certain runtime. And I think that route was really helpful for me because it's like giving me more information about my solution that I developed with my thinking process.
Mario: One student told us that after a test, they used genAI to discuss the answer they gave on a tricky question.
Enaya: Other times, if I'm in like a test and I've like just given an answer and then I leave the room and I'm like was that, was that right? I go back and I like, you know, try and like rehash the question with a large language model and it helps me learn. Maybe a bit too late, but it does help me kind of like get through the parts that I'm not necessarily sure about.
Mario: This is something we definitely didn’t have in school!
Diane: Another whole category of uses is for tasks that feel tedious, and where it seems that offloading the task to genAI won’t in any way affect your learning. This is a clear win. A great example of that is when you have some math that would be tedious to type up in LaTeX or in Word. If you upload an image of the handwritten math, genAI can do the formatting for you. What a gift!
Tanvi: My partner and I, typing in latex, it was super tedious. And so this really worked out well for us.
Diane: I’m encouraging my databases students to do this.
Mario: Some students use genAI to speed up the time-consuming process of wading through documentation.
Chong: I also think genAI is great for, like, saving us some time for reading through documentation. So summary is also very, very good.
Diane: Writing documentation can also be done by genAI. Students should check whether this is encouraged in their particular course.
Falak: Personally I find documentation, writing documentation a bit more, you know, less of a hard work, less of a brain power work, so I'll definitely use AI to, you know, write such redundant documents and stuff like that. So it knows everything, so it's going to just write our good new nice document.
Mario: The last category we saw was all about getting your thoughts together and writing them down.
Enaya: I use it a lot for ideating. GenAI is really good at coming up with ideas, even though it sounds it sounds like it wouldn't cause it's probability, but it does help me like kind of think outside the box and think of things that I hadn't thought of before. Not as in using all the ideas that genAI gives you, but using those ideas as kind of like a launchpad to think more about and think critically about.
Diane: What do you think about this use of AI, Mario?
Mario: Yeah, I mean, I think I've done it as well. I've used AI to help me ideate.
Rudy: When it comes to organizing my thoughts for both learning and coursework, I can sometimes brain dump everything I have like this jumble in my head and have it organise those thoughts into me. Which makes it easier for me to understand the things that I'm thinking.
Mario: What about this one?
Diane: Yeah, I have two reactions. On one hand it sounds like a really good way to get stuff organized. But on the other hand, I know that organizing ideas yourself in your mind is how we get things into long term memory so we can retain them.
Mario: Students also talked about using AI to help with putting their ideas into a polished form.
Tanvi: I normally use it for like writing tasks. I would just give it an idea of what I want to have written with not like proper English, grammatical structure and whatnot, and it would nicely organize the sentences, make sure the grammar the punctuation is correct.
Diane: I think this use of AI is tricky. If you’re organizing bullet points into a professional and polished email, there doesn’t seem to be anything to lose. On the other hand, if you’re writing something substantive to express ideas, writing can be an important part of the thinking, a way of working out your ideas, and you wouldn’t want to deprive yourself of that, depending on the circumstances.
Mario: So, we’ve heard about how students are using genAI and offered a few of our own cautionary notes. The students also had plenty to say about potential pitfalls and how to avoid them.
Diane: An obvious pitfall is that genAI currently – as of late summer 2025 anyway! -- can provide incorrect information and do it with great confidence.
Falak: AI has a lot of misinformation. So as I mentioned before, I suck at writing essays, so I usually go like, you know what, give me all the references and like give me the content. Turns out it gives out nice, blue links there that oh this directs you to a certain page and it's like 98% of the time those links are incorrect. So it's like this too much misinformation out there. Thankfully, I learned that way early on and I know not to trust AI at the face value. So I do double check that.
Mario: Sometimes this can be a huge waste of time!
Falak: Just like an hour back, my supervisor and I were researching something and again, there's so much misinformation and false information that you might never end up getting an answer. So I remember, like an hour back we went like back and forth 7 times on something and we could not find the answer. My supervisor is like go to Google. And try searching there. I went to Google, I wrote a single sentence. It was the first official website link. I clicked there and there was right the answer.
Diane: It’s not difficult to identify links that go nowhere or references that don’t actually exists, but it’s harder to identify misinformation in an explanation, especially when AI is explaining something that you are just learning.
Tanvi: When you're, let's say, using it to learn something new. And that means you have no information about the topic, right? You don't know what inform like - whatever responses you get - you don't know how real that is, or how true, how exact that is, and it could be very misleading. Because generally the responses of any AI tool sound very confident. But it may not be the correct thing.
Mario: Another problem raised is that AI isn’t always great at finding bugs – again, at least for now.
Chong: I found that I have to identify the bug first. If I don't, it will just go in like, a negative self-feedback loop. It would just hallucinate and it would never find the bug. Yeah. So it's up to myself.
Diane: So far, these pitfalls are due to failures of the AI itself. But every student we spoke with raised the issue of how we use AI – and the risk of using it in a way that undermines learning. The temptation is vast.
Falak: World is moving so fast, you have to deliver features, you have to deliver your work, there are deadlines, and if you get caught in all that you know, I know for a fact you're just gonna, you know, get things done with AI as soon as possible. And you'll forget to learn.
Mario: And sometimes when you ask for an explanation, AI will offer you a full solution.
Enaya: It would go a bit too far with its recommendations. Try and do the entire thing for me. At that point, I think you just have to be critical and realise that you're not gonna learn anything. If it does it for you. You have to think and you have to try it and get messy with it first.
Diane: We heard many times that it is hard to resist over-relying on AI even though it may come at the expense of learning.
Tanvi: GenAI it's very powerful, right? Like you could you could get basically answers to anything you want which also means that it eradicates the whole process of going through actual resources and understanding the depth of the topic, when you can just get an answer straight up. So I've been trying to like not rely much on it, especially when when you're learning something new.
Enaya: So a pitfall that I have both avoided and not avoided, and a pitfall that I think a lot of people can relate to is: reliance. When you start using AI it's addicting and you don't want to stop. You don't want to think. You don't want to use your energy on something that you don't think you need to, which I think is fair. You don't you want to use it as a tool that will push you forward and not hold you back. So yeah, but I think of it kind of as like a dopamine hit of information. When you ask it for information or do you ask it to do something for you? It it feels good, but it's cheap dopamine. You're not learning anything.
Mario: Chong and Falak shared specific examples of when their learning was impacted by over-reliance on AI.
Chong: I have asked that to write a cookie for the website, and I didn't actually know how a cookie worked, but it just it just worked. So I didn't bother debugging it. But because I didn't debug it, I didn't actually get to learn it. And then so after that, after I finished that task, I actually went back and looked up how the cookie works. I realized I actually didn't understand it. I just, I just yeah. So that is a very common pitfall. Like even though the code works, I still had to go back and understand it.
Falak: You should have the will to learn and only then you can learn. So I remember I wrote this one essay. It was a fantastic essay. No one could tell that oh, it wasn't written by me or it was written by AI. And I think I submitted it, got a great mark on it and later when it came to an exam, I realized, wait, no, I don't even know what it is about. Like, I don't know anything about it. And after that I talked to my professor. I came clean. I'm like, you know what? I want to redo the essay so that while I'm redoing it, I at least can understand what it is.
Mario: When Chong saw that his learning was suffering, he limited his use of genAI to help solve assignments.
Chong: I actually outside of courses a ton and in this summer internship. But during the school year, I actually try to refrain to use them as much as possible when I'm doing the assignments. Even if I use them and I got the correct like perfect grade, I would still not understand the actual concepts and didn't get to practice them before, either for my exam or for my own good, because I wouldn't be able to do an exam and I didn't actually learn the concepts.
Diane: Rudy described how he realized the problem and changed how he used genAI in response.
Rudy: My use of genAI for learning and coursework is something that's evolved over time. Initially when I had just started using it, it was a shortcut to almost speed through all my coursework and learning. But at some point I realized that this wasn't really actual learning, it was just me using it as a way to speed through my work instead of actually assimilating that information with all the other stuff that I know in CS. And that led to me pivoting to using it as a learning copilot. So now instead of just using it to complete assignments, I use it to understand concepts at varying levels of difficulty. I treat it as a teaching assistant rather than a homework doing tool.
Mario: We'll wrap this section up with some sage advice from him about how to protect your learning and future-proof yourself at the same time.
Rudy: In addition to completing the assignment I need to understand what I've learned. I need to be able to explain those concepts in my own terminology, my own words, in my own thoughts so that I can re-explain them or reuse them to answer questions on midterms and exams. So if I just simply copy paste and get the assignment done and look at it as a one and done thing, I'm not really helping myself right? I'm creating bottlenecks for myself in the future where I will struggle.
Rudy: As we've seen in the past, like 6 months, if anything like models are getting smarter at an insane speed, right. There's no doubt that just two or three years down the line models will reach the level of skill needed to replace many of us in our jobs. And that is something that is a huge concern for me is I want to be able to future proof myself and my skills in a way that I don't get replaced by AI in the future. So if the way I look at is, if I don't build the fundamental skills needed today to be able to represent myself as an orchestrator of AI rather than just a user, I will struggle a lot in the future to put it like very lightly, right? So I need to focus on developing deep knowledge of the best practices, approaches and conventions now so that I can verify things in the future. I can orchestrate and have the AI do what I wanted to do rather than passively let it do its thing.
Diane: We invited our guests to add any further comments on the use of AI. Privacy was a big theme.
Falak: As I mentioned, I'm interning, and obviously we have clients, there's so much client information, so I believe putting in that client information can be like detrimental. So it's like there's so much sensitive information out there and you have to be so careful with what you're, you know, putting into AI.
Mario: Students mentioned their own privacy as well.
Diane: I’m glad to know that they are thinking about that. There was a time when many of us didn’t think twice about giving out our information -- “I’ve got nothing to hide”.
Tanvi: A few months back there was this whole Instagram or social media trend where people would ask chatGPT, specifically, giving a prompt like based on what you know about me, draw an image of what my day looks like and I tried it myself. And it was pretty accurate. It knew I went to UofT, what major I was. It knew I was a girl and I don't think I've explicitly ever provided such personal information about myself, but it somehow still knew, maybe based on the tone of my writing, or I really don't know. So that's kind of concerning.
Chong: One of the big thing is privacy. I think even during, yeah, during my work, I'm working with patient data. So it's very important for me to not share those with ChatGPT. Another thing is about my own privacy is that I'm scared that activity is going to get a too good of a picture of me, since I would ask it anything. I'll ask about vacations, I'll ask about my hobbies. Recently there was a moment where it just knew my name in a new chat. Like I didn't even ask it. And it's just like, hey, Chong, do you need help with this? And then I was like, well, I didn't even tell you my name. How did you know? And then when they when the response came back, it said that all the chats I actually backed up and they actually are fully connected. And it does spook me out a bit because I didn't know. I thought each chat was its own entity and couldn't refer to the other chats, but then it's actually all fully connected.
Diane: Several students also raised issues related to broader impacts on society. I was so interested to hear Enaya’s thoughts on creativity.
Enaya: Can AI be an artist? Can AI write? If AI is being trained on data that was copyrighted, how is that? How is that fair? But also what is the future of creation. I think there's a Picasso quote that all art is just stealing or good artists steal something like that. I wonder if there are limits to that, but I also wonder if they're if we're going to be using a tool that has been trained on everything, all language that ever was, where that goes, where that can take us with our creativity.
Mario: And Chong argued that we need to be thinking about regulating AI.
Chong: We actually have to know how to use genAI to be able to control it, to be able to, put regulations on it and frame it in a way that is beneficial for humanity and society in general. I know that sounds really like abstract and out of reach, but I think we're at a point in history where it's actually important to really think about, like what regulations we need to put on it, whether in terms of our own use and in terms of its impacts.
Mario: These were some very thoughtful remarks about using genAI in learning and coursework.
Diane: Yes, everyone has been talking about AI incessantly it seems. I really enjoyed hearing from our students on this burning topic.
Mario: You know, regardless of what you've heard in this episode, the first thing you should do before relying on AI for a course is check your syllabus. Some courses may not allow AI at all. Other courses may do so depending on the assessment, right. You really need to check that syllabus to make sure that you're not committing any kind of academic offense.
Diane: And ask for clarification if it isn't clear. And I'd like to also add in the benefits of using the UofT provided platform for AI over other systemns. Currently its Copilot. Couple of really big benefits. One, it's free and has no limits on your use and also your data stays within the university ecosystem and is not used to train any models. And this means it's okay to share even copyrighted course material with the UofT platform, which wouldn't be appropriate if you're going to upload it to ChatGPT or something.
Mario: Right.
Mario: What a great episode! Stay tuned for episode 2, where we'll crossover to the other side of the classroom and hear from two faculty experts, plus us, on teaching and learning with AI.
Mario: My name is Mario Badr.
Diane: And I'm Diane Horton.
Mario & Diane: And you are in the loop.