Suspicion, Cheating and Bans: A.I. Hits America’s Schools
![Suspicion, Cheating and Bans: A.I. Hits America’s Schools Suspicion, Cheating and Bans: A.I. Hits America’s Schools](https://static01.nyt.com/images/2023/06/28/multimedia/28daily-ai-schools-image/28Daily-AIEducation-jtlk-facebookJumbo.jpg)
This transcript was created using speech recognition software. While it has been reviewed by human transcribers, it may contain errors. Please review the episode audio before quoting from this transcript and email transcripts@nytimes.com with any questions.
From “The New York Times,” I’m Michael Barbaro. This is “The Daily.”
[MUSIC PLAYING]
Since its introduction less than a year ago, ChatGPT, the artificial intelligence platform that can write essays, solve math problems, and write computer code has sparked an anguished debate inside the world of education. The question has become, is ChatGPT a useful research tool or an irresistible license to cheat? Today, my colleague Stella Tan speaks with teachers and students as they finished their first semester with ChatGPT about how they’re using it, how it’s changing the classroom, and the thicket of ethical questions that artificial intelligence raises at school.
[MUSIC PLAYING]
It’s Wednesday, June 28.
Let’s start by having you introduce yourself for me.
So I’m Alyana Nurani. I’m 19 years old, and I’m studying at UCLA.
My name’s Aiden. Right now, I’m in grade 12 in high school.
My name is Terry Dickerson. I go to Montclair State University. And I’m a junior.
I am 22, just graduated from college in Maryland.
And do you remember the first moment you heard about ChatGPT?
I actually first came across it through TikTok.
[MUSIC PLAYING]
I think it was TikTok or Instagram.
I started TikToks about it. I started seeing YouTube videos about it. And I was like, what is this thing?
And then I got back to school. And every single syllabus I have says on it, “if you use ChatGPT, you will immediately fail. And it’s like, whoa, what is this? Then I started exploring on it. And I was like, oh, I have to use it.
And I was like, asking it questions. Like, do you feel anything? Do you ever get lonely? [LAUGHS]
I was taking an Aristotle class. And I’m like so confused in it. And so I typed in “Aristotle metaphysics.” And then it told me all about Aristotle metaphysics. I was like, oh, my gosh.
I turned to my friend. And I was like, this is real, right? This isn’t a prank? It felt so otherworldly. Like, I was watching a movie that was based in the 3000s.
The way it types it out for you. And you can watch it think. It’s like mesmerizing.
It’s zeros and ones speaking back to me.
And I think I’ve used it for most major assignments since.
All right, my name is Andrew Reeves, and I’m 46. I’m a history professor at Middle Georgia State University. My specialization is medieval history. But being a small department in a smallish school, I also teach ancient history. I teach early modern history and world civilization survey courses.
And what made you want to become a professor?
Oh. So, one, is that I find this stuff fascinating. Two, because I find it so fascinating, I want to share it with other people. That is the biggie. And three, there is the sense that for a functioning liberal democratic republic, it does help to have the college-educated people have a basic knowledge of history.
So there’s the sense of duty about it, but then there’s also just the sheer joy in this. And because I like this stuff, I want other people to it. I want other people to see that this is just as fascinating as I find it.
[MUSIC PLAYING]
So I excitedly started a new semester, just hunkering down, ready to go. And I was sitting in my office. And first discussion board post graded for world history I, I read the post, and I looked at it. And I immediately thought, this does not read like a student.
[MUSIC PLAYING]
And then I remember thinking, I bet it’s ChatGPT. And it’s hard to put my finger on what it was exactly, but it’s a combination of polish together with impersonal, kind of like if Wikipedia had a voice. But I felt, OK, we’ve got ChatGPT detectors.
So I ran it through a ChatGPT detector. And bang, it said 99 percent chance of being fake. And so I just emailed and said, all right, here, you’ve been given a 0 because you had an AI write this.
So the first time I was not super flustered. I thought, OK, this is just another instance of dishonesty. We can deal with it. But what was more upsetting to me was in an upper level class. It was, I want to say about, three weeks later. And I was teaching Atlantic world.
what’s Atlantic world?
It covers the early modern Atlantic, about 1,400 to the mid 1700s. And I raised some questions about the Stuart Monarchy continuing the policies of Cromwell’s protectorate. And I read a post.
And it was a post that talked in too much detail about Stuart policies for what I had covered in the readings and lectures I assigned for that week. It was with a voice — that’s the thing. It was with the voice of someone who was fully familiar with Foreign Relations in 17th century England.
[LAUGHS]:
And it was that too-familiar voice. Because the thing about even a bright undergrad, eight and a half times out of 10, even when the material already, because they’ve been enthusiastic about this, their voice doesn’t read like distilled Wikipedia.
There’s still their own sort of personality. There’s still their own sort of idiosyncratic understanding. There’s still something they may have missed, or there’s something supremely insightful because they haven’t been swimming in the material the whole time, and they can come at it from an outside angle. But this post had none of that.
And I saw that. That was when I got this sinking feeling.
[MUSIC PLAYING]
What did it feel like?
It felt like all of my certainties about what teaching was going to be like were crumbling, because that was when I realized that, now, I can’t go to any student posts and know if I’m reading what a student wrote. Even in upper level classes, even among history majors, I’m going to be fighting this. And that just — it was an effing gut punch.
I was disappointed. I was sad. It kind of felt like a little bit of betrayal. But it was like, all at once, all at once this feeling of, what in the hell am I going to do now?
I don’t endorse it, but I understand why a student who is taking world history one in order to get their social science elective. I can understand why that student would consider discussion board posting to be a hoop to jump through. But if you are majoring in history, you should be interested enough in this that you can sit down and write 600 words about it over two weeks.
So every time now, I see a student post that looks like they know the material well, my first thought isn’t, aha, this is a good student who’s read the material. My first thought is, did a robot write this?
Every discussion board post that I see, it’s gone from a sense of delight and anticipation when I click to say, what has a student said? To a sense of suspicion and irritation that I am having to turing test each and every post. So there was this growing anger, unease, distress that, from here on out, there was going to be this pall of suspicion hanging over every student post.
I mean, did you ever doubt yourself? Did you doubt your suspicions and also the GPT detectors, which have been shown not to be entirely accurate?
Ah-ha. Yes. There is a vast gray zone of areas where I don’t know. Is this a student who knows the material well and is grammatically capable? Is this a robot?
And I am near certain that robot-generated posts have gotten past me because I don’t normally go in guns blazing in terms of disciplinary action. And I would rather there be false negatives than false positives. And I don’t like going into course discussions with an adversarial mindset. I want it to be where people learn things.
Yeah. I’d love to talk more about that. How does that affect your approach to teaching and your relationship to your students?
So, clicking open the discussion board and realizing that I might be reading text-generated by a machine that has no inner life, I feel like — do you remember the psychology experiment when they got the little baby monkey, and it was hugging the model of a monkey because it needed something to have contact with, but it was hugging this piece of machinery?
Am I’m being overly dramatic? I might be being overly dramatic. But I feel like. It’s like, am I a duck swimming up to a decoy excitedly trying to mate with a piece of floating wood?
That’s what I feel like. I mean, the joy of talking to a student is what? It’s that there’s this other human out there that I’m connecting with. And when they’re talking to each other on the discussion boards, that they are connecting to other humans.
And this idea of robots and eventually maybe robots talking to robots and no humans there. Yeah. I don’t know. I sort of had this terrible feeling that the future is this road going into darkness, and I have no idea what’s at the end of it.
[MUSIC PLAYING]
After the break, stories from students.
[MUSIC PLAYING]
So tell me a little bit about how you’ve used ChatGPT for school.
When I had first heard it a while ago, I was like, I’m not going to touch it. I really pride myself who I am as a student, and I’m very dedicated to my academics. I was like, I’m not even going to go near this thing. I’m just going to be authentic in the work that I submit.
See, I don’t use it to analyze things or write my answers for me because I’m paying so much for this education. It kind of feels like it would just undercut all the value of it to cheat like that because I’m learning how to critical thing. So why would I want AI to do that for me?
But after using it with my friend, I was like, OK, there is a way for this resource to be used, aside from how students have been using it for plagiarism.
I did use it for an English assignment where I was analyzing Beowulf. And I had no idea what was going on in the story because it was impossible to understand. So I typed in Beowulf chapter 40, whatever, into ChatGPT, and it pumped out just a quick summary.
I had this exam to study for. I’m not very good at math. And I was like, maybe this thing can explain to me how this math is working right now.
So I gave it all of the stats. And immediately, it populated a step-by-step process of how to do the division, what everything meant. And I was like, wow, this was so so, so helpful.
How would you have handled that situation before ChatGPT?
To be so honest, I probably would have just had to muster up the courage to just ask a classmate, but it kind of eliminated that awkward step.
I think I would have just had to ask a friend or try harder, which, now that I’m thinking about it, I probably should have just done that.
Do you feel like you lost out on anything by using ChatGPT in that way?
I think there is something missed when you don’t just go up to your classmates and ask for help. In the moment, I just really needed the answer.
But I think the downsides of constantly going to ChatGPT would be that you miss that one-to-one connection that you could be having with a classmate, that you could be having with a professor, and really connecting and bonding over the material that you’re discussing.
Sometimes I worry that I’m not a good critical thinker, and that I’m only getting the shallow level of things I read. And the good thing about ChatGPT is it kind of like tells you what you’re missing, which I guess is making me a worse reader because I’m to do it on my own.
But it didn’t give me an analysis. So as long as it’s not giving me ideas, see, it’s hard to know what it’s going to crank out. So it’s hard to know whether it’s going to give you ideas. But I just — yeah, the line is getting a little blurry now.
Yeah, I think as I’m giving you examples, I’m realizing, oh, my gosh, it really is just for critical thinking, and that’s what I’m using it for. And I have this internal law that I will not find shortcuts in analysis, but I totally am doing that. And I don’t think there’s a way to use it and not do that because it will think for you. If you give it a question, it’ll just think for you.
[MUSIC PLAYING]
I’m 22 years old, and I went to a university in California for computer science.
I am a recent grad, and I studied finance.
I am 24 years old. I just completed my masters in health sciences.
And you want to be anonymous?
Yes.
Yes.
Yes, please.
Have you ever directly copied and pasted from ChatGPT and turned it in for an assignment?
Yeah.
It was towards the end of the semester. I had a good amount of papers and final projects coming up for a multitude of classes. I was feeling very stressed about it. And for me, especially, I despise writing papers. Like, I’ll do math problems and econ problems all day. But researching and writing papers feels like the bane of my academic existence.
I started using it to sort of create the structure of my papers. I would just do like stream of consciousness on a dock, and then it would package everything that I said in it. And then I would be able to tweak it.
The first time I used it, I was writing a discussion post or whatever. So I just copy/paste the prompt. And then it gave me a response. Oh, my god. It was so good. It was so good. It was amazing.
It just wrote everything out clearly. And the sentence structure was so good, too. It wasn’t the most detailed. But I was like, hey, I can work with this. I can rearrange sentences, adds some little more specific details.
Out of paranoia, I used a lot of plagiarism platforms to see if it even detected AI. And then my use of it became more and more complex as I began to trust it.
So I was going to write about a criminal justice system. And as the deadline encroached on me, I was like, yeah, I’m not really finishing this. So I started having a conversation with ChatGPT where I was like, OK, I want you to write 500 words talking about this topic.
And it did it. I read through it. And I was like, OK, this is good work, but it doesn’t really sound like me. So then I gave ChatGPT — I was like, I’m going to give you something that I’ve already written. And I want you to rewrite this paragraph to make it sound like me.
Wow.
So I copy/pasted the page of what I’d already written. And then it rewrote that paragraph. And I was like, OK, this works. And then I pasted those sections into my final paper.
And I reworded a little bit to make it sound like less robotic. And then I’m like, beep, beep, boop, boop, turn it in. Submit. 10 out off 10. Good job. Nice work. Well, thank you.
It was 10 out of 10?
Yeah.
I got an A. And my professor actually liked the ChatGPT part of the paper more than mine.
Huh.
She gave us back our edits. And the first half of the paper that I’ve written is full of correct this here, correct this there. And then literally, around the part where I started using ChatGPT, the corrections just reduced drastically to a sentence per page or something for the next four pages. And I was like, so ChatGPT literally made better quality work than I did.
Every opportunity I had from the, I would use ChatGPT, you know what I mean? Like, first case, hey, what’s up, ChatGPT? How do I do this?
[MUSIC PLAYING]
Did you have any feelings about using ChatGPT to do your work in this way?
Not really. Do I feel guilty about using ChatGPT?
Mm-hmm.
No. In the moment, I was like, even more excited. I was like, dude, it’s too easy. I can go outside. I can touch grass or something.
I felt a little like, damn, is this what I came to school for to not do the work? I had a good excuse. I felt like this class was not important to me. It was a requirement, but it was an antiquated requirement.
It was a Brazilian studies class. Like, I’ve got to take a history class to fulfill a requirement. It’s not something I’m really interested in. It’s cool to learn about Brazil. But at the end of the day, I’m just trying to pass the course.
You don’t feel like you kind of cheated yourself out of learning by using ChatGPT in these ways?
I didn’t really feel that bad about it, to be honest. If GPA is everything, then students are only going to care about grades. And GPA is pretty much everything.
Your internship applications, your job applications, they literally have all these GPA requirements. So I feel like that emphasis on grading and GPA, that pushes students to use whatever means that they can.
College, we pay hella money, right? We’re just trying to get a degree, that’s it. Credentials. Just like, I’m stressed I really need to get this degree so I can get a job so I can survive. Oh, man, oh! So when that’s your mindset, then it’s really hard to be like, oh, no, let’s learn everybody. Yeah!
Well what would you say to students who might say it’s not fair that you use ChatGPT like this?
I mean, I can understand feeling pissed off when you did all that work, and you found out that other people got there easier. And it just makes you feel like all your work was invalidated. I understand their feelings of unfairness because I also used to feel that way, too. But it’s like, the whole world is unfair already. People are going to get jobs and get picked for positions just off of knowing someone or some weird connection or whatever. The system is already unfair, so it’s up to you to try and get it how you get it.
I would say it’s going to be the same case for people who didn’t learn how to type when computers came out. And I feel like over time, it’s going to be used in a variety of different industries. So it’s not cheating. It’s just like learning how to use a new tool.
[MUSIC PLAYING]
My viewpoint on education has changed quite a bit. I’ve been in school for 13 plus years. And I realize that the workforce that we’re entering after being students, it’s fast-paced, and it’s who can get the information there the fastest, the quickest, and in the most concise manner. And if you could identify tools that could help you do that while still having somewhat of a foundational understanding of what it’s doing, I see it as more of an extension of our abilities as opposed to cheating.
We live in an age now where, literally, any piece of information I want, I could have it immediately. So it’s just like what is the incentive to really, really learn, to memorize, if you have this external brain that could do it for you?
So Andrew, earlier, you mentioned a feeling of betrayal when you found out that your students were using ChatGPT.
Yes.
And I’m wondering what the betrayal is of. Is it of you, of the material, of the ideal of learning in a college setting? I mean, who or what is being betrayed here?
I think it’s a betrayal of the purpose of a university class. We’re on this journey together, is my feeling about a class.
Mm-hmm.
And so I suspect one reason it hit so hard for me is that a great many students never saw themselves on a voyage of discovery along with me. They saw themselves en route to a credential. And to some extent, the upset is my realizing that not everyone is going to see this as a magnificent voyage of discovery. [LAUGHS]
Well, have you talked to your students about ChatGPT?
I need to do that at some point. I think that ChatGPT opens up a teachable moment for getting students to think about, why do you learn to do stuff that a machine can do? Why do you learn to do the math when I can realistically just, you know, pull out my phone and get my phone to do it? Why do math students still learn 1 plus 1 is 2, for example? Why do we still learn these things?
And I think — I hope this is not me just telling myself I am still relevant — but I think that the writing process, especially, or the thinking process, what’s important for a student is they’re learning how to organize their own thoughts. They’re learning how to really express themselves and to think clearly. And that’s the value of having people write, and it’s the value of having people think about stuff they read.
It’s not just sucking up the facts and spitting them back out. It’s taking in — you’ve got to have the facts in your head to chew on, to think about, to train yourself as a thinker. And I think that, as I’ve had this conversation, it’s helped me to think about how this could be a very teachable moment, especially at the outset of the semester, in a way to say, here’s why we’re doing this, rather than, I will come down on you with the wrath of God if you decide to cheat with robots.
Andrew, thank you so much for this conversation.
Well, it was a privilege getting to think out loud about all this stuff. Thank you very much for talking to me about this.
Take care.
You, too. Bye.
We’ll be right back.
Here’s what else you need to know today. On Tuesday, the Supreme Court rejected the legal theory that state legislatures have almost unchecked power to create the rules for federal elections and draw partisan election maps without oversight from state-level courts.
The case stemmed from a voting map drawn by North Carolina’s Republican-controlled legislature that was later thrown out by the state’s Supreme Court as overly partisan. North Carolina Republicans seeking to restore that map asked the US Supreme Court to intervene, saying the state court had no say over the matter. By a vote of 6 to 3, the Supreme Court’s justices disagreed, saying that state courts do have a role to play in such disputes. Many legal experts celebrated the ruling, saying it would avoid introducing even more partisanship into federal elections.
Today’s episode was reported and produced by Stella Tan and Diana Nguyen with help from Michael Simon Johnson. It was edited by Lisa Chow with help from Ben Calhoun and Paige Cowett. It was fact-checked by Susan Lee, contains original music by Elisheba Ittoop, Marion Lozano, Dan Powell, and Rowan Niemisto. And was engineered by Chris Wood. Our theme music is by Jim Brunberg and Ben Landsverk of Wunderlich.
That’s it for “The Daily.” I’m Michael Barbaro. See you tomorrow.