AI & Plagiarism Detectors for Editors?

[SatansFavouritePet]

Hey fellow editors,


So . . . Who's using AI & plagiarism-checking software? What do you use?

Unfortunately, I have a need to use it. But so far, I've found wildly different results when I check a single work using various engines--everything from 0% to 95% AI-generated/modified! I suspect models like GPTZero and Originality.ai are better-trained than others (their %AI results seem to fall in the high middle-ground of the extremes), but who knows? Anyone?

If you want to pm me about this, that's A-okay.


-SFP

 

[AnnieLit]

Either you like the text and want to work with the author on it, or you don't. Why does it matter who and how wrote it?

Plagiarism is a different story, but there google is usually enough. Just put a full sentence from a few random places into the search and see what comes up.

 

[jsmiam]

This brings a side topic to the foreground. Personally, I would want to know up front if an editor is using AI. I think it’s something that should be disclosed.

Just as an editor can feel justified in wondering if their author really wrote something, an author should feel justified in wondering what methods an editor is using.

In spite of how helpless we are when AI bros take anything posted on the internet without permission, it’s quite different for someone to put someone else’s writing into an AI tool.

Definitely worth disclosing.

 

[SatansFavouritePet]

This brings a side topic to the foreground. Personally, I would want to know up front if an editor is using AI. I think it’s something that should be disclosed.

Just as an editor can feel justified in wondering if their author really wrote something, an author should feel justified in wondering what methods an editor is using.

In spite of how helpless we are when AI bros take anything posted on the internet without permission, it’s quite different for someone to put someone else’s writing into an AI tool.

Definitely worth disclosing.

Indeed, and I do disclose it when I've felt a need to use such tools.

This began as my attempt to see what an AI detector reported for stories that have been rejected for suspected AI-generated content. I was hoping it might inform me as to what might have led to their rejection. It's hasn't been very useful so far.

It's giving me ideas for stories, though.

 

[SmilingLez]

AI detection tools suck. Here is a passage from the Gettysburg Address tagged as 100% AI generated. Did a time traveler use AI to write Lincoln speech? Me thinks not

 

[StillStunned]

Did a time traveler use AI to write Lincoln speech? Me thinks not

Did the plagiarism machine train on the speech? Highly likely.

So it's going to recognise those patterns as something it would produce, because those exact patterns are already in its memory.

 

[AwkwardMD]

None of which applies to Literotica

 

[SatansFavouritePet]

AI detection tools suck. Here is a passage from the Gettysburg Address tagged as 100% AI generated. Did a time traveler use AI to write Lincoln speech? Me thinks not

I knew it . . . Lincoln was an android!

 

[SmilingLez]

None of which applies to Literotica

Actually it proves that AI detection tools are flawed and I'm sure at least 10% of AI rejected stories were actually written by humans and falsely rejected by literotica. Or do you believe the fairy tale that a human being actually reads each and every story?

 

[AwkwardMD]

Actually it proves that AI detection tools are flawed and I'm sure at least 10% of AI rejected stories were actually written by humans and falsely rejected by literotica. Or do you believe the fairy tale that a human being actually reads each and every story?

It's like I haven’t spent the last year being very meticulous and consistent about all of this.

Yes there are false positives. Is it ten percent of all rejections? Maybe. Even then, false positives are identified on evidence of something even if it's not explicitly LLM-generated text.

Of course the process has been automated. No, Nobody reads first time submissions. Never in the last 12 years was that the case. Manual reviews only ever happen on a resubmission with a note.

 

[SmilingLez]

Manual reviews only ever happen on a resubmission with a note.

And I have got some beachfront property in Kentucky for sale. Nothing is reviewed manually.

 

[AwkwardMD]

Okay

 

[PennyThompson]

If you have access to someone's document for editing purposes, why not just look at their version history or track changes? It should be fairly obvious if someone is writing like a human with all the changes and deletions and rearranging, or if they're just pasting in huge chunks of mechanically pristine text?

 

[SatansFavouritePet]

If you have access to someone's document for editing purposes, why not just look at their version history or track changes? It should be fairly obvious if someone is writing like a human with all the changes and deletions and rearranging, or if they're just pasting in huge chunks of mechanically pristine text?

I think that would only be possible if they share it online through some Microsoft cloud thingamajigger. Usually authors send me attached docs, pdfs, and cut-and-pasted txt, which carry no history. Also, if they send me a doc, I put it straight to Google Docs, so I don't have to open it on my computer.

 

[PennyThompson]

I think that would only be possible if they share it online through some Microsoft cloud thingamajigger. Usually authors send me attached docs, pdfs, and cut-and-pasted txt, which carry no history. Also, if they send me a doc, I put it straight to Google Docs, so I don't have to open it on my computer.

You should be able to do Track Changes on an offline desktop version of Word too, though the original writer might need to turn it on in their document.

If it were me and I were editing or beta reading for someone I didn't already know and trust, I would probably make it a non-negotiable workflow requirement.

I feel strongly that "proof of work," is going to be an increasingly important part of any creative artifact, in the future. People will want and demand evidence that the thing they're enjoying and interacting with was actually made by human time and effort.

 

[SatansFavouritePet]

You should be able to do Track Changes on an offline desktop version of Word too, though the original writer might need to turn it on in their document.

If it were me and I were editing or beta reading for someone I didn't already know and trust, I would probably make it a non-negotiable workflow requirement.

I feel strongly that "proof of work," is going to be an increasingly important part of any creative artifact, in the future. People will want and demand evidence that the thing they're enjoying and interacting with was actually made by human time and effort.

Yeah, I tried it, but the option isn't even available.

 

[millamomud]

I feel strongly that "proof of work," is going to be an increasingly important part of any creative artifact, in the future. People will want and demand evidence that the thing they're enjoying and interacting with was actually made by human time and effort.

It would be nice to be able to actually prove authorship like that on here!

 

[SkyBubble]

This began as my attempt to see what an AI detector reported for stories that have been rejected for suspected AI-generated content.

An equally interesting experiment might be to put stories written and published before AI writing became a "thing" and see what percentage the detectors come up with.

 

[SatansFavouritePet]

An equally interesting experiment might be to put stories written and published before AI writing became a "thing" and see what percentage the detectors come up with.

Check out @SmilingLez 's reply above re: Abe Lincoln.

Also . . . I've put small passages of my own writing through a number of AI-checkers . . . and all came back as 100% human! Thank science, 'cause I was seriously worried that I might be a simulation. . . . Right?!? Anyone else?

 

[AnnieLit]

Actually it proves that AI detection tools are flawed and I'm sure at least 10% of AI rejected stories were actually written by humans and falsely rejected by literotica. Or do you believe the fairy tale that a human being actually reads each and every story?

You do know that this site existed ling before AI came around, right? At that time everything was really read, or at least skimmed, by the same person. I would guess that with AI the site was flooded with generated submissions, it became impossible to read all of it, so Lauren started to use automation to weed out AI submissions. What's wrong with that? If your story gets wrongly tagged as AI, resubmit and attach a note explaining how AI was used to assist in writing. That is if it was used at all.

 

[AnnieLit]

Check out @SmilingLez 's reply above re: Abe Lincoln.

Also . . . I've put small passages of my own writing through a number of AI-checkers . . . and all came back as 100% human! Thank science, 'cause I was seriously worried that I might be a simulation. . . . Right?!? Anyone else?

I use AI assistance from time to time. Not to write start to finish, but to format the text, especially dialogs, and to fill out gaps when I need the story to get from point A to point B. Sometimes I like its ideas, sometimes I don't.

I ran a few of these hybrid texts through different AI detection sites, usually I would select three paragraphs where two are written by me and one is 100% AI. Every time I do this, all sites say that it is 100% human text, even though I know for sure it is 30% AI.

PS: none of it was submitted to Lit, so I don't know what their automation would say about these texts.

 

[SmilingLez]

You do know that this site existed ling before AI came around, right? At that time everything was really read, or at least skimmed, by the same person. I would guess that with AI the site was flooded with generated submissions, it became impossible to read all of it, so Lauren started to use automation to weed out AI submissions. What's wrong with that? If your story gets wrongly tagged as AI, resubmit and attach a note explaining how AI was used to assist in writing. That is if it was used at all.

Then how do you explain the fact that there are many writers who have written dozens and dozens of (posted) stories and just within the last year their stories get sent back with AI rejection? Don't believe me, search all of the AI rejection threads and see for yourself.

Quite naive to believe that 1 person reads anything but maybe a page or two (and I'm being generous) of every single story submitted. Imagine what that would do to the human psyche. Following your train of thought, why isn't this 1 person who reads everything not dinging the poorly constructed incoherent messy stories?

One last thing, where did you get this inside information as to how literotica operates? Did you pull it out of thin air?

 

[StillStunned]

Quite naive to believe that 1 person reads anything but maybe a page or two (and I'm being generous) of every single story submitted. Imagine what that would do to the human psyche. Following your train of thought, why isn't this 1 person who reads everything not dinging the poorly constructed incoherent messy stories?

Quite likely that Laurel (may she live forever) likes a low barrier to publication for amateur writers and doesn't feel that messy stories don't pose the same ethical or legal qualms as AI-assisted writing does.

One last thing, where did you get this inside information as to how literotica operates? Did you pull it out of thin air?

It's commonly accepted knowledge, from back when Laurel (may she live forever) posted regularly on the forums. There's no reason to suppose her methods have changed except to weed out AI.

 

[AnnieLit]

Then how do you explain the fact that there are many writers who have written dozens and dozens of (posted) stories and just within the last year their stories get sent back with AI rejection? Don't believe me, search all of the AI rejection threads and see for yourself.

Quite naive to believe that 1 person reads anything but maybe a page or two (and I'm being generous) of every single story submitted. Imagine what that would do to the human psyche. Following your train of thought, why isn't this 1 person who reads everything not dinging the poorly constructed incoherent messy stories?

One last thing, where did you get this inside information as to how literotica operates? Did you pull it out of thin air?

As I said, they DID add AI filter to slow down inflex of generated texts. And every now and then those filters will produce false positives, no way around it. So I am not surprised that SOME human written stories get flagged.

As for why some badly written stories are allowed... Lauren never said that she is judging the quality of writing. Her only goal is, and always was, not to allow stuff that shouldn't be on the site for legal reasons - underage, celebrities fanfic, etc.

Where did I get the inside info ... I have been around long enough to read actual conversation with the owner. Plus I know that when I submitted my stories (before AI time) I included some questions to the admins in there and they always got answered.

 

[SmilingLez]

As I said, they DID add AI filter to slow down inflex of generated texts. And every now and then those filters will produce false positives, no way around it. So I am not surprised that SOME human written stories get flagged.

As for why some badly written stories are allowed... Lauren never said that she is judging the quality of writing. Her only goal is, and always was, not to allow stuff that shouldn't be on the site for legal reasons - underage, celebrities fanfic, etc.

Where did I get the inside info ... I have been around long enough to read actual conversation with the owner. Plus I know that when I submitted my stories (before AI time) I included some questions to the admins in there and they always got answered.

Thank you for admitting that the AI tools are flawed and non-generated AI stories get rejected (though it is more frequent than "every now and then"). I disagree with you when it comes to poorly constructed and/or error riddled stories. More than a few errors and these stories should get rejected with an appropriate message on how to correct said errors.