A whole new ballgame in cheating. Introducing ChatGPT

[Parasaurolophus]

It's pretty dinky. Remember that they think parental leave means you get paid a percentage of courses you already taught...

[Sun_Worshiper]

I'm pretty sure many of my students are using this for discussion boards in which they respond to quick prompts. Contrary to many naysayers in this thread and elsewhere, the responses are good enough to get the points and I have no credible proof that students are cheating. Is it good enough to write an academic paper or a work of great fiction? Obviously not (yet) but, just as many of us expected, it is easier for students to cheat and harder for faculty to stop them.

[Wahoo Redux]

IHE: Leading Scientists Worldwide Are Victims of Fake Articles

They are planning legal action over pieces written with artificial intelligence.

In recent months, academics at leading universities in Australia, Europe and North America have been alerted to low-quality scholarly articles—often little more than a page long, probably written by a language-scraping algorithm—appearing under their names in titles published by Prime Scholars, an open-access publisher registered to a west London address. That office, where hundreds of British companies are incorporated, is also home to other digital periodical companies whose authors are usually from India, the Middle East or developing economies.

[Caracal]

I'm pretty sure many of my students are using this for discussion boards in which they respond to quick prompts. Contrary to many naysayers in this thread and elsewhere, the responses are good enough to get the points and I have no credible proof that students are cheating. Is it good enough to write an academic paper or a work of great fiction? Obviously not (yet) but, just as many of us expected, it is easier for students to cheat and harder for faculty to stop them.

I'm sure some of your students have always just cut and pasted. Perhaps, done some light paraphrasing of their own. I'm sure CHATGPT does just fine with this stuff. It sounds like you use these sorts of assignments the way I do, which is just to create an expectation that students are going to do the reading and engage with it at least a little bit. These kinds of assignments are always vulnerable to cheating because we don't want students to need to spend lots of time writing them, don't want to punish them for misunderstanding a reading before we discuss it in class, and we definitely don't want to spend lots of time grading low stakes assignments.

I start off with the assumption that some students will never do the readings and just try to evade any assignment through some combination of cheating and BS. I'm really just trying to communicate to everyone else that the readings aren't optional. It generally works, enough people do the readings to have a discussion in class. It's always worth keeping an eye on whether things remain effective. The danger with cheating is when it becomes so pervasive that people start doing it even when it really isn't an effective strategy for them. Kind of like how I realized in the first semester of teaching at a large school with lots of commuters that I had to take attendance because by the end of the semester less than a third of the class was coming and even the better students were starting to skip classes. Nobody was doing themselves any favors, if they didn't come to class they were missing a lot of content that was on the exams and the grades reflected that, but at some point people do become effected by the larger social context.

Students who actually do the reading so they can write a response on a board or complete a quiz are aware that they could write the response without doing the reading. For most of them, cheating or even BSing probably just makes them nervous-and they understand that it wouldn't actually be an effective work around in the long run. If they do the reading, they can understand what's happening in class and will do better on exams. They may need the nudge of an assignment to do it, but they aren't going to evade that nudge. It's always possible CHATGPT could make evading the reading so easy, so simple and so widespread that it will become a problem. Or not. Post COVID, I still count attendance and drop a few attendance grades, but I excuse any absences above that if students just tell me they are sick. Abseentism is probably up a little, but not that much. I'm sure some students lie and tell me they were sick when they weren't, but most people don't actually feel great about lying-when they're laying in bed thinking about how nice it would be to just not go to class and the thought occurs to them "I could just write him and say I'm sick," they think, "no, no, I'd just feel guilty and besides I should go to class" and they get up and do it.

[apl68]

It occurs to me that we're soon going to have absolutely enormous quantities of junk written by AIs all over our internet.  Individually most of these items probably won't be able to fool anybody who knows how to take some care with what they read.  But it's going to be contaminating the ocean of material that the AIs use to "write" their pieces.  Which suggests that we may be creating a feedback loop of some kind that's going to have unpredictable, but surely not good, consequences.

[Parasaurolophus]

It occurs to me that we're soon going to have absolutely enormous quantities of junk written by AIs all over our internet.  Individually most of these items probably won't be able to fool anybody who knows how to take some care with what they read.  But it's going to be contaminating the ocean of material that the AIs use to "write" their pieces.  Which suggests that we may be creating a feedback loop of some kind that's going to have unpredictable, but surely not good, consequences.

I believe that happened with some AI artbots earlier this year or late last year.

[Hegemony]

Somewhat relatedly, I work for a magazine that publishes fiction. The number of submissions written by ChatGPT has absolutely exploded. They are all completely terrible. ChatGPT may write passable expository prose, but its fiction shows that it doesn't get the concept. The people submitting these things presumably don't get the concept either, and can't tell that the submission is awful.

So far most of the submitters are foreign, and their cover letter is something like, "I want to working for wonderful publication. This is column story I write for you enjoy."

Then the story itself is something like "CHAPTER ONE. A little girl was walking down the road when she met a monster. She had never met a monster before but she was happy to see one. CHAPTER TWO. She became friends with the monster and learned that new friends can be everywhere." I mean that's not the summary, that's the actual story.

It's so exhausting weeding these out that the submission system is developing an AI way to spot and eliminate them. Spam for publications. I wonder how many stupid ChatGPT articles the academic journals are getting.

[Katrina Gulliver]

Wow. I'm not surprised students use it for an essay for a required class: but people who (presumably?) want to be writers are firing off AI short stories to Ploughshares and the New Yorker?! That's incredible.

[Caracal]

Wow. I'm not surprised students use it for an essay for a required class: but people who (presumably?) want to be writers are firing off AI short stories to Ploughshares and the New Yorker?! That's incredible.

It helps to remember that this is usually teenagers. Adults in Nigeria and China also have better things to do with their time.

[Sun_Worshiper]

I'm pretty sure many of my students are using this for discussion boards in which they respond to quick prompts. Contrary to many naysayers in this thread and elsewhere, the responses are good enough to get the points and I have no credible proof that students are cheating. Is it good enough to write an academic paper or a work of great fiction? Obviously not (yet) but, just as many of us expected, it is easier for students to cheat and harder for faculty to stop them.

I'm sure some of your students have always just cut and pasted. Perhaps, done some light paraphrasing of their own. I'm sure CHATGPT does just fine with this stuff. It sounds like you use these sorts of assignments the way I do, which is just to create an expectation that students are going to do the reading and engage with it at least a little bit. These kinds of assignments are always vulnerable to cheating because we don't want students to need to spend lots of time writing them, don't want to punish them for misunderstanding a reading before we discuss it in class, and we definitely don't want to spend lots of time grading low stakes assignments.

I start off with the assumption that some students will never do the readings and just try to evade any assignment through some combination of cheating and BS. I'm really just trying to communicate to everyone else that the readings aren't optional. It generally works, enough people do the readings to have a discussion in class. It's always worth keeping an eye on whether things remain effective. The danger with cheating is when it becomes so pervasive that people start doing it even when it really isn't an effective strategy for them. Kind of like how I realized in the first semester of teaching at a large school with lots of commuters that I had to take attendance because by the end of the semester less than a third of the class was coming and even the better students were starting to skip classes. Nobody was doing themselves any favors, if they didn't come to class they were missing a lot of content that was on the exams and the grades reflected that, but at some point people do become effected by the larger social context.

Students who actually do the reading so they can write a response on a board or complete a quiz are aware that they could write the response without doing the reading. For most of them, cheating or even BSing probably just makes them nervous-and they understand that it wouldn't actually be an effective work around in the long run. If they do the reading, they can understand what's happening in class and will do better on exams. They may need the nudge of an assignment to do it, but they aren't going to evade that nudge. It's always possible CHATGPT could make evading the reading so easy, so simple and so widespread that it will become a problem. Or not. Post COVID, I still count attendance and drop a few attendance grades, but I excuse any absences above that if students just tell me they are sick. Abseentism is probably up a little, but not that much. I'm sure some students lie and tell me they were sick when they weren't, but most people don't actually feel great about lying-when they're laying in bed thinking about how nice it would be to just not go to class and the thought occurs to them "I could just write him and say I'm sick," they think, "no, no, I'd just feel guilty and besides I should go to class" and they get up and do it.

Yes, students have always cheated - but now it is way easier for them (and for more of them) to cheat and to not get caught.

Edit, sorry I'm being a little knee jerk - yes I get it that we have to evolve and that not all students will cheat, but let's also be realistic about the problems that are being created here.

[Larimar]

This is disturbing.

https://www.npr.org/2023/03/02/1159895892/ai-microsoft-bing-chatbot

[Cheerful]

This is disturbing.

https://www.npr.org/2023/03/02/1159895892/ai-microsoft-bing-chatbot

Thanks for posting that, Larimar.  Informative and concerning.

[marshwiggle]

This is disturbing.

https://www.npr.org/2023/03/02/1159895892/ai-microsoft-bing-chatbot

Thanks for posting that, Larimar.  Informative and concerning.

From the article:

Even scholars in the field of AI are not exactly sure how and why chatbots can produce unsettling or offensive responses.

The engine of these tools — a system known in the industry as a large language model — operates by ingesting a vast amount of text from the internet, constantly scanning enormous swaths of text to identify patterns. It's similar to how autocomplete tools in email and texting suggest the next word or phrase you type. But an AI tool becomes "smarter" in a sense because it learns from its own actions in what researchers call "reinforcement learning," meaning the more the tools are used, the more refined the outputs become.

They're basically doing what all kinds of scifi AIs and/or aliens have done over the decades. "Observing" human behaviour (either by studying history or monitoring online content) makes it obvious that humanity runs the gamut from saintly to evil. For millenia it has generally been assumed by all major religions that good behaviour takes effort. It's only relatively recently that materialists have suggested that benevolence is somehow natural. Maybe AIs are giving us some metaphysical insight.

[the_geneticist]

As a crocheter (and a composition teacher), I absolutely loved this article:  "ChatGPT Keeps Imploding Because of Crochet. (Seriously.)."

I'm off to ask it to write me a pattern for a stuffed elephant.

I love this!  Almost tempted to get an account just to ask it for patterns.  I've been telling folks that knitting and crochet patterns are a lot like coding: lots of abbreviations, conventions, logic, and can go hilariously wrong if you make a mistake.

[Langue_doc]

As a crocheter (and a composition teacher), I absolutely loved this article:  "ChatGPT Keeps Imploding Because of Crochet. (Seriously.)."

I'm off to ask it to write me a pattern for a stuffed elephant.

I love this!  Almost tempted to get an account just to ask it for patterns.  I've been telling folks that knitting and crochet patterns are a lot like coding: lots of abbreviations, conventions, logic, and can go hilariously wrong if you make a mistake.

Ha! That's the understatement of the year. I still remember knitting what I thought was a little purse, except that there was no opening due to not paying attention to one of the details. I had to start all over again after my mother had unraveled the non-bag bag.