RE: Why we shouldn't plagiarize with AI! 👟👟👟
You are viewing a single comment's thread:
Yes, indeed it is tricky. ChatGPT 3.5 underlines the lines that were paraphrased according to its matching algorithm. Also, we don't flag anything unless it is way more than 50% paraphrased according to the openai detector. It's a fine line we navigate, however, the benefit of the doubt will be given in all the borderline cases. We're only going after the 80-90% matches for the most part.
0
0
0.000
I understand, so anyway many could circumvent the mechanism by paraphrasing at least 50 per cent of the published article, in which case the benefit of the doubt is rightly given.
The thing I wonder is if a person is a repeat offender? That is, if it happens once for a single article, you turn a blind eye. But if a person publishes 50% paraphrased content every day, you take action right? In that case you can't have the benefit of the doubt all the time :)
Exactly, we take it on a case-by-case basis. We do examine older writings look at their history and compare it to new ones. Depending if there is a drastic change or not can play into the equation too. Once someone is warned then the next 50% is worse, and a string of them would mean they have not learned a lesson. Anyway, two 50%s add up to 100%, so we won't turn a blind eye to someone gaming the system like that. We always follow up after a first flag or a warning. I think half AI is just as bad personally.