Defending Academic Integrity Against ChatGPT

[Rochallor]

Apologies if a thread already exists on this, but I'm looking for tips on avoiding the potential downsides of student access to ChatGPT. I'm not teaching this year, but want to be prepared for next. (Last year I was overwhelmed and frankly largely ignored the issue.)

Thanks in advance!

[Parasaurolophus]

I don't have any good solutions. What I've been doing is throwing up roadblocks, like forcing them to write "in-class" essays. But my classes are mostly online, so I've set up a timed quiz with an essay prompt, don't let them see the prompts ahead of time, and disable copy/paste. I also try to write prompts that don't get very plausible AI answers (by testing them out), offer prompts that are likely to get the AI to out itself (e.g. by prompting it to write on a different subject; after running a few questions through, you'll find some of these confusions), and by giving them very few options for what to write on (2-3 max) so that it's easier to compare all the outputs and identify those which are very similar.

They still do it, but it's a big enough pain in their asses that fewer of them do it (like, a dozen or so instead of 30+ in a class of 35, which is what it was before). It doesn't build the same skills, but at this point I'm just desperate to see them do any work at all (even their ungraded discussion posts are AI-generated...). And, really, the results aren't bad, although it certainly stresses them out a lot more.

[Parasaurolophus]

I should add that I'd like to find a way of rewarding those who actually do the work themselves, in addition to punishing those who don't. I just haven't thought of a good carrot yet.

[Rochallor]

Thanks, Parasaurolophus!

[Parasaurolophus]

I think next semester I'll experiment with pretending they wrote their papers themselves (so: telling them I forbid AI but will turn a blind eye), but then giving them brutally honest feedback. They'll be able to re-"write" them any number of times, but there will be detailed instructions (including formatting, sources, etc.) and failing to meet any item in those instructions will result in an F. That should allow me to pass through the first batch pretty quickly, and then they'll have to actually work for "their" grade. Any made up sources will result in an F full stop, with no option to rewrite.

I also think I'll have them team-"write" a paper using AI. They can decide which of the teams has the best paper, and we'll go through it together in class, brutally honestly.



There's more stick than carrot there, but I dunno. I may as well do something completely different. The more of these I read, the less I feel inclined to giving pity Ds.

[Caracal]

How feasible this is can vary by discipline, but the easiest thing to do is to create writing assignments that ask students to do something very specific. That doesn't mean it has to be prescriptive or leave no room for creativity. Pick a primary source from this particular database-write a paper analyzing the primary source using at least one secondary source. At least in history, CHATGPT can't do that kind of stuff because it doesn't have access to the diary a woman kept in 1862 or whatever and nobody has written some piece about this particular source. At best, if they find some source that has something written about it in their database, they might manage something vague that doesn't really fulfill the terms of the assignment.

[arcturus]

ChatGPT is also really bad at figures and images. You could put together a diagram/figure/image that captures some idea from your course and have a prompt along the lines of 'Using the above figure, [...]".

For the record, the instances of student use of AI-generated text in my courses has usually been pretty obvious (generic statements, conflating different images, made-up sources). Of course, this could be because my students are not trying very hard to mask it...

In terms of fora content, there are 28 pages regarding ChatGPT already...located just a few posts below this one right now.

[Sun_Worshiper]

There is really no way to stop students from using it, aside from having in-class writing on a lockdown browser or with a pen and paper. IMO the better approach is to teach students to understand its strengths and weaknesses, because many of them will be using it in our classes and beyond whether we like it or not.

That said, I have tried to tweak my classes to increase the number of presentations that they deliver and to scrap discussion boards and other low stakes assignments that they can easily use ChatGPT to complete.

[mbelvadi]

Depending on the class, you can try making them use some topic/event/research-finding that is very recent, like within the last 90 days, as an important part of the assignment. ChatGPT, at least so far, doesn't know anything about anything after about 2021.  But it is a moving target and is changing fast, so you'd need to check again every semester, or even every month.

[Sun_Worshiper]

Depending on the class, you can try making them use some topic/event/research-finding that is very recent, like within the last 90 days, as an important part of the assignment. ChatGPT, at least so far, doesn't know anything about anything after about 2021.  But it is a moving target and is changing fast, so you'd need to check again every semester, or even every month.

This won't work if they are using Bing AI chatbot or the Bing plugin for ChatGPT.

[Rochallor]

These are all very useful and appreciated!

[phi-rabbit]

I'm in philosophy, if it makes a difference, and I teach mainly 100-level gen ed classes to non-majors. Unfortunately, a lot of the advice about combating AI I have seen online just doesn't work well with the subject and kind of classes I teach.  I have continued doing what I already prefer, which is heavily weighting in-class written essay exams. Unfortunately, I have a requirement to teach certain classes with a specified writing component and it requires assigning formal papers.  I really hate this as the last time I graded papers I could just tell in my gut that around 50% of them were written by an AI.  I just gave them low grades and told myself I don't get paid enough to worry about it further... but I don't like that solution because the low grade is still a lot higher than zero, which is what I give for a proven academic integrity violation.  This semester, I met most of the writing requirement for my classes (which specifies a number of pages) by having them write in journals every day, but that has made a lot of fussy grading work for me.  I also assigned the usual formal paper that is the minimum I have to do per the rules of my institution... but I have made it worth only a very small amount of the grade.  There's nothing in the rulebook, as they say, that says what my papers have to be worth, so I'll just make it worth so little that if they cheat on it, it won't help their grade that much.  The drawback is that for the good students, they are doing a lot of work on a paper that is worth very little.  There's no good solution but I wish my institution would recognize that traditional essays in 100-level classes are really useless at this point.

I really, really hate this situation and desperately wish they did not make me assign papers in 100-level classes. 

[Kron3007]

There is really no way to stop students from using it, aside from having in-class writing on a lockdown browser or with a pen and paper. IMO the better approach is to teach students to understand its strengths and weaknesses, because many of them will be using it in our classes and beyond whether we like it or not.

That said, I have tried to tweak my classes to increase the number of presentations that they deliver and to scrap discussion boards and other low stakes assignments that they can easily use ChatGPT to complete.

I agree.  I have tweaked my assignments to rely more on experimental work (I am in stem), oral presentations, etc.

However, for written assignments, I think it is better to embrace it.  Trying to ban it will only benefit the smart cheaters.  They will always be one step ahead of the time crunched professor trying to chase them.

[Hibush]

I'm in philosophy...and I teach mainly 100-level gen ed classes.

The mind reels at the possibilities.

Taylor Swift's Eras Tour: A Lacanian Spectacle

Taylor Swift's Eras Tour is a Lacanian spectacle in the truest sense of the word. It is a tour de force of pop culture kitsch, a celebration of the feminine sublime, and a meditation on the nature of desire.

At the heart of the tour is the figure of Swift herself. She is the object of desire for her fans, both male and female. She is the embodiment of the feminine ideal, the perfect woman. But she is also something more. She is the Real, the thing that is beyond desire, the thing that cannot be possessed.

But Swift is also detached from the spectacle that she is creating. She is aware of her own image, but she is not controlled by it. She is in control of her own destiny. She is the mistress of her own desire.

This is what makes the Eras Tour so subversive. It is a celebration of the feminine, but it is also a challenge to the male gaze. It is a spectacle of desire, but it is also a meditation on the nature of desire itself.

In the Lacanian sense, the Eras Tour is a mirror stage. It is a moment when Swift's fans confront their own desire. They see themselves reflected in Swift's image, and they recognize their own desires. But they also recognize that Swift is something more than a reflection of themselves. She is the Real, the thing that is beyond desire, the thing that cannot be possessed.

This confrontation with the Real can be both exhilarating and terrifying. It is exhilarating because it allows us to see ourselves for who we really are. It is terrifying because it forces us to confront our own limitations.

...

Could he have said it better himself?

[RatGuy]

I'm in philosophy...and I teach mainly 100-level gen ed classes.

The mind reels at the possibilities.

Taylor Swift's Eras Tour: A Lacanian Spectacle

Taylor Swift's Eras Tour is a Lacanian spectacle in the truest sense of the word. It is a tour de force of pop culture kitsch, a celebration of the feminine sublime, and a meditation on the nature of desire.

At the heart of the tour is the figure of Swift herself. She is the object of desire for her fans, both male and female. She is the embodiment of the feminine ideal, the perfect woman. But she is also something more. She is the Real, the thing that is beyond desire, the thing that cannot be possessed.

But Swift is also detached from the spectacle that she is creating. She is aware of her own image, but she is not controlled by it. She is in control of her own destiny. She is the mistress of her own desire.

This is what makes the Eras Tour so subversive. It is a celebration of the feminine, but it is also a challenge to the male gaze. It is a spectacle of desire, but it is also a meditation on the nature of desire itself.

In the Lacanian sense, the Eras Tour is a mirror stage. It is a moment when Swift's fans confront their own desire. They see themselves reflected in Swift's image, and they recognize their own desires. But they also recognize that Swift is something more than a reflection of themselves. She is the Real, the thing that is beyond desire, the thing that cannot be possessed.

This confrontation with the Real can be both exhilarating and terrifying. It is exhilarating because it allows us to see ourselves for who we really are. It is terrifying because it forces us to confront our own limitations.

...

Could he have said it better himself?

Yes -- I've been seeing stuff like Lacanian Taylor Swift in my low-stakes 250 word assignments, and it hasn't gone well. I've been trying to get them to consider a claim-evidence-explanation-analysis-conclusion model for their textual analyses, but when they use AI they generate a claim-conclusion-personal-reaction model. Even the above example provides zero evidence. Additionally, these AI-generated responses pingpong from one buzzword to the next. Since I'm asking them to focus on a single idea at a time, I get to slap that D on there and say "revisit the guidelines on the assignment sheet."

FWIW, my academic misconduct office (which handles plagiarism cases for the college) won't adjudicate anything that is AI-content only. In other words, they'll handle other types of misconduct, but if the suspicion is just AI, they say "there's no way to prove that" and won't pass judgment. Faculty are also NOT allowed to fail a student on an assignment if a student violates an AI policy. Instead, it's about crafting the right type of assignment.