A whole new ballgame in cheating. Introducing ChatGPT

[Caracal]

^^ This. ^^ So far I've had one, and the student caved immediately.  The "tell" was the illiterate word salad BS of all her work to date, only to have the most recent paper pretty much flawless in its writing, though vague and ephemeral in content.  Nope.  Not going to fly. 

One suggestion from our admin is to allow students to use ChatGPT to create a first draft, then edit that for the content that ChatGPT is incapable of. The purpose is to skip the illiterate word salad stage so that the student can engage more with the learning. Instructors seem receptive at least to not have to read so much crud. The expectation for sharp arguments may rise.

This just doesn't make any sense. The point of most academic writing is to make some sort of original argument. You can't figure out what you're trying to say if you outsource your first draft. Worse, making good original arguments is exactly what ChatGPT is bad at. It makes pretend arguments that lack any real point. That's the worst thing you can have in a draft. If your paper lacks any real ideas, you can't just revise it, you have to basically start over. It doesn't matter if you have polished words on the page that seem to follow some clear organizational structure if there's no point. Bad grammar, unclear phrasing, messed up organization, transitions and argument development are all things you can fix with revision, but there has to be something worth revising in the first place. The point of that revision is to bring out the actual ideas so your reader can actually figure out what they are and to make them persuasive so they might be convinced by them. If there are no worthwhile ideas, you're just polishing a turd.

[Sun_Worshiper]

^^ This. ^^ So far I've had one, and the student caved immediately.  The "tell" was the illiterate word salad BS of all her work to date, only to have the most recent paper pretty much flawless in its writing, though vague and ephemeral in content.  Nope.  Not going to fly. 

One suggestion from our admin is to allow students to use ChatGPT to create a first draft, then edit that for the content that ChatGPT is incapable of. The purpose is to skip the illiterate word salad stage so that the student can engage more with the learning. Instructors seem receptive at least to not have to read so much crud. The expectation for sharp arguments may rise.

This just doesn't make any sense. The point of most academic writing is to make some sort of original argument. You can't figure out what you're trying to say if you outsource your first draft. Worse, making good original arguments is exactly what ChatGPT is bad at. It makes pretend arguments that lack any real point. That's the worst thing you can have in a draft. If your paper lacks any real ideas, you can't just revise it, you have to basically start over. It doesn't matter if you have polished words on the page that seem to follow some clear organizational structure if there's no point. Bad grammar, unclear phrasing, messed up organization, transitions and argument development are all things you can fix with revision, but there has to be something worth revising in the first place. The point of that revision is to bring out the actual ideas so your reader can actually figure out what they are and to make them persuasive so they might be convinced by them. If there are no worthwhile ideas, you're just polishing a turd.

If you give it the right set of instructions (including the original argument you'd like it to make), then ChatGPT can give you a pretty solid draft, including the motivation and some facts and figures (which must be vigorously fact checked). From there, revising is necessary.

Like it or not, this is where a lot of professional writing is going.

[Caracal]

^^ This. ^^ So far I've had one, and the student caved immediately.  The "tell" was the illiterate word salad BS of all her work to date, only to have the most recent paper pretty much flawless in its writing, though vague and ephemeral in content.  Nope.  Not going to fly. 

One suggestion from our admin is to allow students to use ChatGPT to create a first draft, then edit that for the content that ChatGPT is incapable of. The purpose is to skip the illiterate word salad stage so that the student can engage more with the learning. Instructors seem receptive at least to not have to read so much crud. The expectation for sharp arguments may rise.

This just doesn't make any sense. The point of most academic writing is to make some sort of original argument. You can't figure out what you're trying to say if you outsource your first draft. Worse, making good original arguments is exactly what ChatGPT is bad at. It makes pretend arguments that lack any real point. That's the worst thing you can have in a draft. If your paper lacks any real ideas, you can't just revise it, you have to basically start over. It doesn't matter if you have polished words on the page that seem to follow some clear organizational structure if there's no point. Bad grammar, unclear phrasing, messed up organization, transitions and argument development are all things you can fix with revision, but there has to be something worth revising in the first place. The point of that revision is to bring out the actual ideas so your reader can actually figure out what they are and to make them persuasive so they might be convinced by them. If there are no worthwhile ideas, you're just polishing a turd.

If you give it the right set of instructions (including the original argument you'd like it to make), then ChatGPT can give you a pretty solid draft, including the motivation and some facts and figures (which must be vigorously fact checked). From there, revising is necessary.

Like it or not, this is where a lot of professional writing is going.

Maybe for technical writing, but definitely not for argumentative writing of the kind that people in the humanities and many of the social sciences do. 

[Sun_Worshiper]

^^ This. ^^ So far I've had one, and the student caved immediately.  The "tell" was the illiterate word salad BS of all her work to date, only to have the most recent paper pretty much flawless in its writing, though vague and ephemeral in content.  Nope.  Not going to fly. 

One suggestion from our admin is to allow students to use ChatGPT to create a first draft, then edit that for the content that ChatGPT is incapable of. The purpose is to skip the illiterate word salad stage so that the student can engage more with the learning. Instructors seem receptive at least to not have to read so much crud. The expectation for sharp arguments may rise.

This just doesn't make any sense. The point of most academic writing is to make some sort of original argument. You can't figure out what you're trying to say if you outsource your first draft. Worse, making good original arguments is exactly what ChatGPT is bad at. It makes pretend arguments that lack any real point. That's the worst thing you can have in a draft. If your paper lacks any real ideas, you can't just revise it, you have to basically start over. It doesn't matter if you have polished words on the page that seem to follow some clear organizational structure if there's no point. Bad grammar, unclear phrasing, messed up organization, transitions and argument development are all things you can fix with revision, but there has to be something worth revising in the first place. The point of that revision is to bring out the actual ideas so your reader can actually figure out what they are and to make them persuasive so they might be convinced by them. If there are no worthwhile ideas, you're just polishing a turd.

If you give it the right set of instructions (including the original argument you'd like it to make), then ChatGPT can give you a pretty solid draft, including the motivation and some facts and figures (which must be vigorously fact checked). From there, revising is necessary.

Like it or not, this is where a lot of professional writing is going.

Maybe for technical writing, but definitely not for argumentative writing of the kind that people in the humanities and many of the social sciences do.

If you mean the sort of argumentative writing that we as professors publish in academic journals, then you are correct (for now). If you mean argumentative writing that social science undergraduate students frequently do as homework assignments, then I must respectfully disagree.

[Sun_Worshiper]

And it can help people doing more serious academic writing as well, even though it can't write the whole piece or a serviceable draft just yet. I explain some ways I've been using it above and I must say that it is a very useful tool.

[Caracal]

And it can help people doing more serious academic writing as well, even though it can't write the whole piece or a serviceable draft just yet. I explain some ways I've been using it above and I must say that it is a very useful tool.

Sure, I just think the suggestion that undergrads writing a first draft could helpfully use chatgpt is silly.

[Caracal]

This is a good piece. https://www.washingtonpost.com/opinions/2023/04/24/artificial-intelligence-consciousness-thinking/

I especially liked this part, "Chat is like a person who is barely paying attention, which is understandable because — being unconscious — Chat is definitely not paying attention. It couldn't possibly be paying attention." This really helps to explain why chat actually does right like many of my students. Many of them are also not paying attention.

[Stockmann]

This is a good piece. https://www.washingtonpost.com/opinions/2023/04/24/artificial-intelligence-consciousness-thinking/

I especially liked this part, "Chat is like a person who is barely paying attention, which is understandable because — being unconscious — Chat is definitely not paying attention. It couldn't possibly be paying attention." This really helps to explain why chat actually does right like many of my students. Many of them are also not paying attention.

And if those students were paying attention, perhaps they'd worry that AI is doing the same as them - but cheaper and faster. It used to be that skilled white collar workers mostly benefited from automation (although PCs could replace basic secretarial work) but that era seems to have come to an end and it will not be better when these students graduate. Combine that with grade and degree inflation (the US used to teach Latin in HS instead of having remedial English in college), and college costs, and the weaker students should be really, really worried.

[quasihumanist]

This is a good piece. https://www.washingtonpost.com/opinions/2023/04/24/artificial-intelligence-consciousness-thinking/

I especially liked this part, "Chat is like a person who is barely paying attention, which is understandable because — being unconscious — Chat is definitely not paying attention. It couldn't possibly be paying attention." This really helps to explain why chat actually does right like many of my students. Many of them are also not paying attention.

And if those students were paying attention, perhaps they'd worry that AI is doing the same as them - but cheaper and faster. It used to be that skilled white collar workers mostly benefited from automation (although PCs could replace basic secretarial work) but that era seems to have come to an end and it will not be better when these students graduate. Combine that with grade and degree inflation (the US used to teach Latin in HS instead of having remedial English in college), and college costs, and the weaker students should be really, really worried.

I think we'll soon have a society where half of us are disabled.

[aprof]

I gave ChatGPT (3.5-the free one) a chance at my final exam for my MS-level engineering class.  It was pretty ugly.  I didn't calculate an exact grade but I would say it was worse than the lowest student score. As usual, it provided verbose answers that sound close if you give just a glance but much of it made no sense, incorrect equations were selected, wrong units were used, and equations were evaluated incorrectly.  Surprisingly it did poorly on some of the multiple choice and true/false conceptual questions that I thought would go ok.  Of course if we take a ratio of correct answers to time spent, it did quite well since it took ~2 minutes to compose answers for which most students took >2 hours.

I worry about young adults today seeing the constant advice from influencers that you can "learn anything in a day" from ChatGPT.  It spits out accurate-sounding pseudo-science like a well-trained flat earther, and most non-experts are going to have a difficult time discerning truth from hallucination.  It's likely to make our Dunning-Kruger problems worse.

[apl68]

ChatGPT sounds like a Vegematic for producing word salad.

[Larimar]

ChatGPT sounds like a Vegematic for producing word salad.

LOL! Good one!

[the_geneticist]

I've found at least two students using ChatGPT on their online makeup lab assignments. That's an easy 0 for the grade book. 
The mismatch between the questions and answers makes it pretty darn obvious since the program didn't watch the video, look at the dataset, read the protocol, etc

[marshwiggle]

I've found at least two students using ChatGPT on their online makeup lab assignments. That's an easy 0 for the grade book. 
The mismatch between the questions and answers makes it pretty darn obvious since the program didn't watch the video, look at the dataset, read the protocol, etc

But, to be fair, it probably didn't spend the evening before the assignment was due drinking and smoking weed.

[history_grrrl]

I'm pleased to report that I had an extremely productive meeting with my uni's academic integrity coordinator (who I didn't know existed until recently). Person was very knowledgeable and actually thinks it's important to promote academic integrity! Yay.