# AI cheating runs wild on campus | Front Burner

https://www.youtube.com/watch?v=RM3X4liaE5I

[00:02] Hi, I'm Elaine Chao filling in for Jamie Poissant.
[00:11] We've had a lot of conversations on the show about the growth of the AI industry in the last few years.
[00:17] From whether it's a bubble that's about to burst to how hard it is to govern.
[00:21] The technology is evolving so quickly and unpredictably that it's upended a wide array of sectors from entertainment to manufacturing and to higher education.
[00:32] Today I'm talking to James Walsh about that.
[00:35] He wrote a piece in New York magazine this month called everyone is cheating their way through college.
[00:41] The use of generative AI has become rampant on college and university campuses across North America.
[00:49] According to KPMG who surveyed over 400 Canadian students about this in the fall, around 60% use AI chatbots like chat GPT for their school work.
[01:00] A recent poll from study.com in the US found that
[01:03] 90% of a thousand students who were surveyed used it on the majority of their assignments.
[01:11] Today, how students are using AI to cheat, why professors and institutions are struggling to keep it in check, and how this is ratcheting up a debate about the value and future of higher education in North America.
[01:31] Hi James.
[01:32] Hi Elaine.
[01:33] You spoke to dozens of students while reporting on the story, but there was really one in particular that really stood out to me.
[01:40] His name is Roy Lee and he goes to Columbia and he told you that AI wrote about 80% of every essay he turned in, which to me was pretty astounding.
[01:52] And tell me a little bit more about Roy and how he was using AI in his school work.
[02:02] Right, Roy was fascinating to me.
[02:03] He
[02:06] Worked really hard to get to an Ivy League school.
[02:11] He had first gotten into Harvard in college and had that admission rescinded because of a disciplinary issue when he was in high school.
[02:20] And then he spent a year, sort of gap year, and then a year at community college and eventually found his way to Columbia where he proceeded to use AI to kind of cheat his way through every course.
[02:33] And it wasn't because he, you know, had any sort of academic deficiencies or anything.
[02:40] It was simply because he saw homework assignments as hackable and therefore he felt it was like permission to use AI because, you know, he he believed, "Hey, I'm going to have this tool in my life forever.
[02:55] I might as well use it."
[02:58] Right, he even created software that would allow other students to cheat on their coding interviews.
[03:03] Was that right?
[03:05] Right, so he has this ambition to to be
[03:09] A kind of a classic startup Silicon Valley founder.
[03:11] So he went to school with this mission of of two two-pronged mission.
[03:18] He said, "School is good for finding a co-founder and and finding a wife."
[03:24] Um, sure.
[03:25] Uh, eventually he he did find a co-founder and they landed on this idea of using AI to build this platform which allowed coders to cheat their ways through way through remote job interviews.
[03:39] Coders applying to jobs at really big tech companies often have to do these kind of riddles, these puzzles that Roy kind of saw the same way he saw, you know, homework assignments that that they were stupid and hackable and, you know, not really a reflection of the work a coder would be doing in the workplace.
[04:00] And so, he built this platform um and advertised it by hacking his way through an Amazon job interview.
[04:08] Hey guys, I'm about to join my interview.
[04:10] For my final round Amazon away at Amazon internship, and I'm using Interview Coder.
[04:15] So, here's how you can do it.
[04:18] Just ended the meeting.
[04:19] Uh I don't know if you could tell, but I was pretty much copy-pasting exactly what Interview Coder gave me, but um down to the comments, I was commenting, I was like writing down exactly what Interview Coder gave me, so uh that feels like an offer, so I'll be back if I get an offer.
[04:33] That's how you can use Interview Coder to get a real Amazon internship.
[04:38] He got the job, which he turned down and posted a video online to promote this platform.
[04:44] He's clearly enterprising and certainly.
[04:47] And that was a very enterprising use of AI.
[04:49] But uh for many of the students that you spoke to, AI was being used also in like some very mundane ways.
[04:54] Like I want ChatGPT to help me write an essay for my first-year English class.
[05:01] Like how would that process work?
[05:05] Sure.
[05:05] So, essay writing was a was a big part of this.
[05:07] You know, of course, students are using it for their coding assignments or they're using it for math.
[05:12] Or data analysis and sciences, but essay writing was really a big part of my reporting.
[05:16] And for example, you know, I spoke to a student at a top university in New York City, and she really walked me through her process for writing essays.
[05:27] This is somebody who told me she enjoys writing.
[05:29] She enjoyed writing essays on her own kind of cold turkey when she was in high school.
[05:35] But, you know, then she got to college and sort of took the same view as Roy's that essays were now hackable because of AI.
[05:45] So, she would sit down with the assignment and she would prompt Chat GPT.
[05:49] She would first give it some information about herself.
[05:51] You know, I'm a first-year student.
[05:56] This is the class I'm taking and we've been reading, you know, these different authors and then here is the essay prompt that my professor has given me.
[06:07] My conversation with her was really insightful because she started off our conversation saying, "I am totally"
[06:13] Against cheating.
[06:15] She knew that plagiarism was against the student handbook and so I'm against, quote, copy and pasting from Chat GPT.
[06:25] But, what she did instead of copy and pasting, you know, an essay from Chat GPT into a document was she asked Chat GPT to generate an outline and topic sentences for each paragraph of an essay.
[06:36] So, essentially this chatbot was coming up with the idea, the topic sentences and she kind of had to do a kind of paint-by-numbers for her essay.
[06:46] And so, she saw that as the work.
[06:50] Right.
[06:50] I Yeah, you know, she wrote her essay in the sense that she literally wrote it.
[06:54] She was very proud of how quickly she can do this.
[06:57] You know, she said she started the assignment the morning it was due, woke up early and spent two hours on it and filed the assignment when normally it would have taken her, you know, four or five days of writing, she could just kind of bang it out.
[07:22] I feel like a through-line in a bunch of your interviews was kind of this as a time-saver, right?
[07:26] Like you spoke to a student, you call her Sarah in the piece and from Wilfrid Laurier University in Ontario and I think she said something to the effect of I have to spend a lot of time on Tik Tok so I don't have much time to organize my school work.
[07:43] Sure.
[07:44] Sarah yeah, Sarah was a first-year student at Wilfrid Laurier, and she told me that she was fully addicted to using ChatGPT.
[07:53] In fact, she had taken to Reddit to confess her addiction and earnestly ask for help because she recognized herself coming to depend on AI in a way that kind of freaked her out.
[08:07] But you're right, you know, she told me that addiction it was sort of a cyclical thing where she would spend hours and hours on TikTok until her eyes hurt, she said.
[08:19] And that meant that using ChatGPT to write an essay just kind of saved her.
[08:23] Time.
[08:23] I think we've all had the experience of being on social media and time flies by and you think, what did I accomplish?
[08:31] But that's probably increasingly a challenge for college students who, you know, will get lost in social media and then suddenly look up and think, oh, I have a deadline.
[08:39] And so it's easier for me, the time-saver, to write an essay using ChatGPT and write an essay that, you know, it may not be a perfect essay, but it's safe enough that, you know, I'll get a good enough grade.
[08:53] Um, like I've heard some students even describe it as part of their learning style.
[08:57] Well, certainly be.
[08:57] I think that probably comes from all of the positive or creative use cases, you know, there are plenty of positive use cases for AI in education.
[09:11] And we are delighted to finally announce the grand opening of the Carnegie Learning Office.
[09:15] The main service that's being provided in this office is the development of our K-6 math software product called Mathia Adventure.
[09:21] It's
[09:23] Actually producing greater hints and suggestions for students to improve their mathematical abilities, but it's not giving them the answer in any way, shape, or form.
[09:30] And so we're using it in a very subtle and intentional way to help students improve their math outcomes.
[09:38] AI can be a good study buddy.
[09:40] It can help students understand concepts that maybe they didn't really have a good grasp on during class or they can produce study guides that are really useful.
[09:50] So, I do think there is this kind of sense among a lot of college students that it's like this, you know, Socratic method of learning and they're having a conversation with a chatbot.
[10:03] And so, I think a lot of college students sort of see that as their study method.
[10:08] That's how they learn now.
[10:12] So, we've talked a bit about the student side of things.
[10:13] I want to ask you about how teachers are dealing with all of this.
[10:18] And from the folks that you spoke to, you know, what have they been noticing in assignments since AI became?
[10:27] So prevalent on campus?
[10:29] Yeah, and I think right away, certainly professors who taught writing classes began to notice students handing papers that sounded robotic or suddenly had perfect grammar.
[10:44] Or one professor pointed out it was strange how counterpoints and counterarguments were suddenly presented just as rigorously as central argument.
[10:55] So, these papers didn't really sound like college freshmen or sophomores if they even sounded human at all.
[11:04] That started very soon after ChatGPT launched and really has only gotten worse.
[11:11] And I spoke to quite a number of professors who kind of felt like they were in this state of despair because for a number of reasons.
[11:20] Number one, there's really no good way to catch and prove somebody's using AI.
[11:29] There is software that these AI detectors and their actual effectiveness varies widely between platforms.
[11:39] And so, the professors I spoke to just did not feel comfortable relying on those detection platforms to sort of bring a case against students.
[11:48] It's kind of turned, it's made the student-teacher dynamic.
[11:51] It's made professors kind of have to build this Perry Mason case against students.
[11:55] And it also just makes this sort of cat and mouse student-teacher dynamic and makes professors kind of have to build this Perry Mason case against students.
[12:04] And this is something that is so prevalent that, you know, it would require quite a bit of a professor's time to kind of catch every single AI plagiarist.
[12:12] And students also know this.
[12:15] You know, students have said, "If you're caught, just deny, deny, deny because they can't actually prove that you're using AI."
[12:22] Um so, it really has been extremely frustrating for professors.
[12:29] Secondly, a student can write their own.
[12:31] Paper and it cannot be detected as AI generated.
[12:34] But, if you really dig into it, had that student um come up with their own topic for the paper?
[12:41] Have they rewritten AI?
[12:44] You know, it's quite easy to kind of game the system.
[12:47] Uh other students told me about um laundering AI essays through other AIs so that detection software won't pick up on it.
[12:55] You can insert typos.
[12:57] You can have AI insert typos.
[13:00] There are AI platforms that now, you know, advertise themselves as having kind of greater authenticity, authentic language.
[13:06] You can upload past essays so that it will write in your own voice, assuming that, you know, at this point you have past essays that are your own voice.
[13:14] Uh so, so there are any number of ways to kind of fool professors and it's just led to this kind of intractable situation where where professors don't really know what to do.
[13:33] I found the details of this uh one study about AI detection software so stark that researchers use fake student profiles to hand in work that was 100% AI generated.
[13:43] And after professors graded this work, they only caught about 3% of the AI generated work.
[13:51] Imagine that this has made it all quite challenging to come up with policy around dealing with this.
[13:56] You mentioned a bit some instructors would put in Trojan horse white text into their assignments and there seems to be a just a wide array of policies being implemented uh to to deal with this.
[14:13] Yeah, the Trojan horse to me was a really good example of of kind of the the desperation.
[14:17] Uh this is these are professors who and a number of professors do this who will insert, you know, between paragraphs of their prompt.
[14:25] Let's say they assign an essay and the assignment has two paragraphs and in between those paragraphs they'll insert in little white text something.
[14:35] That is a total non sequitur.
[14:37] It says, "Mention Finland. Mention Dua Lipa."
[14:40] You know, um "Mention broccoli."
[14:42] And the idea is that if a student were to blindly just copy and paste that prompt, throw it into Chat GPT, Chat GPT or whatever platform produces an essay and then they copy and paste that essay and put it back into a document and hand it in, you know, the essay will have some strange off-ramp about broccoli or Finland or Dua Lipa.
[15:05] And so it's just kind of fascinating because these are students who not only didn't write their own essay, but they also didn't read their own essay before handing it in.
[15:13] And that's sort of what it's come to that professors are resorting to these Trojan horses.
[15:18] Um and so it just one of the kind of the more depressing anecdotes that I learned about while writing this report and the story.
[15:27] As much as professors are frustrated with students using AI, um at the same time some are using AI in their own work.
[15:34] Like US survey of 1,800.
[15:37] Higher education instructors this year showed that about 40% use AI for their courses.
[15:42] There was also a recent piece in the New York Times about how some teachers were using chat GPT to create lecture notes, organize presentations, even provide feedback on students' essays.
[15:56] And how do you think that kind of complicates the debate around all of this and how to deal with AI in higher education?
[16:04] You know, I don't know if it complicates it.
[16:06] I think it's a reflection.
[16:09] This story really, this is about the early adopters of AI.
[16:12] You know, today's college students are tomorrow's, you know, entry-level people in the workplace.
[16:18] And so the idea that professors are using it, um, just as other people in other jobs are using it is not all that surprising.
[16:25] Of course, to the extent that it's violating kind of the student-professor compact, that's something that of course we have to be concerned about.
[16:33] You know, if a professor or teacher is using AI to...
[16:38] Leave comments and feedback on a paper, that would certainly feel to me like I was being short-changed if I'm paying tens of thousands of dollars for my education.
[16:48] And it also opens the possibility, um which a number of teachers and professors pointed out to me, that if an essay is AI-generated and um a professor is leaving...
[17:00] Right.
[17:00] You have two robots grading each other.
[17:04] Yeah.
[17:04] Or or maybe just even one robot.
[17:07] Of course, it raises all kinds of concerns.
[17:09] I I also spoke with professors who were using AI in ways they really felt was beneficial to the teaching process.
[17:17] I spoke with a professor who would used AI to generate an entire textbook for a comparative literature class and she told me that the textbook had freed up time with her TAs um and this was the best class she taught in her career because she felt like it was such a good textbook and and the students were so engaged.
[17:37] So, there.
[17:39] Are these positive use cases?
[17:41] Right.
[17:41] And I I also read that there are high schools in Miami that have introduced AI chatbots to allow students to almost speak to historical figures.
[17:51] You could talk to Abraham Lincoln and to learn about his life and so forth.
[17:54] So, there are There's that side of things as well.
[17:58] What Yeah, like what strikes me is that there is Like even in talking about the use of generative AI in education, there there is such a spectrum of its use, right?
[18:07] Um which complicates, you know, when when we're we're talking about the ethics of something that, you know, using it to generate like ideas for curriculum is different than using it to grade something, right?
[18:18] Like so, there is such It just strikes me that there is such nuance in looking at all of this.
[18:25] I Yeah, I also just don't think it's apples-to-apples in terms of overworked professors relying on it for help versus students who, you know, do not have fully formed brains and have never known a college without AI and the sort of downstream.
[18:40] Consequences of that might be a little bit different.
[18:53] Another layer to all of this is that many of these universities where, you know, students are using ChatGPT on their assignments, cheating with ChatGPT, many of these universities actually have partnerships, financial partnerships with ChatGPT's parent company OpenAI, right?
[19:10] Which would I would imagine make banning it as a way of like dealing with the situation quite unlikely, right?
[19:18] There is like money and profits at stake here as well.
[19:22] Certainly.
[19:22] I mean, AI platforms are generated by people and you know, your nonprofits or companies or companies within nonprofits in the case of Open AI.
[19:32] A lot of these these schools have partnerships with Open AI.
[19:35] Whether or not that plays into ban that you know, the question about banning, I
[19:41] I don't know because even if they were to to ban it on campus, I don't know how that would even be enforceable.
[19:48] So, I think that would be very very difficult.
[19:50] But Open AI all the platforms that offer products like Chat GPT EDU, which provides some guardrails and is tuned sort of more to the learning experience, they say.
[20:05] But they also provide Chat GPT Plus, a platform that costs $20 a month for free during finals.
[20:11] Is like target marketing to students, right?
[20:14] Certainly.
[20:14] Certainly.
[20:15] You know, the these these platforms are in a race to capture users.
[20:20] And younger users, teens and and college students are really, you know, an important demographic for them.
[20:28] I had kind of the surreal experience as we were finishing this story, I got a push notification that Google had just sent out an email to parents saying that they were about to offer a chatbot to I believe children.
[20:43] 13 and and and under. So,
[20:46] it's certainly
[20:48] like any any company that they are
[20:49] trying to capture these these young
[20:52] users.
[20:53] James, what do you think are the wider
[20:55] implications of all of this on the
[20:59] higher education system more broadly
[21:02] here?
[21:03] I think we're a long past the time when
[21:05] anybody kind of thought about college in
[21:08] this this sort of ideal way as a place
[21:11] where, you know, students go off to
[21:13] expand their minds for a few years, you
[21:16] know, I think Roy, that Columbia
[21:18] student, was really um instructive not
[21:21] just because of his views on AI and
[21:23] cheating,
[21:24] but also because of his understanding of
[21:27] college as this transactional place, a
[21:29] place where he would go to, you know,
[21:31] simply to meet a co-founder, not because
[21:34] he wanted to engage with the core
[21:37] curriculum, which was supposed to be at
[21:39] Columbia, you know, intellectually
[21:40] expansive and personally transformative
[21:42] as the as the school advertises it or
[21:45] describes it. And and he was just there
[21:48] kind of as a stepping stone to to
[21:51] Silicon Valley.
[21:52] I do think that there's going to be a
[21:53] very fundamental reframing in how we do
[21:56] almost every bit of knowledge work in
[21:59] the future. Essays, writing is not going
[22:01] to be in the same, tests are not going
[22:02] to be connected the same, memorization
[22:04] will not not not need to be happening.
[22:06] So, we're headed towards a future where
[22:07] almost all of our cognitive load is
[22:08] offshore to to LLMs. And uh I think
[22:11] people need to get with the program.
[22:13] I I think this is kind of AI the
[22:16] introduction of AI has forced a lot of
[22:19] educators and then administrators to
[22:22] kind of rethink what they are assigning
[22:25] and and what they value in terms of of
[22:27] education. Is it this transactional
[22:30] place where students are there to get a
[22:33] degree that can increase their earning
[22:35] potential or to get a certain skill that
[22:37] will um increase their earning
[22:39] potential? Or is it a place where
[22:41] students go to, like I said, ideally
[22:44] expand their mind and develop critical
[22:46] thinking skills so that they can, you
[22:48] know, lead more fulfilling or or
[22:51] enriching lives?
[22:53] James, did the reporting on the story
[22:54] change the way that you
[22:56] reflect back on your own education?
[22:58] You know, I I admit, as I was reporting
[23:00] the story, I did like think about the
[23:03] handful of times when I went on
[23:05] SparkNotes or you know to to kind of get
[23:07] the reading like that.
[23:08] Really revealing your age here.
[23:12] Cheating, you know, cheating is nothing
[23:14] new. You know, cheating has been around
[23:16] forever. But as one student put it to me
[23:19] like the floor of cheating is still
[23:21] there. It's just that the ceiling has
[23:22] been blown off and the very definition
[23:24] of cheating is is starting to change. I
[23:27] shiver to think if I had this perfect
[23:30] cheating tool,
[23:32] you know, I I shiver to think about how
[23:34] my education may have been different.
[23:37] And in that sense, I it made me think
[23:39] like this is not some crisis. This is a
[23:41] moral crisis, right? Like it's not about
[23:44] this younger generation having these
[23:46] ethical lapses. It's that they have this
[23:49] tool that's available to them and we as
[23:51] a society haven't figured out this
[23:53] tool's place and how it should be
[23:55] regulated in the classroom much less
[23:58] across society.
[24:00] James, thank you so much for your time
[24:02] today.
[24:03] Oh, thanks for having me.
[24:09] That's all for today. Front Burner was
[24:11] produced this week by Matthew Amha, Joy
[24:13] Tasha Gupta, Lauren Donnelly, McKenzie
[24:16] Cameron, and Marco Luciano.
[24:18] Our intern is Katie Teeling. Our video
[24:21] producer is Evan Agard and our YouTube
[24:23] producer is John Lee. Our music is by
[24:25] Joseph Shabason. Our senior producer is
[24:28] me, Elaine Chow.
[24:29] I've also been filling in for Jamie
[24:31] Poissant along with Jonathan Mupeti. Our
[24:33] executive producer is Nick McCabe Lokos.
[24:36] Thanks for listening to Front Burner.
[24:38] Jamie's back Monday.