resistancemoney on Nostr: I’ve been on my college's integrity committee for over a decade, often as its ...
I’ve been on my college's integrity committee for over a decade, often as its chair. Plagiarism, academic dishonesty, cheating, and the like: these are our remit. I've seen it all. But times are changing. A few thoughts, then, on ChatGPT.
The most important message I have for students is not that using ChatGPT in completing their assignments is unethical, or plagiarism, or pawning off someone else's ideas or words as your own. It can be those things, but needn't be.
My message about ChatGPT, instead, is this: you've got to cover your tracks.
The advice is not about hiding misconduct or dishonesty. Because if you actually do as I suggest, you'll have no misconduct to hide.
ChatGPT-generated prose has some well-known signs, including:
- vague generalities
- hemming and hawing, especially in a summative sentence or paragraph
- 'delve', and other artifacts of the way large language models are currently trained
- pointless filler words everywhere ('complex dynamics', 'multi-faceted analysis')
- generic verbs instead of specific ones ('explores the topic' instead of 'argues that')
- claims without backing textual citations (or worse, with fake ones: pure hallucinations)
- repetitive and flattening use of the present participle
There's an architectural reason why ChatGPT and similar tools have these irritating tics. They write to the statistical median. And the statistical median isn't ungrammatical. It is, instead, mediocre. Boring. Flat. Without any real voice or message. Dull in both style and substance. It doesn't sound like anyone in particular, because it's a blend of everyone all at once, sort of like those generic 'average faces' you can find online — all smoothed over and sort of vaguely pretty but also off-putting and inhuman.
If you can identify and eliminate infelicities like the ones noted above — and doing this thoroughly requires a sentence-by-sentence look at the entire document — no one need ever know that you used ChatGPT to write your essay. Your secret will be safe.
But along the way, you'll also have made the essay your own. You'll have thought about every word. You'll have exercised agency. The words that remain will be there because you put them there, and for a reason. You'll have actually done some writing.
So, ChatGPT-using students, here is my advice once more. Cover your tracks. Ruthlessly eliminate the junk ChatGPT generates. In doing so, you'll make your paper better, and you'll make it only yours. And that is precisely what your instructors wanted from you anyways — your own best work.
The most important message I have for students is not that using ChatGPT in completing their assignments is unethical, or plagiarism, or pawning off someone else's ideas or words as your own. It can be those things, but needn't be.
My message about ChatGPT, instead, is this: you've got to cover your tracks.
The advice is not about hiding misconduct or dishonesty. Because if you actually do as I suggest, you'll have no misconduct to hide.
ChatGPT-generated prose has some well-known signs, including:
- vague generalities
- hemming and hawing, especially in a summative sentence or paragraph
- 'delve', and other artifacts of the way large language models are currently trained
- pointless filler words everywhere ('complex dynamics', 'multi-faceted analysis')
- generic verbs instead of specific ones ('explores the topic' instead of 'argues that')
- claims without backing textual citations (or worse, with fake ones: pure hallucinations)
- repetitive and flattening use of the present participle
There's an architectural reason why ChatGPT and similar tools have these irritating tics. They write to the statistical median. And the statistical median isn't ungrammatical. It is, instead, mediocre. Boring. Flat. Without any real voice or message. Dull in both style and substance. It doesn't sound like anyone in particular, because it's a blend of everyone all at once, sort of like those generic 'average faces' you can find online — all smoothed over and sort of vaguely pretty but also off-putting and inhuman.
If you can identify and eliminate infelicities like the ones noted above — and doing this thoroughly requires a sentence-by-sentence look at the entire document — no one need ever know that you used ChatGPT to write your essay. Your secret will be safe.
But along the way, you'll also have made the essay your own. You'll have thought about every word. You'll have exercised agency. The words that remain will be there because you put them there, and for a reason. You'll have actually done some writing.
So, ChatGPT-using students, here is my advice once more. Cover your tracks. Ruthlessly eliminate the junk ChatGPT generates. In doing so, you'll make your paper better, and you'll make it only yours. And that is precisely what your instructors wanted from you anyways — your own best work.