Nearly four years after the release of ChatGPT, colleges are still struggling to contend with AI cheating. What have teaching experts learned? Will AI lead to an arms race of detectors and humanizers, or will it bring on bigger changes in teaching that might make college better? Jeff talks with four experts with different takes on the issue, including an official whose company sells an AI detector, and a professor who worries that such technology will only make things worse.
Will AI lead to an arms race of detectors and humanizers, or will it bring on bigger changes in teaching that might make college better?
Related links:
Signs of AI writing, on Wikipedia
“How An AI Detector Made Me Trust People Less,” by Marc Watkins.
Sponsor:
This episode is brought to you by Studiosity. Real learning happens through real thinking and real experiences. It's something that takes effort, not shortcuts.
For over 20 years, Studiosity has partnered with educators to protect the thinking, the essential work that the student must do for themselves, not detection, not an us versus them culture in teaching and learning. Just the real journey behind every degree that is worth holding. They're doing it for the institutions that trust their students, for the institutions that back their teachers to teach, not police, and for the institutions that stand behind their degrees. Keep the learning real with Studiosity.
Get the walkthrough at studiosity.com/learningcurve.
Jeff Young:
Metropolitan State University of Denver runs a workshop these days where they teach professors how to cheat on homework. Specifically, the session shows professors how to use new AI tools to cheat.
Jeff Loats:
The core idea is we want faculty to have a very student-like experience of what it means to work with these tools in a learner kind of position.
Jeff Young:
That's Jeff Loats. He's the director of the university's Center for Teaching, learning, and design to make sure that professors across the university get the training, even those who have not much interest in AI, they have been doing these sessions at departmental meetings that faculty attend anyway. The title of this workshop is meant to be provocative. It's called "Cheating Required.
Jeff Loats:
And we send the faculty ahead of time two quizzes in our learning management system. The first one we call Speed, and it's a 10-question multiple-choice quiz. These are questions taken from real courses at MSU Denver because the center I work with helps design a lot of courses, so we were able to get permission to use those. But it's from a hugely eclectic range of topics, and it's for both of these quizzes. We are specifically choosing questions that are not the expertise of whatever department we're visiting.
Jeff Young:
They're stumpers in this case for these professors.
Jeff Loats:
That's right. You wouldn't know, right? You don't know. You don't have this expertise. You're not in that class, and so we tell them to go try and answer these quiz questions using generative AI as quickly as you can. Right, fastest time wins.
Jeff Young:
On another exercise, the professors are given short answer questions, and they're taught some other tricks that students have been known to use.
Jeff Loats:
And so that's a place where we give the faculty a little prompt, maybe three sentences. It takes four lines or so, and on the website about ‘here's how to prompt the AI to give an answer that doesn't sound like AI to give a student-like response.’
So here's where we're trying to show faculty that it's really pretty easy to prompt AI to write in a way that doesn't necessarily immediately trigger people's detector … internal mechanisms for detecting it.
Jeff Young:
Yes, there are telltale signs of AI written work. Apparently, in fact, there is a Wikipedia page called "Signs of AI Writing" that lists some of them. But of course, students have figured out you can just ask the chatbot to rewrite its output to try to avoid these ticks. One takeaway from the workshop is that even if you think you can detect AI, you probably can't, especially if someone is trying to fool you.
Jeff Loats:
The analogy that I like to use is there's a you know on the on the list of logical fallacies that's out there. There's this one called the toupee fallacy, which is if you meet somebody that says I can always spot a toupee, they're always terrible. the The problem is, of course, you've only spotted the ones that were bad, and all the good toupees that you've seen in your life escaped your notice by definition. That's what made them a good toupee, right? So, can you spot poor use of AI by students? Absolutely. Can you spot all AI use by students? Not a chance. Like the literature is super clear on this. The actual scholarly work is just rock solid. We're not very good at detecting it when we when it's done with a little bit of care. If it's done sloppily, then sure.
Jeff Young:
This is the reality that educators are dealing with these days. Metaphorically, there are many more convincing toupees than ever. It has been almost four years since the release of ChatGPT, and this whole time, this has been the most pressing challenge for many teachers that I've heard from. That students have this easy and very hard to detect way to have a computer do their work for them. Of course, many students are very honest and don't use these tools this way. And as I learned on the episode where we went to university and talked to a bunch of students, students might use AI in one course that they don't really feel the professor is paying attention to them in. And be very careful not to use it in another course.
But I've been curious to know what is the latest advice on how to respond to this new reality. Basically, what have teaching experts learned from this new AI reality? Hello, and welcome to Learning Curve, where we look at how education is adapting to the rise of generative AI.
I'm Jeff Young, a longtime education journalist.
For this episode, I talked with four different experts with different takes. You'll hear from someone whose company sells an AI detector, and an expert who worries that such technologies might only make things worse? Back to those workshops by Jeff Loats in Denver. What does he hope that professors will do once they realize how easy it is to cheat with AI?
Jeff Loats:
We start the workshop with them by saying we're not going to leave you today with a bunch of really rock solid answers because I don't think they're out there right now. I think everybody is kind of fumbling in the dark about how are we going to preserve the important parts of higher education and grapple with the way that AI has laid bare some really bad problems with how we've always done assessment. Like it just made it has exacerbated those problems. So my take, if someone were to ask me that in the session in the room, I would say, well, first of all, what I think this means is you should not be relying on any sense of I will spot it if it is used. Like you should just let that go, right? Let that let yourself off the hook, because that takes a lot of energy and effort, and I think that that there's a way in which that kind of effort, even if it's something where you turn to a so-called AI detector, I think it kind of poisons the relationship that we're really hoping exists between students and professors, right? It turns it into sort of cops and robbers, and I and arms race, and who has the better prompt or the better tool?
Who has paid the AI companies for the right tool in this moment, right? Those kinds of things.
So we're really trying to steer faculty away from that and toward thinking carefully about how the curriculum needs to change. Where do we see these are features of the skills we want? Like we have learning objectives, right?
So there's learning objectives associated with every class on campus, and one of the things that we're pushing these departments to do is maybe just start with one course, pick a course, and as a department, go through the learning objectives for for this course, and think about is this something that we really need students to have at their own fingertips? If you want to be an expert that can be trusted to use AI well, you're going to need these skills yourself. So, like that's sort of one bucket, and if that's the case, then we intentionally build, you know, resources and learning activities and assessments that all rely on just the student doing it themselves, and we build in as much structure and as we can to make that the right approach.
Then there's maybe some. Right, the other end of the spectrum would be there are some things where you are not. You're just never not going to use AI to do this in your future role, right?
I'm working with faculty across all departments, but certainly there's some. Like I go talk to marketing. You're not going to be ignoring AI if you have a marketing job in five years. I don't know what it's going to look like, but it's going to look like something. And so we just can't pretend to graduate students with this expertise if we haven't exposed them and kind of forced them maybe to use some of these tools.
And then there's a bunch of things in the middle, right? Of is it AI assisted? Do you use AI to get things started, or do you use AI to wrap to polish things toward the end, and that's where, like, I think there's going to be tons of really interesting and difficult conversations in these departments about, ‘How do these skills and learning objectives land on that spectrum?’
Jeff Young:
So there are no easy answers, but Loats argues that AI is different from earlier technologies like the internet, as far as how tempting it is for students when they are under pressure to finish any given homework assignment.
Jeff Loats:
I mean, I always make an analogy of like generative AI, the free models that are out there right now, feel to me like if someone had come to my dorm room when I was 19 and installed a free candy vending machine outside the dorm room door, like it's right there. It's incredibly convenient. It's basically free.
Like my ability to to hold back from you know eating way too much candy would have been. I'm not sure that would be good for me now, right? As an as a full grown adult, so we're gonna have to really work to say like here's why I'm not why I don't want you to go to that vending machine or here's the structures I've put in place that mean you should be nice and full from this nutritious meal and that will hopefully help you be less tempted by this vending machine or whatever those structures look like we're gonna have to work really hard at it because man our right our humans evolved.
Your brain is trying to keep you from discomfort constantly. Like that's just one of its basic jobs: is keep you out of discomfort. But learning has some facets that are where discomfort is necessary, not in the like you know you got to work hard till you bleed kind of you know, but just that learning often requires confronting ideas that we thought that we hadn't thought about before, or we had thought about them and we were wrong, right?
I mean, a huge part of the physics research on good teaching is how do you help people get away from their misconceptions about physics concepts, and it's really hard. You can't just tell them the right answer and then sort of wipe your hands and walk away like, great, that's solved. You got to dig in there. They have to really confront an idea that doesn't work and see that it doesn't work, and then maybe they're ready to hear what does work, and that's just that's uncomfortable. So when discomfort is part of what we're after, and there's this button that can free them from discomfort, we're gonna we have a big challenge.
Jeff Young:
Just making me think of the freshman 15 and the dining hall with the ice cream unlimited. You know, it's sort of a little bit of experience of college, but now it's in the academic realm with the yeah.
Jeff Loats:
I mean, but at least that was like across campus, right? That wasn't like right here, right?
Jeff Young:
And only at set times. Yeah,
Jeff Loats:
Exactly.
Jeff Young:
For a national perspective, I checked in with Tricia Bertram Gallant, who is the director of academic integrity at the University of California at San Diego, and she's also a leading expert on this question. She's co-author of a new book, Academic Integrity in the Age of AI. I told her about that workshop in Denver, called “Cheating Required.”
Tricia Bertram Gallant:
Um, it's clever. Um, I would never run a workshop with that title. I'm not much of a, you know, that's uh, that's a what do you call that? Like hyperbolic or or you know to try and get people's attention, and I get that.
I think the gist of it is correct, right? Running, you know, using your new. I would. What I would do is I'd say to I'd pair faculty up. I say give each other one of your assignments and pretend you're a student and trying to do it. But I probably wouldn't call it cheating required.
Yeah, but I you know I think most faculty at this point have done it. that's the awareness is there. I would say for the majority of faculty, at least that I interact. Of course, I'm interacting with people who have brought me to talk, and so the people that show up are interested. You know, awareness is only the first thing. Desire is another thing.
Again, you know, knowledge of how to change it. So it's a baby step, but it's at least helping probably a lot of faculty.
Jeff Young:
What is Tricia Bertram Gallant's advice for professors?
Tricia Bertram Gallant:
From an individual faculty member's point of view, a key takeaway is at the very basic level. If you haven't yet run your assessments through the free version of ChatGPT, which is what majority of students use. Although I'm meeting with more that are using Claude now, then you should do it. You got to at least know, you know, what's happening with your assignments when the students put them in there.
So if you don't even have that awareness, how can you possibly identify the solution?
Jeff Young:
So how are professors responding to this new reality, and what are some bigger changes that colleges can make to respond? All that after the break.
Sponsor:
This episode was brought to you by Studiosity.
Real learning needs a struggle, but for educators, assessment security just means more detective work, and that's not what students came for, and it's not what the teachers came for. True learning and authorship isn't found in a probability score or a surveillance log. It's found in the journey.
Studiosity makes assessment security easy by shifting the burden away from every educator and into a proactive institution-wide infrastructure, from viva-style inputs to clear binary validity reporting, Studiosity is the multi-layered assessment security process, turning institutional policy into practice. Just a clear signal to the teacher, owned by the student, that I did this thinking.
De-risk your assessment flow and reduce workload. Give your teachers back the time to teach, not police, and give your students the agency to show their own growth. Keep the learning real with Studiosity. Get the walkthrough at studiosity.com/learningcurve.
That's studioocity.com/learningcurve.
This episode is also brought to you by the journal Social Research.
The new issue of the interdisciplinary journal Social Research explores how the way we write and the meaning of writing are changing rapidly. It asks how artificial intelligence transforms norms and expectations that, for centuries, associated writing with voice, originality, and agency. If writing is a technology that restructures thought, as philosopher Walter Ong put it, then which written forms of thought and reflection will survive AI, or will be enabled by it? What new literacies are required? How will new technologies rework the relationship between writing and the marketplace?
In nine groundbreaking articles, the summer 2026 issue of the journal tackles these and other questions around the title, "Writing in the 21st Century.”
Social research is the flagship journal of the New School for Social Research in New York City. You can find the journal on Project Muse, or at www.socres.org. That is www.socres.org. I hope you'll check it out.
Jeff Young:
One approach that many instructors have taken considering this new reality is to use AI detection tools. Lots of companies sell them these days, and they can be connected to course management systems so that professors can, at the click of a button, get a report on how likely it is that AI was involved in writing something, or what percent of the document seems AI written.
One big complaint about these detectors, though, is that they are not perfect, and since colleges and schools deal with large volumes of student papers and work, even a small percentage of error could add up and affect a lot of students.
And in some cases, disputes about these detectors are winding up in court.
Last year, a student at Adelphi University on Long Island sued the institution after a professor accused him of cheating when an AI detector flagged his paper as AI generated. The student says it was a false positive, and that he spent 15 to 20 hours on the paper. Though he did say he got help from an AI tutor. I was curious to hear from a company that offers an AI detector to get their perspective, so I connected with Annie Cecatelli, Turnitin's chief product officer, and I asked her about this issue of accuracy of AI detectors.
Annie Chechitelli:
We have a less than 1% false positive rate for a document, right? So meaning that we will flag less than 1% of documents of having more than 20% AI, you know, out of 100 like you know, and we fix that. Like that is fixed, meaning the rest, like the how much we catch is based on that, right? So if we wanted to change the 1% we're like, you know what? We'll do a 5% We'll catch a lot more, right? We'll see a lot more.
And so this is interesting because you hear it from both sides. There's, it’s a knob, right? The more sensitive it is, right? The more you'll catch, but the more false positives you can have, the less sensitive it is. Is you know we believe it's safer, but we're gonna miss stuff, right? And so we anticipate we probably miss meaning like miss meaning like we're gonna not call something AI that we think maybe we about 15 to 20% right, of the time.
But we also that changes, right? It changes with models, and we do updates, you know, to our models.
But as new different LLMs come out, we we have a system that monitors our performance against top LLMs. We also monitor what's happening in the world of contract cheating. What I mean by that is, I think many people don't realize the amount of startups and companies who are have come out to profit on students cheating has grown. I don't know if it's exponential, but it's pretty pretty linearly high, right? Because of AI, or just grown?
Because of AI, and so we track about 100 different companies that call themselves humanizers or bypassers, and you can see them online.
Jeff Young:
They advertise on YouTube directly to students, and they use your name of your company sometimes, right? To say like we can beat the the Turnitin AI detector, whatever they might even use your product names, you know.
Annie Chechitelli:
They absolutely do. But the one part I do like is that at least we require a premium. You have to buy the premium version. It says to get us Turnitin, but it is a lot. It's not cheap. I mean, they go they run between like six. And 30-$5 a month for a student to buy this, and you think about buy
Jeff Young:
The ‘defeat the detector?’
Annie Chechitelli:
Yeah, the ‘defeat the detector’ software.
So there's a whole new economy for trying to profit off student misconduct, which I do not like, and I do not like the people don't realize it. Like it is very, and you can still have somebody write your paper for you.
Like we forget about the old-fashioned ways of cheating, but those still exist. If you go to Google and you write, can I have somebody write my paper for me? Right, and that economy still exists in in essay mills. Now the interesting thing is those essay mills are using AI.
Jeff Young:
Some have argued that if these tools are going to have false positives, maybe they shouldn't be sold at all.
Annie Chechitelli: 20:44
The theoretical conversation of do we need to make sure students are doing their work? Like I think you do. I think there are plenty of examples of reasons of showing that it is important. I don't want a doctor or a nurse or I would say a lawyer or somebody who's in my best interest to not have completed and learned the things they were supposed to learn, and we have an obligation.
And sometimes we feel like, oh well, it's just humanities, or what? It is our obligation to make sure when we're conferring a degree to somebody that they have learned the things that they said that they were supposed to learn along the way, and is it perfect? It's absolutely not perfect. Is it different stakes for different disciplines? Absolutely, right. But I think it's essential and part of the underpinning of the value of a degree in this country.
Jeff Young:
The company stresses that its detector should not be used as the sole evidence of whether a paper is the student's work or not.
Annie Chechitelli:
It's part of one data point to something much larger, but the conversation is less about is AI present, but more of like how it was used and bringing that together. I think you know, and use the example of like there's any every test that people do or people take or smoke alarms, TSA, like medical tests, like that. That's the nature of tests, right?
And we've really tried to instill that any of our tools are for educators to have conversations with students, not to use it as a judge and a jury for telling a student that they're not that they've done something wrong. Right, have the conversation first and understand how it was used. And so we've since you know launched a lot of a lot more tools around clarity, so that teachers have more insight.
Jeff Young:
Marc Watkins is a close watcher of AI in education, and he's raised concerns about the misuse of AI detectors. He is assistant director of academic innovation at the University of Mississippi, where he directs an AI institute for teachers. He also co-wrote a new book, *The Norton Guide to AI-Aware Teaching.*
Marc Watkins:
AI detection is fascinating. It's not just using a classifier, aka old-fashioned AI detector too. There's now process tracking. There's watermarking. There's lots of different variables that are playing out in the field right now. The challenge is that if you accuse a student of using AI, they still have due process rights, and that is very time consuming.
That is also going to be something to going forward. That is putting a lot more labor on faculty members' parts as well.
Even the best AI detectors still have significantly high false positive and false negative rates, and there are lots of what they call adversarial prompting techniques to get past it. I think one of the better models on the market right now for AI detection is Pangram, and Pangram has some very high accuracy rates, but you can still get by it by using a humanizer. I think one of the most popular ones is called Walter Wrights instead of Walter White from the Breaking Bad series. That's what they call it. And then the other version of it too is that there are novel LLMs that come out every day that it's not been trained on.
One that came out recently was called Talky, which was based on pre-1930s writing from newspapers, and that is actually a way too that you can so-called wash text with a sort of smaller LLM to get past the AI detection.
So, if you use AI detector, it's very possible you're going to catch students that are using AI, but it's also very possible too. You're going to mislabel them, and you're probably only going to catch the students too that have the lowest levels of AI literacy or access to these other tools to get past it.
So it does become a little bit of a equitable situation too if you're going to be able to fairly get the students too that are cheating versus letting. Students too that know how to cheat better through the process.
Jeff Young:
But Tricia Bertram Gallant says that if used correctly, detectors can have a place.
Tricia Bertram Gallant:
The best way to figure out if a student wrote something or not is to have a conversation with them. Now that's not scalable if you have 900 students in a class, right? So there, it the problems are when we are. I find that we're using tech to solve other problems.
I think the real problem is 900 person classrooms. Right, that there's no way a professor can get to know their students. There's no way that a professor can really assess does this student have the knowledge and skills that I am trying to evaluate in a class that big, it's just not no way. I should not say no way. There is very challenging, right? Especially if it came to writing, a skill like writing, and it's not sufficient to say, well, instead of writing essays, you're going to write a blue book because that's totally different conditions than well, you know, what someone might write under.
So, no, it's not enough. Can it be one useful point of data? I think with some of the good programs, yes, but more for having a conversation with the student and for saying like their Google Doc version history. The conversation with them. Do they can they talk about the best thing is can they talk about what's in their paper? Can they talk about their process, and that all put together can be some information. But it should never be used as, ‘the detector says you used AI therefore zero on the assignment.’
Speaker 5
I think the other one that I'm seeing some faculty think about is to decide what level of a writing project, for example, needs to be assessed in a secure framework.
Jeff Young:
That is Marc Watkins again.
Marc Watkins:
Whether you're using a tool like a lockdown browser, or if you're having them come on site to actually do a test, if they're an online student versus you know student in class, they're making those judgments and they're making their calls. The reality is AI is so ubiquitous; it's very hard to keep out of a scaffolded process like writing.
So, if you can, for my my intent too is that if I teach a class that's writing heavy, which I did generally do every semester, try to get as much of the first draft in a place too that is secure for students too, where they can't have access to AI in some way, and then after that's done there too, there's going to have opportunities for them to look at different AI tools that it could actually help improve their writing to, or if it's used, you know, uncritically, could actually harm it. But we want to do this in a way that's open to them to have conversations about it.
Jeff Young:
What are a couple key takeaways? Because it is, it's a wicked problem. Like it's a very challenging scenario for a an educator right now,
Tricia Bertram Gallant:
Yeah, I think ultimately there are some things that need to happen at the institutional level because we keep focusing on what faculty need to do differently, right?
And again, faculty one, the majority of them were not trained how to teach when they did their PhDs, right? They were trained how to do research. They certainly weren't trained how to design valid assessments, as my colleague in Australia, Phil Dawson, says. Like, don't worry just about students undermining the validity of the assessment by cheating. Worry about the validity of the assessment to begin with, right?
And if the assessment's not a valid measure of knowledge, he's not saying it doesn't matter if students cheat, but he's saying that that's not what decreased the validity, and so we've got a situation where teaching is so important, right?
We all know we can all reflect on the teachers that really made a big difference in our lives, and they're mostly from the K through 12 area because they actually got trained on how to teach and how to assess learning, and so the institution of developing future professors needs to change, I think. And institutions, if they do say they value quality undergraduate education, then they should reward. They should, you know, like give faculty course release to redesign their course with the help of an instructional designer, just like we give them course release to do research or to write a book, but we don't give them course release.
We say you have to redesign this course and these assessments while you're while you're teaching it, right? So that's a fundamental structural thing that has to change, as well as the probably the tenure promotion reward system, especially at research universities that that tend to favor, you know, research over over teaching. Then faculty can have the space and the support they need because the majority I talk to want to change things. Right? They like being complacent. They just when are they supposed to do it? They need sleep. They need family time. They're human.
So I think that fundamentally needs to change at the institutional level, and then you know system wide. Obviously, there's some some big changes.
Marc Watkins:
It's also about the labor model of higher education.
Yeah, so many faculty right now that teach these labor intensive writing classes are in contingent positions. You know, I get yelled at too by K 12 teachers like you. We don't earn them like you college professors. Like, look, you don't quite understand what college is like in terms of the Stafford structural faculty. Adjunctification has been here now for at least a generation, if not longer, and there's no sign of this slowing down because of it.
And you're asking faculty that are generally speaking non-tenure track, in some cases teaching from one contract to the next to try to figure this out on their own, it is not going to work. You know, faculty will say thank you so much, I appreciate this, but I'm going to now go log off and go home and probably go to my second or third job.
So we have to be really sensitive to the fact that the labor model of education is not really matching that sort of intensification that this is causing us.
Jeff Young:
Tricia Gallant notes that even when faculty are doing the extra work to revamp assignments to make them more AI proof, there is a sense that something is lost.
Tricia Bertram Gallant:
You know, we just had gotten through the pandemic when this stuff all started hitting us, and truthfully, you know, a lot of the stuff that we've been doing is proven in decades of learning science research that it works or it used to work, right? And so it's hard to give up that idea.
I talked with one professor. This was early on in this era, and he said, you know, he said, "I've always done it this way. It's always worked. I got great writing from students, great insights from students, and now you're just telling me it's not going to work anymore. And that was grieving. And I really don't think we've given faculty the grace to grieve. What? How much has changed? And then I had another professor …
Jeff Young: 31:40
You mean since 2022 and the release of ChatGPT, essentially, yeah?
Tricia Bertram Gallant: 31:42
Yes, and I had another professor who had this great assignment. It was a a blog where they wrote about their leadership philosophy. It was in a leadership class, and it again, it had always worked. And she said, "I got to change it because people are using AI. So she tweaked it one way, and that didn't work. She tweaked it another way. She worked really hard on three different iterations of this assignment that had proven in the years prior that really inspired and motivated students to think and really reflect and and contemplate deeply, you know what kind of leaders they wanted to be, and she said after her third iteration, I think she just said, "I got to give it up. I'm just getting slop, and that again was a grieving process for her. But unlike the other professor, she worked. The other professor retired. He could have just yeah.
Jeff Young:
We're hearing that's one that's one thing. Right.
Tricia Bertram Gallant:
But this professor was not at the retirement age. She don't want to retire, and she tried really. She really cares about student learning, and she tried really hard, and just nothing worked. And so I think we really have to give faculty that grace. And so yes, there are still faculty out there who are at all different spectrums of this, and there are some who are just saying, "This is easy. I know what works in class exams. And then I'll say, "Well, what about the meta glasses? And what about you know all of this and they say oh I can spot them well maybe now because right now they look like sunglasses but Meta just came out with a whole bunch of new glasses that look more and more like you're wearing Jeff.
Jeff Young:
Like normal glasses yeah. my good old-fashioned prescription glasses that do not have AI in them yeah but you're referring to these new products that are hitting market fast with AI
Tricia Bertram Gallant:
yeah and they're working on ways for all sorts of privacy invading things, let alone cheating technology. So it's a tricky problem. Faculty aren't given the time, training, and support compensation to redesign their courses and assessments. So there, that's where we are.
Jeff Loats:
I think it's a lot of work that we're up against, right? One of the genuinely, I don't recall if this was my own idea, or I may have heard it or read it somewhere. But like one of the comparisons that I'm really valuing right now is asking someone on my campus or anywhere, what do you think is going to have the longer lasting impact on higher education, the COVID pandemic, or generative AI, and like to me, it's 100% generative AI. Like no question, hands down, that this is not even a contest. And then I say, okay, so think about the resources that your university brought to bear when the pandemic hit. Like at my institution, we center that I happen to run, we stopped everything we were doing, and we did nothing but help faculty do the emergency emergency transition to remote teaching in spring 2020 Right, so we we just brought and and we recruited and deputized people all over campus to help with that. Like, hey, do you know a little bit about technology? Great, come do this session for us, right? Because you're going to help these …
Jeff Young:
All hands on deck.
Jeff Loats:
Absolutely, we got money from the federal government, right?
We paid all of our faculty to do a summer training in the summer of 2020, so they'd be ready for not quite emergency, but still very quick transition to remote for that fall. So we think about those resources and how that. Shifted what we're doing and things like that, and then I try and point at the current moment and say, if you believe that AI is a bigger change to higher education, it's a slower change, and humans are terrible at slow disasters. Right?
Slow disasters are brutal for us. We're not really good at that. We can respond to an acute disaster super well, but we got to be thinking about like what are the resources we're going to bring to bear. We don't have to do it all in once. It doesn't have to happen in one single summer. But something at that same scale or bigger is is coming. We're going to have to do the work.
Jeff Young:
So is AI just causing an arms race of detectors and humanizers, or is it leading to bigger changes in teaching that could make education better? I asked Tricia Bertram Gallant whether she's optimistic that AI can lead to positive change.
Tricia Bertram Gallant:
I'm a hopeful pessimist.
I got that from the from a book, the name, but with the same name, which means that I am pessimistic because of the industry, the AI industry driving this conversation because of the motives and the pro the not yet you know the the profits associated with that the fact that these general purpose chatbots were not designed for teaching learning and assessment.
So I'm pessimistic from that side of things I'm hopeful because humans have agency and if we just embrace that agency, and we stop saying things like "the genies out of the bottle, "the train's off the tracks. We have no choice; it's inevitable. If we stop saying that and start realizing that as citizens, we have agency to lobby our governments to ask for regulations, for ask for guidelines and and boundaries.
We have agency as students to not use it when we're not supposed to, to make a different choice.
People talk a lot about students as, oh, well, they had no choice. Blah blah. Don't strip their moral agency away from them. Of course, they had a choice. The agency of instructors to make changes within their own class, but also to act as a body to advocate for teaching and learning and time and and pay for that.
So I'm hopeful in the sense that humans have always found a way to exercise their agency and solve problems, and so that's that's where my hope lies. And I'm I'm constantly it depends on the minute whether I'm more pessimistic or more hopeful. Depends on what news is coming at me.
Jeff Young:
For Jeff Loats of Metropolitan State University of Denver, the challenge for colleges comes down to one of leadership.
Jeff Loats:
I feel a little bit like each different constituent of the campus is waiting for the others to show how important they think this is. So I imagine a faculty member is thinking, you know, my chair hasn't really mentioned like how the big changes we have to make around AI, so I must be okay. Or the chair is thinking, you know, my dean hasn't really said anything about this. Or you know, the AVP for Faculty Affairs is thinking, well, chairs haven't really mentioned to me that they need something big around this.
And so it feels a little bit like we don't we we all need know we're all kind of antsy we all know something has to change we don't really know what it is which is makes it really difficult actually that's one of the hard parts compared to COVID like at some level basically every institution had some people on campus whose job was to know what does effective online teaching look like, right? So we have experience.
Jeff Young:
Yeah, yeah, they've been doing it for 10 years. We had experts, yeah.
Jeff Loats:
We had like research papers that were a decade old or two decades old about what makes a good online course, and we just don't have that for this, right? There is no decades old paper on how to handle generative AI in your online course, so we're all kind of trying to make our way.
So, for my take, the work for leaders is to show that they know how important this is and how big the change is that's going to be required.
Jeff Young:
This has been Learning Curve.
Every episode, we tackle big questions about AI and education. If you have feedback or want to share your big questions about AI, email me at jeff at learningcurve.fm.
And please make sure to follow Learning Curve wherever you listen, so you don't miss an episode.
We are gearing up to launch season two for the fall, so stay tuned for that.
This episode was put together by me, Jeff Young. The episode art was generated by Midjourney. Theme music is by Komiku, with other music by Blue Dot Sessions.
We'll be back in two weeks with more. Until then, thank you for listening.