AI Detector

[Neapolitan Ice Cream]

So I've been working on my dissertation for my CBT diploma and the topic is one close to my heart, so I wrote it from my heart… now, I passed with through the plagiarism checker. No plagiarism. Good. Now for the AI detector. 27% AI written! The hell?

 

[fla4fun]

So I've been working on my dissertation for my CBT diploma and the topic is one close to my heart, so I wrote it from my heart… now, I passed with through the plagiarism checker. No plagiarism. Good. Now for the AI detector. 27% AI written! The hell?

How does an AI detector work? What is it looking for? I mean, if you write a factual paper and the facts are out there on the internet, is it going to assume you pulled them from the internet using AI? Does it tell you which parts of your paper contain the alleged AI content?

I am all for being able to test for plagiarism and using AI to cheat in an academic setting, but the software has to be able to determine these things with great precision if a person’s grade depends on the result.

 

[Neapolitan Ice Cream]

It's a highly personal bit of matter, which is why it feels like such a blow to be told that you sound like a bot. I think it operates on the language you use, repeating a certain word or something. But for the life of me I can't figure that out. The text that it highlights as AI generated are actually the most personal texts that I wrote!

 

[fla4fun]

It's a highly personal bit of matter, which is why it feels like such a blow to be told that you sound like a bot. I think it operates on the language you use, repeating a certain word or something. But for the life of me I can't figure that out. The text that it highlights as AI generated are actually the most personal texts that I wrote!

Not knowing the content of your paper, could you use phrasing to indicate that those bits are your opinion or somehow indicate a personal connection? Would that throw off the detector? My experience with AI is a bit limited, but it appears to summarize and generalize a lot, so taking the phrasing in the opposite direction might help. Just a guess on my part . . .

 

[Neapolitan Ice Cream]

Not knowing the content of your paper, could you use phrasing to indicate that those bits are your opinion or somehow indicate a personal connection? Would that throw off the detector? My experience with AI is a bit limited, but it appears to summarize and generalize a lot, so taking the phrasing in the opposite direction might help. Just a guess on my part . . .

I've tried funnelling it through QuillBot, a paraphraser that can make you sound more academic, casual, short, elongated, etc, but it just makes the AI detector go berserk! I've primarily tried re-writing the dodgy paragraphs, to no avail...

 

[BrianL]

Let me ask you, Neapolitan, are you dreaming about unicorns?

 

[Neapolitan Ice Cream]

Let me ask you, Neapolitan, are you dreaming about unicorns?

THEY'RE SO FLUFFY I COULD DIE!!!

 

[BrianL]

THEY'RE SO FLUFFY I COULD DIE!!!

 

[fla4fun]

I've tried funnelling it through QuillBot, a paraphraser that can make you sound more academic, casual, short, elongated, etc, but it just makes the AI detector go berserk! I've primarily tried re-writing the dodgy paragraphs, to no avail...

I wouldn’t use additional software at this point, because the wording the software is coming up with may be just the type of wording the AI detector is looking for. Go back and rewrite a section in your own words as it was before you started the editing process (or if you still have a copy of your original draft, start there) and run the detector to see if it makes a difference. My guess is that the paper has been run through so many software programs to check it that it has taken a lot of “you” out of the paper and that is what the detector is picking up on. Good luck!

 

[Neapolitan Ice Cream]

I wouldn’t use additional software at this point, because the wording the software is coming up with may be just the type of wording the AI detector is looking for. Go back and rewrite a section in your own words as it was before you started the editing process (or if you still have a copy of your original draft, start there) and run the detector to see if it makes a difference. My guess is that the paper has been run through so many software programs to check it that it has taken a lot of “you” out of the paper and that is what the detector is picking up on. Good luck!

Thanks, @fla4fun! Makes a lot of sense!

 

[georgina]

It is a known problem with neurodivergent writers. AI thinks they sound too much like a bot. I have seen several reports of this recently.

https://medium.com/bouncin-and-beha...ua Williams,autistic people write and speak.”

 

[Neapolitan Ice Cream]

Still grappling with this. One short answer came up with 100% AI written! Do I sound so like a robot?! My course material says that sourcing data from the internet is fine, but this takes the Jerry.

 

[Uk145081]

AI detectors seem to rely a lot on syntax patterns and sentence structures, so even if you write something entirely on your own, it can still get flagged. I’ve noticed this happening when I try to be concise or formal in writing, somehow, that “AI tone” sneaks in even when it's all my own words.