New inventive ways of cheating

[Hegemony]

Boy, the GPTchat has been busy writing my students' papers this semester. (Yes, I am not letting them get away with this. By which I don't mean that I am patiently explaining academic dishonesty to them yet again — I am launching the process for a full university investigation, with take-no-prisoners penalties if they're found guilty. Enough is enough.)

And now I have several I suspect of having run some text through a paraphraser. Typically, the paper doesn't address the question — it's just a meandering series of superficial observations about the subject, in incredibly convoluted language. For instance, "This ignites many wonders, educing your mind to consider outlooks."

Is there a way to try to trace this kind of thing back to its original form, short of just making guesses?

[Morden]

Interesting question. I haven't tried this myself, but what if you ran the gobbledygook sentences through an online paraphraser--would it unscramble it enough for you to find the original?

[Hegemony]

There are a lot of paraphrasers out there. When I ran some sentences through two of them, I usually ended up with something coherent, but then when I googled the result, plus the name of the text, I got nothing.

[Caracal]

Boy, the GPTchat has been busy writing my students' papers this semester. (Yes, I am not letting them get away with this. By which I don't mean that I am patiently explaining academic dishonesty to them yet again — I am launching the process for a full university investigation, with take-no-prisoners penalties if they're found guilty. Enough is enough.)

And now I have several I suspect of having run some text through a paraphraser. Typically, the paper doesn't address the question — it's just a meandering series of superficial observations about the subject, in incredibly convoluted language. For instance, "This ignites many wonders, educing your mind to consider outlooks."

Is there a way to try to trace this kind of thing back to its original form, short of just making guesses?

Is it worth it? If the paper doesn't address the question and doesn't make any argument, can't you just give it a D or an F and move on? If you see definitive evidence of cheating or some checking turns something up, then by all means, pursue it, but if cheating results in bad grades, then you're combating it.

[Caracal]

There are a lot of paraphrasers out there. When I ran some sentences through two of them, I usually ended up with something coherent, but then when I googled the result, plus the name of the text, I got nothing.

Well, it isn't going to paraphrase it into the original, it's just going to paraphrase it in into something else.

[downer]

You look for proper names, numbers and possibly technical terms that can't be paraphrased or put into other words. Then you put them into Google and hope to find the passage the student used. Once you find it, you can easily compare the two and show the obvious parallels.

Paraphrasing tools are getting better. Some produce rather good English, which is making it much harder to detect.

I am starting to think that a large proportion of my class time will be giving multiple choice tests. I will also get them to do presentations in class. I will provide lectures online for them to watch, or they can just do the reading. I may occasionally get them to write essays in the classroom in Blue Books, but I really don't want to have to try to decipher their handwriting. It looks like it will soon be pointless to assign work submitted online.

[fishbrains]

Boy, the GPTchat has been busy writing my students' papers this semester. (Yes, I am not letting them get away with this. By which I don't mean that I am patiently explaining academic dishonesty to them yet again — I am launching the process for a full university investigation, with take-no-prisoners penalties if they're found guilty. Enough is enough.)

And now I have several I suspect of having run some text through a paraphraser. Typically, the paper doesn't address the question — it's just a meandering series of superficial observations about the subject, in incredibly convoluted language. For instance, "This ignites many wonders, educing your mind to consider outlooks."

Is there a way to try to trace this kind of thing back to its original form, short of just making guesses?

Is it worth it? If the paper doesn't address the question and doesn't make any argument, can't you just give it a D or an F and move on? If you see definitive evidence of cheating or some checking turns something up, then by all means, pursue it, but if cheating results in bad grades, then you're combating it.

This is the direction I'm starting to go in. I'm finding it hard to see another path, at least for my composition courses.

[dismalist]

These are fascinating examples of how markets get destroyed by information asymmetries. It's nothing new. Remember the used car market ca. 1970? Sellers, who knew the quality of their cars had no incentive to tell the truth about them. Lesson: Don't buy a used car, 'cause you'll get a lemon! [And don't sell your good car 'cause you'll only get the price of a lemon for it.] Things have improved since then through warranties and such.

Easy cheating pollutes all on-line activity, teaching, testing, whatever. I foresee a technological arms race to prevent cheating. Arms races cost money, lot's of it.

If the control of cheating does not win, even with lots of arms, on-line anything is dead meat.

I suppose colleges can win if they keep the on-line offers secret! :-)

[Hegemony]

Is it worth it? If the paper doesn't address the question and doesn't make any argument, can't you just give it a D or an F and move on? If you see definitive evidence of cheating or some checking turns something up, then by all means, pursue it, but if cheating results in bad grades, then you're combating it.

Yeah, I could just give them a bad grade, but these cheaters are just intolerable and I want to levy maximum penalties. The maximum penalty here, once the academic dishonesty is proven, is an F for the course, and I will happily fail this lying cheating student. It also goes on the student's record, and after a certain number of verified reports, the student is expelled. This is the shot across the bow: "We take cheating seriously here. Don't even try." So — I don't have to pursue it — but I want to.

[the_geneticist]

Well, you could always try the "tell me how your wrote this" approach.  Ask for notes, drafts, etc.  They might just confess.

[Hegemony]

Yes, there's some vocabulary in the paper (apart from the suspicious "educing") that I suspect no average undergraduate would know. I am going to ask the student to explain some of the vocabulary to me. It's annoying not to be able to find the source text, though. It's most likely on the internet in an obvious place, since I don't think a student would work very hard to find it.

[Caracal]

Yes, there's some vocabulary in the paper (apart from the suspicious "educing") that I suspect no average undergraduate would know. I am going to ask the student to explain some of the vocabulary to me. It's annoying not to be able to find the source text, though. It's most likely on the internet in an obvious place, since I don't think a student would work very hard to find it.

Yeah but if the student took some random snippets from a few different places and the paraphraser mangled it thoroughly, it may be totally unrecognizable. It's also always possible the student put their own words through the paraphraser...At some point, it might be best to just decide it's a very bad paper by a student who has no idea what they are doing and not spend any more time trying to figure out the origin story of the terrible paper.

[apl68]

These are fascinating examples of how markets get destroyed by information asymmetries. It's nothing new. Remember the used car market ca. 1970? Sellers, who knew the quality of their cars had no incentive to tell the truth about them. Lesson: Don't buy a used car, 'cause you'll get a lemon! [And don't sell your good car 'cause you'll only get the price of a lemon for it.] Things have improved since then through warranties and such.

Easy cheating pollutes all on-line activity, teaching, testing, whatever. I foresee a technological arms race to prevent cheating. Arms races cost money, lot's of it.

If the control of cheating does not win, even with lots of arms, on-line anything is dead meat.

I suppose colleges can win if they keep the on-line offers secret! :-)

It would be ironic if online cheating ended up wrecking online education.  Talk about Silicon Valley scoring on its own goal!

[history_grrrl]

I remember the first time I saw one of these bizarre papers full of just slightly "off" synonyms that didn't fit. I think the tip-off was "collieries" instead of "mines." What student would even know the word "collieries," let alone use it in a paper? I posted a query on the old Fora and learned about spinners. Mind blown.

I saw an online discussion about this recently; a student had referred in her paper to - get ready to guffaw - Steve Occupations.

Agree about having to move back to in-person work. My ChatGPT papers with quotations and citations are easy-ish to catch, because the quotations and citations are made up - and that's falsification whether by the student or AI. It's the essay exams with no detail that are stumping me, because I can assign poor grades but can't prove AI. What I hope my chair will agree to for my current cases is my offering Stu either 0s on his exams (he pulled this bs in two classes) or a chance to take a new exam in person. And yeah, I think I'll have to go back to timed exams in the giant gym, which I hate; most students do much better on my exams when they can take their time, rifle through their notes, etc. These cheaters are ruining it for everyone.

[apl68]

I can just about see a student knowing "collieries."  Hey, I once had one correctly use the word "crapulent" in an essay.  Multiple odd word uses like that all at once would be a real red flag, though.