Tortured phrases: A dubious writing style emerging in science • Hacker News Board

Re: Tortured phrases: A dubious writing style emerging in science

#41

Post by WesolyKubeczek » Sun, Aug 08, 2021, 6:13 PM UTC

eecc wrote: post #20

WesolyKubeczek wrote: post #15

Earlier quoted context omitted.

I’m wondering what happened to good old reading with comprehension. Ain’t nobody got no time for that? If so, doesn’t it make those papers worthless?

Nope. Time is the only non-fungible asset being burned here and everyone is desperately defending their own allotment.

Then can we at least draw a border around such "science" so that serious people who have work to do know to not waste time with it?

Re: Tortured phrases: A dubious writing style emerging in science

#42

Post by PragmaticPulp » Sun, Aug 08, 2021, 6:13 PM UTC

tarboreus wrote: post #12

doubtfuluser wrote: post #7

Maybe a future direction would be to train new models to identify plagiarism by training on this information. Use „non matching backtranslations for training classifiers. It’s again the typical cat and mouse game I guess

Or someone could...read the papers.

The number of papers being published is growing at a staggering rate. This requires proportional growth in the number of people reading these papers, which inevitably means the plagiarists and cheaters themselves are being pulled into the review system as well. They don’t care about letting fraudulent papers slip through because they never really cared about the science in the first place.

They see it as a game that they’re playing and they’re doing their best to put as little effort as possible into the game while extracting as much reputation upside as they can.

We really need to make publishing fraudulent papers a career-ending move across academia and even the industry. The only reason this continues to happen is because it has a lot of upside but very little downside. Caught publishing fraudulent papers? Oh well, just leave them off your resume and apply somewhere else.

Re: Tortured phrases: A dubious writing style emerging in science

#43

Post by itronitron » Sun, Aug 08, 2021, 6:13 PM UTC

I like that "citation of non-existent literature" is also a feature of these articles, although I wonder if the non-existent literature was previously cited in other papers. (https://irthoughts.wordpress.com/2009/07/15/the-most-influen...)

Re: Tortured phrases: A dubious writing style emerging in science

#44

Post by phkahler » Sun, Aug 08, 2021, 6:14 PM UTC

gzer0 wrote: post #35

This is anecdotal evidence at best, but it is worth considering. I know of several individuals who were able to complete their entire Master's thesis utilizing a combination of AI generated content (GPT-3) and a paraphrasing tool. The generated text was well over 50 pages, completely bypassed all known content/plagiarism checks and was even included in the Universities "exemplary examples". To this day, it is still t…

Sounds like automated review of automatically generated papers. And people pay money for that...

Re: Tortured phrases: A dubious writing style emerging in science

#45

Post by cperciva » Sun, Aug 08, 2021, 6:18 PM UTC

cpach wrote: post #40

Pixelbrick wrote: post #14

This will come as no surprise to FE/HE lecturers the world over...

What is FE/HE?

I'm guessing Further Education / Higher Education.

Re: Tortured phrases: A dubious writing style emerging in science

#46

Post by mrfusion » Sun, Aug 08, 2021, 6:18 PM UTC

I wonder if there are any phrases to detect industry funded papers? I know they tell you their funding sources but it doesn’t seem to always help.

Re: Tortured phrases: A dubious writing style emerging in science

#47

Post by riedel » Sun, Aug 08, 2021, 6:18 PM UTC

jokoon wrote: post #31

What's the point of this? To waste the time of foreign scientists? Would we call this science warfare?

It is the result of a misguided science system that relies mostly on external quality checks (peer reviewed publication) and flooding the world with so much "novelty" that there is no way to digest it. At least you can use the output to train language models up to now: will machines now have to train themselves...

Re: Tortured phrases: A dubious writing style emerging in science

#48

Post by varjag » Sun, Aug 08, 2021, 6:19 PM UTC

First thought after the opening paragraph, "these have to be mainlanders". Scrolling down, yup.

Re: Tortured phrases: A dubious writing style emerging in science

#49

Post by mrfusion » Sun, Aug 08, 2021, 6:20 PM UTC

I wonder if any phrases or styles could detect group think or studies following the crowd.

Re: Tortured phrases: A dubious writing style emerging in science

#50

Post by 13415 » Sun, Aug 08, 2021, 6:20 PM UTC

gzer0 wrote: post #35

This is anecdotal evidence at best, but it is worth considering. I know of several individuals who were able to complete their entire Master's thesis utilizing a combination of AI generated content (GPT-3) and a paraphrasing tool. The generated text was well over 50 pages, completely bypassed all known content/plagiarism checks and was even included in the Universities "exemplary examples". To this day, it is still t…

Please include a link to these theses, because as it stands this anecdote sounds extremely implausible. I don't know what university you were, but I've been at a few in Europe and at every one of them Master theses were evaluated from the start to the end by several humans. GPT-3 is unable to produce even two pages of coherent text, let alone 50 pages good enough to be accepted as a Master thesis in any discipline at any university I could think of (even the worst ones).

I can imagine that plagiators use paraphrasing software quite extensively, though, and that it is a problem.