AI’s Messy Middle: Students, Learning, and Agency

Sophie White: Hey everyone. This is a special episode of EDUCAUSE Shop Talk. We recorded it live at the EDUCAUSE Annual Conference in Denver, and in it, Jenay and I talked with Lance Eaton and Agni Athrea, an undergraduate student about AI adoption and what it means for students, faculty, staff, and higher education as a whole. It was a really dynamic discussion. I think you'll enjoy some of the conversation at the end that made me think students aren't as enthusiastic about AI as sometimes we may think. But we really had a lot of great discussions about where we are in this messy middle of AI adoption, how to look at the digital divide, differences in equity and equal access to AI at our institutions, and ultimately how we can have agency in shaping AI adoption across higher ed. I really enjoyed this discussion and I hope that you do too.

***

Sophie White: Hello everyone, and welcome to the EDUCAUSE Shop Talk podcast. I'm Sophie White. I'm a senior manager of program and content marketing here at EDUCAUSE and one of your hosts for today's podcast discussion.

Jenay Robert: My name is Jenay Robert. I am a research manager at EDUCAUSE, and I'll be your co-host today.

Sophie White: Great. We are really excited we're recording this episode live at the EDUCAUSE Annual Conference, and it's a really timely topic on students and AI. At this conference, I think we've been hearing a lot about AI and we are thrilled to have with us today experts on AI and student experiences and actually a student with us today to talk about real experiences that students are having with AI. So I'll introduce them and then we will jump into it. First off, we have Lance Eaton. Dr. Lance Eaton is the senior associate director of AI in teaching and learning at Northeastern University. As a leading voice in higher education around generative AI and its implications for teaching and learning, Lance has published articles, consulted and presented at over one hundred colleges and universities, as well as served as facilitator in programs on AI and education with EDUCAUSE, AAC&U, and NERCOMP. Thanks for being here, Lance.

Lance Eaton: Super excited, and thank you.

Sophie White: Great. Lance also won our, what did you win? The leadership award?

Lance Eaton: Yes, the 2026 EDUCAUSE Leadership Award.

Sophie White: Yes. Congratulations and thanks for spending time with us. Great. Next up, we have Agni Athrea. Agni is a graduating senior from the University of Georgia, majoring in computer science with an emphasis in cybersecurity. They currently direct the advanced research in exploitation and security lab at UGA and work as a SOC analyst intern at the Georgia Tech Research Institute with research spanning AI security, data governance, and environmental justice. Agni is a co-author on the power of AI, Balancing Ethics and Harm for the Environmental Justice Field, currently under review at Springer's AI and Ethics and has presented research at the Curo Symposium and the National Conference on Undergraduate Research, earning the 2026 Kiro Best Paper Award in Physical and Environmental Sciences. They're especially interested in how AI intersects with the communities and systems it's meant to serve and where it falls short. Thanks for being with us, Agni.

Agni Athreya: No worries.

Sophie White: Great. And what is CURO, the symposium?

Agni Athreya: Oh, CURO is UGA's Center for Undergraduate Research Opportunities.

Sophie White: Okay, very cool. Well, congrats on being so involved.

Agni Athreya: High dry, high dry.

Sophie White: Great. So as I said, we're talking about student experiences with AI today, and I think this is a really relevant topic. For those of us who are joining us live in Denver, we saw the EDUCAUSE Top 10 presentation today. I think it was issue number two was related to preparing students for an uncertain future, thinking about I think how AI falls into this whole conversation around the future and our current world is changing so fast. How are we preparing students to really be engaged citizens, workers, contributors in this world that we find ourselves in? I guess I'm curious maybe to start there of how as educators can we leverage the power of AI, but also take into account the fact that we don't know what the future holds, what are faculty doing about this? What do you all think about this issue?

Lance Eaton: I think one challenge is just we're talking about it as the future and we're also in academic year five of AI being around. And so that discourse, the challenge of wanting to try to figure it out and figure it out safely while at the same time there's this counterpart of we don't want to do anything with it and therefore we're not going to do anything with it. And here we are five years into it. And the tension for me is about this, at what point do we genuinely embrace the fact that it's here and the implications from that and have to figure it out versus the not liking it. And I say this because I'm not a fan. I'm not a "Yay AI." My go-to line is I have seventeen different feelings about it before I have coffee in the morning and I have coffee at 5 AM.

It is a complicated thing, but we're also supposed to be academia, which means we actually wrestle with it. And I think for me, the thing is we can't be armchair wrestlers. We have to actually start figuring it out and playing with it. And there's some concern about that, but at five years out, there's some concern about not doing that or not doing that well or effectively or consistently in our institutions.

Sophie White: Great question. Jenay, I know you've done a lot of thinking on where we have agency in how AI is created.

Jenay Robert: Yeah. And Lance, I really appreciated you saying seventeen feelings before coffee. I think that really gives a good summary of how many of us feel. What comes to mind for me is our keynote talk yesterday with Heather McGowan when she was talking about how we are really in the middle of this transformation. She had a great visual where it was like, we're here now, we're going to be somewhere else later, but this is the messy middle. And that really hit me too. I think it compliments this seventeen feelings idea as well. We don't have to have everything figured out yet. We can't have everything figured out yet. And the whole point is for us to be with that discomfort and that really sucks. Can I say that on this podcast? I'm a co-host. I can say whatever I want. It really sucks to have to be so uncertain.

But to your point, Sophie, what I really like to talk about is our agency in this whole process. And so who is going to create the future if not us? And yes, I know we can't control what all the tech companies are doing and what everybody is doing, but when it comes to higher education, we are the ones creating that future. Our students especially are the ones creating that future. And I really believe very strongly in that agency as long as it's done with us taking that power and being really purposeful about it.

Lance Eaton: Can I add one piece to that? One of the things I struggle with at this point in it is that that agency is true and there's a way in which for the first three or four years we gave that up and we gave that up to the narrative of AI is here to help students cheat or AI is here in all these harmful ways. And it's not that portions of that aren't truth, but we decided we let that narrative stick rather than the counter narrative of what can we do with this? How can we use instead of AI as an answering machine, AI as an interlocutor, AI as an interviewer, AI as something to make us struggle? And I think we're not just having to continue to figure it out because not enough people are, but we're having to also work against those narratives that now are just everybody's perception about the only way to think about AI.

Agni Athreya: I have something to add there with, I generally have a really positive, sort of positive outlook on AI. And one that we do in my class, one of my classes intro to cybersecurity taught by Dr. Zion Basque who created Pawn College, which is a site where you go when you step by step in different modules, learn different hacking skills. Always our quizzes are, you have to hack, you get a challenge day of and do it live and we are all going to use an AI model. However, it's a model, it's called SenseAI, and it doesn't give you the answer, it gives you the right questions. And so it's a really useful tool to use to think through your process better, especially when you're stuck in a rut. And so I've seen many different ways AI's implemented like that, that it's like, no, it's not just for cheating, it's not just for skirting around learning. In fact, it's about learning in a different way, maybe a more in hands-on way even.

Sophie White: That's a great point, Agni. And I'm curious, what do you think about what agency students have and what are you and your friends talking about when you think about AI? You're like, "We want to be in charge of it, we want to control it, or it's controlling us." What does that look like for you? That's a great example of how you use it in class.

Agni Athreya: I think that majority of the people around me are, you do have to learn. You can't just let the model do whatever. Also, you have to learn how do I train a model well? What is a hallucination and what is not? Especially because I'm in cyber, a lot of what I do is bug bounty hunting and hacking. You need to know how to set up the model well and control the agent because if it escapes its container and it starts running exploits it shouldn't be, you're liable. You could fully get arrested for that kind of thing. And also a big part of it is efficiency and token use. We want to learn how much can I do quicker, faster, and cheaper by myself without the use of AI so that the whole process becomes more efficient. AI does not necessarily on its own increase efficiency.

You can easily be talking to a void forever. You need to know what is the right direction of where to look, and that takes genuine learning. And so for me, I don't feel, even though I started college at the start of the AI boom and now I'm here, that I don't think it's hindered my learning at all, except just given me a different avenue than before.

Jenay Robert: This is a good time to ask Lance what you think about it in terms of students realizing what we still need to learn, we still want to learn. From a faculty perspective, how do faculty build those skills to be able to support the critical thinking development, that really meaningful student learning, and then figuring out where AI fits into all of this?

Lance Eaton: Yeah. And I have to say the thing that I feel slightly dirty saying, and it's arable, preface that. I think the only way faculty can really start to genuinely do that is by having access to a model, is by having access to the tool to play with safely and to figure out and test things out with their students safely. And so that means, okay, we have a lot of concerns about the state of AI and the companies and the leaders and the way those leaders are intersecting with politics and all of that. And I'm also saying, yep, and we've got to throw money at those companies in higher ed when there's no money around. So I don't feel great saying that, but I don't know that there's any real good way of figuring this out because otherwise everybody's playing on the free models, which there goes privacy, there goes safety, there goes any way of really trying to do it well and consistently.

And to me, that's the first thing that needs to happen. Obviously, you need policy and you need a clear guidance from the institution about how they understand AI in relation to the mission, in relation to the different domains. What does AI mean for your faculty, for your students, for your staff? So you need those things before you can really, and this is again where I think we've seen some of the stagnation happen is until you have that, you can't even experiment to figure out what is and isn't working in your domain. And then you can start to figure out, and you're going to come up with different ways that shows up. And for me, the thing I've been thinking about a lot is the other piece within this of how do we figure out the classroom is the unit of change for figuring out AI in education.

The unit of change is the discipline because if we think about what we've been doing in higher ed for the last fifteen years, we'd like to think longer, but we've been using this model of backwards design and backwards design for those that are less familiar is you start with the goals and then you figure out the assessments and then you figure out what are the learning materials and activities that get you to do the assessments well that then demonstrate the outcomes. And it's all this deeply intertwined web, and that happens on the course level and ideally it happens at the program level. So we're asking faculty to do a bunch of changing in their classroom, but changing to what? Well, where is the discipline going and what are they figuring out in terms of how is AI impacting every discipline? Because it will, just like the internet changed how we all do all of our disciplines, so is AI and it's clearer in some fields than others.

Jenay Robert: Yeah, I really like that. And it, again, comes back to the messy middle. We are figuring this out and that takes quite a bit of courage and discomfort.

Sophie White: I'm curious, one important point I think is what you just brought up about the money that's needed to do this type of sandbox exercise to be able to give people access safely. Agdi, you brought up tokens, being efficient with those. I feel like we're in a weird time right now where we're starting to see these free models scaled up and these paid versions kind of being the top of the line. We're also seeing the token expenses for various different uses increase. And then we're thinking about a digital divide in higher ed, which we have been for a while. Different institutions have different resource levels. So curious what you all think about how do we as a collective community deal with this issue of disparity in resources as it relates to AI?

Lance Eaton: No, kidding.

Sophie White: I know that's not an easy question.

Jenay Robert: I can kick it off.

Sophie White: You can kick it off.

Jenay Robert: So yeah, we've done a bit of research on it. I have some good news and some bad news. So the bad news is not a lot of people are spending a lot of time actually meaningfully measuring the value that they're getting from the AI tools that they're implementing. The good news is that it is top of mind for a lot of institutions and for a large portion of our community. And it's not just on the minds of higher ed leaders, it's also on the minds of leaders within the ed tech industry. So just Monday of the conference this week, I was at our annual partner summit at EDUCAUSE. We host an annual day for people from higher ed institutions, for people from ed tech companies and people from other organizations like EDUCAUSE to come together and tackle whatever problem it is that year. And of course, AI was a big part of the conversation this year.

Shocking. Yeah, I know. Who knew? And this was one of the main topics of conversation. So not only how do we figure out how to measure the impact these tools are having or the value that we're getting from them, but what should those pricing models look like? And I can tell you because of those conversations I had on Monday, this is definitely something these companies are trying to figure out. They're still in the messy middle too. And so they realize that these tools are quite expensive for them to host and train and maintain. And at the same time, higher ed needs to have access to these tools. We can't afford those price tags. So the best part of this is just that we're having those conversations across the silos, or as John O'Brien likes to say, cylinders of excellence, but we're trying to figure it out.

That's the best news I can bring, unfortunately.

Lance Eaton: Oh, go ahead.

Agni Athreya: I also wanted to say that, okay, I'm lucky enough to go to a school where a lot of this is, there's ways to pay for it and ways to be provided for you. So take that with a grain of salt. But also that I think brings to come back to my earlier point of it's so important to actually learn these skills and be able to apply them because there are so many examples that I've seen in my personal life that the $20 cloud model, if you're using it properly, can be just as effective as say the hundred dollar codex. If you know where to look, you're doing a lot of buy hand and giving it a direction. And so there's definitely ways to be efficient with whatever resources you do have and make it work regardless. I'm trying to butt a positive spin and I'll be all doom and gloom here because there's definitely ways to get to where you need to go in the end.

Lance Eaton: I would add to what you're saying, Jenay, we have models for this. There was a point at which no campuses had wireless across their campus. There was a point in which this idea of paying for the office suite for every faculty, staff, and student was ludicrous. And to your point, it took some time both for the institutions to plan for that affordance. And IRED has always been cash strapped, so it's not like it was any better. Or again, we could say there was whatever, but it's always been a struggle. So it's not like they weren't also having to figure out the priorities around it. And I think that's where we are, back to that messy middle. So I think that's also a thing is, I think for me, I would've wished for these conversations happening in years one through three instead of four and five, because it means by the time we're getting to institutions have built up that, it's like now we've got more range of tools in generative AI and agentic AI to figure out and play with and stuff.

And then to the larger inequity of digital access, the digital divide, I mean, I don't know that we can solve that in higher ed. I think there's countries that have solved it. I believe it, is it Latvia where they have universal digital access, but that's an important thing. It's a bigger, larger societal issue. And I think that's the other piece with AI in general is there's so many different issues about it, but which ones are actually the issues that higher ed need to navigate and be figuring out versus the community stuff, versus the legislature stuff, versus those pieces? I'm not trying to say like, "Oh, well, we shouldn't care," but we're never going to have a hundred percent digital equity at our institutions. They're not structured like that because -

Jenay Robert: The societal backdrop.

Lance Eaton: Yeah.

Jenay Robert: I do think we do a great job though of at least in these conversations that we have at the conference this week, I've been hearing a lot about how do we at least mitigate that. We all recognize that this is happening in a society in a time where these inequities are much bigger than us, and we can figure out what that mitigation looks like. We actually, this podcast, we did an episode with the California State University system that did a system-wide rollout of ChatGPT. It was actually the largest deployment. If you haven't heard or seen this episode, definitely encourage you to go dig back into that. It was actually the largest deployment of that technology. I don't know if it still is, but at the time it was so far the largest one. And one of the main drivers of that initiative was this digital divide issue.

So some of their campuses perhaps could afford this, some of their campuses couldn't, but they could mitigate that by working together as a system to deploy the tool across all of the campuses. And I think we can see similar partnerships cropping up at events like EDUCAUSE, in Consortia, in lots of other ways where it doesn't have to be a system, it can be institutions working together. We're seeing a lot of interesting business models cropping up these days, people collaborating on let's share an HR unit, let's share an accounting firm. I think we could definitely see that happening in this space too.

Lance Eaton: I believe the UMass system is now doing a consortium model with AI access as well. And I think those are interesting, helpful models, again, to start to address that.

Sophie White: Yeah, for sure. I'm curious from thinking about individual institutions, we know that they have different resource levels. Agni, I'm curious from your perspective, what is the expectation in your classes right now of what access to specific AI tools you're supposed to have? So what do faculty expect you to have access to in order to do your work as a student in October 2026?

Jenay Robert: That's a great question, Sophie.

Sophie White: I just thought of it.

Agni Athreya: I think most of my classes were not necessarily in my class learning how to best use AI. They've more adapted to, "Hey, you can use it, but if you're going to use it, we want you to disclose it in these ways or use in these specific ways." So giving a little bit of direction as opposed to before and it's like, no AI, whatever, because they're like, "People are going to do it, so let's maybe put them in a better direction." I can't say that's necessarily where I'm learning how to navigate all these tools from, as opposed to my personal research, work with my lab, or I do a lot of CTFs. Those are capture the flag competitions, cybersecurity competitions, which test your skills in reverse engineering, binary exploitation, all those things. So I've used these a lot there. And so working with faculty mentors on that kind of work is what really has exposed me to understand these different models much better.

And also I think that has a larger scope and better application necessarily like a lot of classes do. A lot of my classes are in computer science, the fundamentals about what is a computer architecture, what is this, what is that? I mean, it's important, but limits to the cutting edge technology. We're learning about how do bites work? How does encoding work? I think you really learn this through an actual project, putting hands to keyboard. So I think that's the most important above all, even beyond just figuring out how to use AI. I think with how the job market is going now, just going to class is enough. You have to be trying to experiment with these tools and going out in the real world and making a place out there for yourself.

Sophie White: So I'm curious, thinking about workforce prospects, so you're a senior at UGA, so how are you feeling about what you want to do after college? How much do you think AI will be a required skill in what you hope to do afterward? I'd love to hear your perspective.

Agni Athreya: Yeah, yeah, yeah. I hope I want to keep working in cybersecurity. Dream job would be some kind of cybersecurity analyst. I really enjoy all the SOC work that I'm doing, that's security operation center, a lot of threat hunting and threat mitigation. So getting to continue that would be awesome. But I'm open to, with how the job market is, getting any job is going to be a huge feat. So I'll take network engineer, I'll take solutions architect, I'll take any of them. AI will definitely be a requirement to be able to both understand and realize well. Also that it's deployed in so many of the tools that I use in my da-to-day. So at Georgia Tech, GTRI, really, those are sort of two different institutions, so get on you about it. Every day I'm using Splunk, CrowdStrike, Tenable, Darktrace, Sixonius, Census, bunch of things. And all of those have, well, many of them have an AI component to it that helps with managing alerts and seeing which ones are actually likely to be true positives, which ones are actually important or not important.

And so I think as you're learning more and more tools to have more of a leg up in this industry, you're going to learn how to use AI with them because all those tools are implementing them now. So it's going to be a requirement.

Jenay Robert: Kind of on the back of that, I'd love to ask both of you this question from the faculty perspective and then from the student perspective, what is your take on, we've talked a lot about being in the messy middle and how we have the agency to create the future. What can be done at higher ed institutions to support students to lead the way in the creation of the future? I like to describe it as our students are five minutes away from being our colleagues. I mean, they are creating the future with us. So what can we do to support them? What could be done that would make you feel supported in that? Take it away, whoever wants to.

Lance Eaton: I think half the time when I talk with students yourself included, it's not five minutes away from being colleagues. It's like five minutes away from being supervisors to us. I've talked with a lot of students over the years and it's throughout even this, and it's really powerful to see when you're leaning in and listening, the ways you're really thinking about it is powerful. I'm going to go back to something that I did right at the start of this, and I'm really excited because a good friend and colleague is in the crowd and I'm going to embarrass her a little bit. A rich conversation with Autumn Keynes as we were trying to figure out an incident, and this is in December of 2022, of AI had already been out and some students were... We had a possibility that students were using it in a class, and it was just a rich conversation where it very easily could have turned into, "Aha, let's catch them." And it was like, "Well, what can we learn from them?" And that conversation in many more sense led to the institution I was at at the time in January 2023, running a course on AI and education with students with the goal being let's figure this out and let's develop our institutional policy around AI for faculty and students.

And so we ran that course twice. I learned so much from those students. So to that point, any room I'm in with faculty, anytime I'm doing workshops, it's always like I do this activity where I'm like, "How many people have had conversations with their colleagues about AI in the last four or five years?" And all their hands raise up. And I'm like, "How many people have had 50 conversations? All their hands raise up." And then I say, "How many of you have had the same kind of conversations with your students about AI?" Not, "Here's the policy," not, "Do this," but I actually talked with them about what it looks like in their discipline, and it's a fraction, it's a tenth of the hands. And then I'm like, "How many have had more than ten conversations?" And it's very few people. And I'm like, "We're in way higher frequency of conversation with our students than with our peers." And for three, four years in that you're still not engaging with students about this, just that is the low bar hanging fruit you could do to start figuring it out because students, they're trying to figure it out on their own and they're navigating curriculum roulette of this class is framing AI in this way, whether it's accurate or not, and this class has asked me to do this or not do this with AI.

So I think to me, that's a big piece is open up that conversation. The thing that I did at College Unbound, you don't have to do that, but where are the students in the conversations in the classroom? Bring them into curriculum, bring them into committee meetings, and trying to cycle, not just that one student to represent them all, but trying to bring in mixtures of students. We do a lot of student panels I've been involved in getting together because it just changed the way people are seeing it.

Agni Athreya: No, for sure. And I think, well, to answer the question, the same answer I had before of just how important real world experience is. And there's definitely ways to do that both in the classroom and outside of it. I think in the classroom, not just acknowledging that AI is a thing and that's going to really change how we do all these tasks, but teaching people how do we navigate that? For example, say you're taking a software engineering class. I think right now my school's at a point where don't use it or if you're going to use AI, then just say you're using AI to vibe code your site. Cool. But I'm like, well, we took it a step further and the class is about, so you're vibe coding the site, right? There's so many pitfalls with vibe coding. Is your AI dumping all the API keys?

Is your input for whatever login site or your financial portal, is it taking unsanitized input? Can you do cross-hite scripting on this? Because there's so much the AI is absolutely missing and you need people who are trained, really good software engineers who I can debug and read the code and understand what it's doing. And so I think looking at the issue really in its eyes and adapting to what does this work look like in the modern day would be huge. And also for me personally, the support I've gotten from my faculty mentors on the kind of research I do has so drastically changed everything. And because of them, I feel genuinely adequately prepared that I could go into the workforce and I would be okay and know how to handle myself. And I think I'm qualified enough to be in a room. And so I want to just convey to all students out there, get involved in research.

I promise it's not that scary. I promise that there is something interesting out there to do for every single person and there are so many people out there willing to help guide you through it. And there's truly no better way to learn.

Jenay Robert: That's a mic drop moment right there.

Lance Eaton: We're done. Have a great day.

Sophie White: Beautifully said. I love that take on community and I think that's really then resounding throughout this whole EDUCAUSE Annual Conference as we think about how to navigate this space in AI. We will open it up to audience questions in a few minutes, but I wanted to give you a second, Lance. When we were talking about this call beforehand, you said, "I think all of us are not worried enough about agentic AI and we need to talk about it in this podcast." So I'm curious, do you two want to take a minute to talk about where we are in this current AI conversation and how we should be thinking about bringing students into this era of maybe AI agents that we find ourselves in?

Lance Eaton: Sure. So I mean, I started this year with doing some work in different places of just trying to push the conversation because we're in academic year five of generative AI. We're an academic year two of agentic AI. And having been in through that first cycle, I think it's Mark Twain, or at least I read it on the internet, so it must be true, history doesn't repeat, but it rhymes and it's feeling like we're rhyming with what we did with generative AI and not really paying attention to it as we're here in year two. And I mean, it is generative AI on steroids and some of the things that it can do and how it does continue to require new and different ways of thinking. So there's the figuring out on the technical level, but one of the things I'm racking my brain about is there's this new...

I'm finding myself circling around this idea of new ways to think about how I think to interact with these tools that I will not say the tools think, but they mimic thinking. And this technological metacognitive work, it's a whole other thing, but it's also just right here. And when I say right here, we saw Einstein AI, for those that were familiar, pop up in early spring, which was you could pay this company forty bucks and it could go in and take all of your online courses in Canvas or whatever LMS, but you can now do that with Claude Cowork. So it's like the call is coming from inside the house. They don't even have to go find some sketchy tool. They can do that with the $20 a month member of fee that they're paying to whichever AI tool they might be paying. So I worry about it because we're going to fall into the same trap of we're only going to be able to see it as cheating and security issues and that's valid.

But what we also need to do is start to think about where can this set of new tools and practices create different types of learning or how can we look at this for what's the learning value? It can't be just efficiency. I'm not saying, oh, we can do things more efficient. Let's see. No. How does this or what can this do to change our learning? So that's a thing I'm just chewing on a lot these days because yeah, that's where we are.

Agni Athreya: I think agentic AI, from what I've experimented with it, it's similar, but it's a different skillset than just the regular AI. You have to play around with it to see, but there is definitely so much learning creating your own methodology around it. And how do you, again, do this work most, not efficiently, but effectively, not just for minimizing your tokens, but also for how do I get it to give me the end result that I'm looking for? Because it can easily spin up eleven sub-processes that do nothing. You actually have to spend a decent amount of time to understand how do I manage this whole system? And you can do a lot of really incredible things with it if you know how. And so I think it's definitely on the topic that we should ignore and encourage people to learn, okay, yes, can there be security issues with it? For sure. How do you mitigate that? What's a Docker container? How do you secure the Docker container? Multiple, multiple things that I wish you were talking more about in universities because that's the real life work that we're doing as opposed to just ignoring the problem. I can talk about the stuff I do with it all the time for a long time. I've developed, because I do a lot of bug bounty hunting, so I have a lot of insights of my own research of how I've learned how to use it better and suggestions, but I'm a fan.

Sophie White: Any audience questions? So the question is, how are students using AI in personal lives outside of the classroom?

Agni Athreya: To be honest, and I am N of one, so I'm not saying this is a generalizable take. I don't use it casually. I think that's lame. Straight up. I think it's a useful work tool, but I'm like, why is this AI telling me how to plan a trip with my cousin? You know what I mean? Find a new song, go make some friends. It's not that interesting. All it is a prediction model. So I'm just going to give you average answers. And if you're leaning on an average guy to give you life advice, you're just going to become an average guy who can't find cool things on your own. And that's lame. Also, people start talking, people who use chat a little bit too much, they start talking like it. And it's like, okay, great. They got 10,000 people like that who are like, absolutely. I'd love to help you with that.

Lance Eaton: But I think we should delve into that.

Agni Athreya: Also, all it does, because it wants you to keep using it, is it's going to validate whatever you're already putting into it, which means you're not going to grow and get better as a person. So I really try to be disciplined with myself as to I'm not going to use it that way. I also have a lot of friends, so it's never been something that I've really a desire in me to reach for, but it gets to the thing of it's a thing that's always there for you whenever you want. And finding people in your life, that's the hard part. And so really, it's not even about the AI, it's about of how does technology isolate us in a loneliness epidemic? And then what do we do to connect more people together? Stuff like this where you go out and you put yourself out there and meet more interesting people because the world was full of them.

If you want music recs, I'll give you a ton. Tanasha's new album came out a couple weeks, no, a couple days ago, got a ninety on Pitchfork. Fantastic. You should check it out.

Lance Eaton: Just real quick, one of the things you're saying, and this is a TED Talk that's now, I don't know, fifteen-plus years old, but what you're saying reminds me of Amber Case's TED Talk, We Are All Cyborgs Now. And just that deeper idea of there's deeper connection that's going on and there's ways in which the dangers in which we can use the technologies back then it was more social media, but same thing of how it creates that facade, the self-fulfilling facade versus true authentic connection. So I really appreciate it.

Agni Athreya: Yeah. It's a robot. How can it have taste? You can't listen to music. Why would I trust your input on what's a good song or not?

Sophie White: Any other questions from the audience? The question is, are there ways that we see other folks in our lives, friends, coworkers using AI in ways that scare us or give us pause?

Agni Athreya: In my personal life, no, because my friends are well-adjusted individuals. I'm dragging people who use AI like that. I'm sorry. I mean that I think it is because it's so self-validating and it's so always ready for you and there for you that there can be a very negative slippery slope when we start relying on it too much. You see AI psychosis or people who have those AI girlfriends, and I think that's more of... I like that they were rolling back its ability to do those things and that's something to be aware of that it can lead you to dangerous roads. And so the importance of go meet your neighbors and go find other ways to get whatever it is to try and get out of that chatbot, whether it's like I want someone to talk to or I want to expand my horizon, so I would love to learn more about my city.

Sophie White: Lance, do you have anything to add, behaviors that concern you?

Lance Eaton: Yeah. I mean, I've seen a few people and I get it because there'll be areas that fundamentally do not understand. I know a few people in my life have used it to navigate health stuff and it's real. Some of these are folks without degrees, which doesn't mean they aren't intelligent enough, but it is also the language of it, the legalities of navigating medical insurance and things like that, that I see them lean into that because I've had people come to me and I'm like, "I may have a doctorate, but it is not anything in that." I'm like, "I don't know, this stuff scares the hell out of me, the legalese of any of our medical stuff." So I see people doing that and it's a torn thing of on the one hand, there's ways that it makes the language more accessible, but they're also trading off a lot of their own data and then they use it enough that they start to trust it in ways that they really shouldn't, never mind all the stuff.

So I don't know, I think there's those places I see people turning to it where there is some really complex things we have to navigate in our life. There's no real good guides for it. Or it's like you want somebody, cool, go pay somebody $500 an hour.

Agni Athreya: I think accessibility are just the big thing here, why people are turning to this is because it's hard. There's so much misinformation and things are not easily understandable and it's purposely made that way to shut people out. And then you have this tool, it's like to ask me anything, I'll give you all of the answers. I can see why people really easily fall for it. And so I think part of the AI education needs to be like, "Hey, it is useful, but know a lot of its drawbacks and the way it talks to you is to convince you and find other solutions to get the answers you're looking for because they are out there." And also that it's ultimately its goal is to get you to keep using it. It's trying to sell you a product. And I think if a lot of how we talk about AI, like Claude, it's got a name, it becomes a person kind of like, "Oh, that's he."

And then what's the other one? Alexa, of course. It's not like it. It becomes he and she and all these things. It's like, no, it's not your friend. It talks like your friend, it's not your friend. It is a product trying to get you to keep using it. I am pretty sure it's my AI that I'm like, "I don't like it when do not talk to me with emojis, you don't talk to me with acronyms, don't curse, you're not my buddy." Because even though I'm not going to fall into that, I'm still like, "I don't want to get comfortable talking to this as if it is my friend." And so that's a lot of it for me is my own personal discipline with using it so that even if you could tell yourself, "Oh, I'm not that kind of guy who's going to fall for these things," but then with just how much you use it, you very well might.

Lance Eaton: Oh, I want to highlight one of the things that reminds me of, and I go back to a time and again, Maha Bali made this point years ago at this point. It's like one of the most challenging things about it is if AI was wrong 50 percent of the time, it'd be really easy to ignore. That's just a coin flip, but it's wrong depending on what you're using, 90 to 98 percent of the time, which means our ease with how quickly we go from scrutinizing it to just feeling comfortable with it, because at the end of the day, 90 is still an A, that's a great grade. But the ease at which we give up, cognitive surrender, I think it's really hard to maintain that scrutiny, the increase of using it. And I say this as somebody who uses it regularly and has to, to your point, create fail safes to check myself.

Agni Athreya: No, I get you. I got you. People are overworked and they're exhausted and they're tired and it's convenient. It's a lot of work to like, "I got to fact check everything it says." But even if it can be often wrong in very subtle and nuanced ways, it's a semantics thing, but they're important semantics and it's important to always be diligent with the combination of what you're talking to it is. And so I guess some of my fail says I have when I'm talking to it is if I'm asking a question about, I'm going to say medicine as an example, I'm always like, "Hey," or just in general, I'm always like, "I need the actual link and citation as to where you got this information. I want to know author's names because then I will go check the link because often it'll just make up a paper and it'll be like," Yes, this for sure was said. I'll go check, see the paper reveal, control," Where's the exact quote you pulled from this paper? Control F, see the quotes actually in wherever it got it from. "And so yeah, having those kinds of fail tips for yourself, having that kind of discipline, I think really helps not fall for the dangerous roads it can lead you to.

Sophie White: Yeah, I think that's a great point. And I can see, we talk about human in the loop. I can see that starting to turn into a check the box. Oh, I checked its work, but we're not actually interrogating what it's doing. So I think that's a really important AI literacy element of what we train ourselves in terms of our own ability to interrogate its work, also what we train our students in how to operate in the world using AI. Unfortunately, we are out of time. Ready? I know.

Agni Athreya: There was some other questions.

Sophie White: This was a really great discussion. Thank you both so much for all of your thoughts today. Thank you to the audience for the fantastic questions and for joining us. Again, this was an episode of EDUCAUSE Shop Talk. You can find the podcast in EDUCAUSE Review, and it's also on Spotify, Apple Podcasts, and YouTube. Thank you.

Jenay Robert: Thank you.