• 9 months ago
ChatGPT and other generative AI are being widely used in higher education in school admissions. But now, students who’ve turned to AI for writing are turning back to people to make that work sound more human—and schools can’t possibly keep up. Forbes technology reporter Alexandra Levine joins Brittany Lewis on "Forbes Talks" to discuss.

Read the full story on Forbes: https://www.forbes.com/sites/rashishrivastava/2024/02/05/chatgpt-college-school-applications-admissions-red-flags-ai/

Subscribe to FORBES: https://www.youtube.com/user/Forbes?sub_confirmation=1

Fuel your success with Forbes. Gain unlimited access to premium journalism, including breaking news, groundbreaking in-depth reported stories, daily digests and more. Plus, members get a front-row seat at members-only events with leading thinkers and doers, access to premium video that can help you get ahead, an ad-light experience, early access to select products including NFT drops and more:

https://account.forbes.com/membership/?utm_source=youtube&utm_medium=display&utm_campaign=growth_non-sub_paid_subscribe_ytdescript

Stay Connected
Forbes newsletters: https://newsletters.editorial.forbes.com
Forbes on Facebook: http://fb.com/forbes
Forbes Video on Twitter: http://www.twitter.com/forbes
Forbes Video on Instagram: http://instagram.com/forbes
More From Forbes: http://forbes.com

Forbes covers the intersection of entrepreneurship, wealth, technology, business and lifestyle with a focus on people and success.

Category

🤖
Tech
Transcript
00:00 (upbeat music)
00:02 - Hi everybody, I'm Brittany Lewis
00:04 with Forbes Breaking News.
00:06 Joining me now is my colleague,
00:07 Forbes tech reporter, Alexander Levine.
00:09 Alexander, thanks so much for coming in.
00:11 - Thank you so much for having me.
00:12 - A really interesting story you have today for us.
00:16 When chatGBT came into the national discourse
00:18 back in November of 2022,
00:21 one of the first questions was,
00:22 would this enable students to cheat?
00:24 You recently reported that students are using generative AI
00:27 in the college admissions process.
00:29 First, you talked to colleges, tell us what they're saying.
00:33 - Colleges really don't have consistent,
00:35 coherent policies in place to address this issue.
00:38 We have seen AI being used across the education sphere
00:42 for brainstorming ideas.
00:44 We've seen it for help putting together sentences,
00:48 but in admissions specifically, the stakes are even higher
00:51 because this is someone's future on the line.
00:53 And also, if you end up getting into a school
00:55 because you use chatGBT,
00:57 you could potentially be taking a spot from somebody
01:00 who may potentially be more deserving.
01:02 So the colleges that we reached out to across the board
01:06 were very hesitant to say anything
01:08 about what kinds of words are red flags.
01:10 However, it was clear when we reached out to them
01:12 that the policies were pretty inconsistent
01:15 and the range was so fascinating,
01:16 the range of what is allowed.
01:18 So some colleges had zero tolerance policies.
01:21 Some of them had sort of middle ground policies
01:23 saying students could use them to brainstorm.
01:25 And then some of them actually allow the use of chatGBT
01:28 in the applications because they say
01:30 that the writing sample is no longer central
01:32 to the application.
01:33 - That's interesting that they're so vague
01:35 because brainstorming, what does that really mean?
01:38 And in your research for your reporting,
01:41 do you find colleges lean into utilizing AI,
01:44 totally against it, or it just changes?
01:47 - It really seems that across the board,
01:49 they are very hesitant.
01:51 They're very hesitant about it
01:52 and they're very worried about it.
01:53 Some of them, however, are leaning in.
01:55 One of the colleges we spoke to
01:56 is actually partnering with OpenAI,
01:59 which is the creator of chatGBT
02:00 to work on some of their curricula.
02:03 But it's very interesting that across the board,
02:06 there really is no cohesive,
02:08 there's no cohesive way to address this.
02:10 And also it is worth noting that one of the colleges
02:13 that we spoke to actually has a prompt
02:15 in their application this year,
02:17 or for honor students asking them
02:20 to generate an essay using chatGBT
02:24 and then to go and critique that essay
02:26 to discuss why it's strong, maybe why it's not strong,
02:30 and to sort of parse the potential issues with it.
02:34 So I thought that that was sort of
02:35 a very interesting twist on this
02:36 that shows schools, even if they don't really
02:38 have a way to approach it,
02:39 may be trying to get ahead
02:40 or sort of look at it from a different angle.
02:43 - Something that you pointed out in your story
02:46 is that there's some dead giveaways
02:47 when it comes to writing your essay using chatGBT,
02:50 using generative AI.
02:52 Let's help some students out.
02:53 What are some of those giveaways?
02:55 - We interviewed a number of essay consultants
02:58 who do freelance work, kind of helping students
03:02 with their essays, reviewing the essays.
03:04 Interestingly, there is a huge cottage industry
03:07 of freelance consultants that has been created
03:10 by generative AI and chatGBT.
03:13 These are people who even advertise on platforms like Fiverr
03:17 that they are able to add a quote unquote human touch
03:20 or human vibes to AI generated essays.
03:23 So students are essentially hiring these people
03:26 to quote, "Unchat GPT" what they have written.
03:29 So some of the words that these essay consultants
03:33 who did not know each other,
03:34 but have all kind of across the board
03:36 looked at dozens or hundreds of these.
03:38 Some of the words that were red flags,
03:40 tapestry, beacon, something aligns seamlessly
03:45 with my aspirations.
03:46 So they're not just buzzwords,
03:48 they're also phrases and sort of also sentence structure
03:52 and syntax that are dead giveaways.
03:54 We saw things like talking about an esteemed faculty,
03:59 a vibrant community.
04:02 There were other sentence structures that were giveaways
04:04 like this is not only X, but also Y.
04:08 So it really runs the gamut.
04:09 And I think the important thing to emphasize here
04:11 is that alone, any of these words is fine
04:14 and certainly could have been generated by me,
04:17 by any human.
04:18 But what these essay consultants said was
04:21 when you see all of the words or several of them
04:25 spread out across an entire package
04:28 or across many applications,
04:30 you start to realize that maybe it's not so common
04:33 to use the word tapestry
04:34 in colloquial everyday language anymore.
04:36 So I think that was one of the most interesting learnings.
04:40 - It's really interesting also when a big talking point
04:43 when AI came onto the scene was,
04:45 am I still going to have a job?
04:47 And this is an interesting facet of
04:49 this is giving more jobs to essay consultants at least.
04:53 It does sound easier just to write your own essay,
04:55 but how easy is it for essay consultants
04:57 or probably more importantly,
04:59 admission staff to sniff out,
05:01 hey, maybe they're using tapestry too much.
05:03 This is AI generated.
05:06 - It's easier than a lot of students would probably know.
05:10 And the essay consultants had a very easy time with it.
05:13 They said, hey, we've just been reading
05:15 dozens or hundreds of these.
05:16 Imagine the admissions committees
05:18 whose full-time jobs are to read these applications
05:21 and they have read over their careers
05:23 probably many thousands of these,
05:24 potentially thousands in just one admission season.
05:27 So we actually spoke to
05:29 the Dean of Undergraduate Admissions at GW.
05:32 And what he said was,
05:34 A, yes, it is extremely easy for us to sniff these out.
05:37 B, we actually trained our committee this year
05:40 with one chat GPT generated essay
05:43 and they were able to find it quite quickly.
05:45 And three, warning to students,
05:48 most of the words that you end up having in your essays,
05:51 if they are generated by AI are hollow.
05:54 He said they're thin, he said they're flat.
05:56 And so I think that a lot of students
05:58 probably don't realize just how easy it is
06:01 for the very lightly trained eye
06:03 to be able to spot this stuff.
06:05 - We both went to college, we're both journalists.
06:08 We know how serious allegations of plagiarism, cheating are,
06:12 especially in the educational world.
06:14 So I am curious if a student was found to use AI,
06:17 are they punished on their college application at least?
06:20 - Again, a range of different responses from the schools
06:24 and some of the vagueness in how they respond,
06:26 how they explain to us what they're doing,
06:29 I think is a reflection of them
06:30 either not wanting to share
06:31 or really just not having a plan.
06:33 So there are some schools that have zero tolerance policies.
06:36 They have very clear language in their,
06:42 in their either fraud policies
06:44 that students need to agree to when they're submitting,
06:46 saying we have not used Chats GPT.
06:49 And students who agree to these policies
06:53 and use Chats GPT anyways, either a copy paste
06:56 or there's other giveaways that they've used it,
06:59 they can be accused of plagiarism.
07:01 And then I think the most extreme example that we saw,
07:03 but perhaps the largest scale one,
07:04 which was very, very important,
07:06 was that the president of Common App,
07:08 which more than a million students
07:11 use to submit their college applications every year.
07:13 And by the way, more than I think a thousand schools
07:17 are part of the Common Application.
07:20 She said that allegations of plagiarism in these essays
07:25 by using AI are investigated.
07:27 She did not say exactly how they were investigated,
07:29 but she did say that they notify the schools
07:32 that the student has been, has applied to
07:34 if the allegations turn out to be true.
07:37 - Obviously a super serious allegation
07:39 to lob against a student.
07:41 So you have to have some concrete evidence, I imagine.
07:44 What are the technologies that these universities,
07:46 Common App are using and are they foolproof?
07:49 - Some universities, and it's not clear exactly how many
07:53 or which ones, are using AI detection tools.
07:57 Now these are tools that the average person could also use.
08:00 They're not only for admissions committees,
08:03 but they're basically tools where you can input
08:05 a segment of text, a large block of text,
08:08 or in this case, an essay,
08:10 and it can spit out a likelihood that AI has been used,
08:13 or it can highlight potentially problematic sentences.
08:16 Now the schools will not say whether or not
08:21 they are using these, but what's very interesting
08:23 is that even the creators of some of these tools
08:26 who we interviewed for our reporting told us
08:29 they're so not foolproof,
08:31 and they're so not ready for prime time
08:33 that we would not suggest to admissions committees
08:36 that they use them.
08:37 Even though these founders had said that schools
08:41 and educators had in fact approached them
08:43 with questions about whether they could use
08:46 these types of solutions in an academic setting.
08:49 - We've seen this a lot in the tech space,
08:51 and you pointed this out in your reporting
08:53 that the technology often outpaces the policy.
08:57 We saw this with social media,
08:58 and now we're seeing this with chat GPT,
09:00 and I am curious, where do you see the policies going
09:04 now that we're a full year of college admissions
09:08 into chat GPT, into generative AI?
09:11 - Yeah, so as you pointed out,
09:12 we are really in the thick of the first full admission
09:15 season that has been since the explosion
09:19 and widened option of chat GPT.
09:20 The technology came out in late 2022 initially.
09:24 I think that we're already kind of most of the way
09:26 through the admissions cycle that year.
09:28 So this is really the first full season
09:30 where we're going to be seeing the effects of it.
09:32 I think the policies are trending more
09:34 towards the schools having things in writing
09:37 or figuring out, especially once they know
09:39 what they're seeing this year,
09:40 figuring out what sorts of language
09:42 that they should be putting in their policies
09:43 and maybe updating their policies.
09:45 I think the issue though is that
09:47 they're already moving slowly,
09:49 and by the time next admission season comes around,
09:52 there will be new use cases for AI,
09:54 there will be stronger tools.
09:56 Detection will be probably getting better,
09:58 it is getting better, but there will be also tools
10:01 that make it even harder to detect.
10:03 So I think that there probably will be more written policies.
10:06 I think there will probably be more public outreach
10:09 and also just education around this,
10:11 maybe done on behalf of the schools
10:12 or maybe done even on behalf of people
10:14 who are in the industry who are creating them.
10:16 But I don't think that there will be a clear way
10:20 for schools to keep up because the technology
10:23 is continuing to improve at a rapid clip.
10:26 - As a tech reporter, as someone who studies this,
10:29 is fully in the thick of it,
10:31 do you think that there's anything missing
10:32 from the national conversation
10:34 when it comes to college admissions,
10:35 when it comes to education as a whole and generative AI?
10:39 - I think one of the most interesting developments
10:41 I have seen, particularly in the last week or two,
10:43 is that some schools are now reinstating
10:46 the requirement that students take the SAT.
10:49 And that happened shortly after our story came out,
10:52 so that wasn't sort of a piece of this,
10:54 but I think anyone seeing that can look at it and think,
10:57 well, is that happening in part because
11:00 all of these other pieces of application
11:03 on which students are judged
11:05 are becoming harder and harder to keep pure.
11:08 And by keeping pure, I mean,
11:09 it's harder and harder to guarantee
11:11 that this is 100% the students' work.
11:13 You can have them sign code of conduct statements,
11:16 you can have them, you can hope that they are young
11:19 and bright and they're gonna be upstanding,
11:21 and I think probably most students really are.
11:23 But I think that you're really seeing that schools,
11:29 and we may see more schools reinstate the SAT
11:31 or other kinds of requirements
11:33 that kind of move away a little bit from the essays,
11:35 schools are probably going to pull different levers
11:39 to try to address this.
11:40 - Alexandra Levine, thanks so much for coming in.
11:43 - Thank you so much for having me.
11:44 (silence)
11:46 (silence)
11:49 (silence)
11:51 [BLANK_AUDIO]

Recommended