278

In a study recently published in the journal Patterns, researchers demonstrate that computer algorithms often used to identify AI-generated text frequently falsely label articles written by non-native language speakers as being created by artificial intelligence. The researchers warn that the unreliable performance of these AI text-detection programs could adversely affect many individuals, including students and job applicants.

top 26 comments
sorted by: hot top controversial new old
[-] cloudless@feddit.uk 37 points 1 year ago

Most AI-generated texts are grammatically perfect. That's not a characteristic of non-native speakers.

[-] AsimovsRobot@lemmy.world 26 points 1 year ago

A lot of non-native speakers can show higher command of the language, because they took the time to study its rules. Just look at how people type on social media.

[-] cloudless@feddit.uk 4 points 1 year ago

I must be one of the lazy ones who didn't take enough time to study English grammar. (͡•_ ͡• )

Yeah I get your point, many non-natives pay more attention to grammar when they write.

[-] hardware26@discuss.tchncs.de 25 points 1 year ago

According to the article grammatical errors are not the reason. The reason is that AI uses simpler vocabulary to mimic a regular conversation of average people.

[-] Absurdist@lemm.ee 23 points 1 year ago

Completely disagree - a lot of non-native speakers have excellent grasp of grammar, precisely because they have learnt the rules. Native speakers rely on stuff sounding right, rather than necessarily knowing the rules. But following grammatical rules rigidly is exactly what I would expect both from a genAI and a non-native speaker (as well as avoiding figurative speech and idioms).

[-] cloudless@feddit.uk 5 points 1 year ago

Sorry I might have overly generalised based on my personal experience. I have been a non-native English speaker for over 30 years, and I keep making grammatical mistakes.

Everyone is different and it depends heavily on how the person learned/acquired the language.

[-] Diabolo96@lemmy.dbzer0.com 10 points 1 year ago

As a non native am keenly aware of every mistake i make. Sometimes, i fix the same text 4 or five times in row.

[-] wahming 7 points 1 year ago

I count 4 mistakes. Conclusion: AI generated

[-] Diabolo96@lemmy.dbzer0.com 4 points 1 year ago

Where ? You got me anxious about my writing skill dude...

[-] DeadlineX@lemm.ee 7 points 1 year ago* (last edited 1 year ago)

Not OC but you forgot the “I” before “am”. I should be capitalized twice. It should say in a row. It’s also a bit weird that you used 4 and then five. Plus it really should be non-native I guess. And since you didn’t specify what you aren’t native to/of, it can only be assumed at this point that you are not a native human. The reason for that? Definitely a bot.

Also honestly don’t worry about any minor errors in your writing. Especially on the internet. Just blame autocorrect.

[-] Diabolo96@lemmy.dbzer0.com 2 points 1 year ago

I didn't specify what I'am not native to because it can be understood from the context. Capitalizing the 'I' feels weird and I juste saw the "in a row" mistake.

[-] DeadlineX@lemm.ee 6 points 1 year ago

In English, I is always capitalized when referring to the self. The “not specifying” comment was me making a joke and calling you a bot. It was just tying the context of the post into my comment.

I usually see mistakes in comments as being caused by crappy phone keyboards, as mine seems to be extra trash recently.

[-] Diabolo96@lemmy.dbzer0.com 1 points 1 year ago

Yep, phone keyboards do be really crappy.

[-] wahming 4 points 1 year ago* (last edited 1 year ago)

Lol. No big deal. It's the sort of thing your eyes skip over when checking your grammar.

As a non native am keenly aware of every mistake i make. Sometimes, i fix the same text 4 or five times in row.

Should be

As a non native I'm keenly aware of every mistake I make. Sometimes, I fix the same text 4 or five times in a row.

Two of the mistakes were just capitalization of 'I'. You should probably also use either digits or spelling for both numbers, not a mix.

[-] Diabolo96@lemmy.dbzer0.com 2 points 1 year ago

Thanks for the correction.

[-] Kolanaki@yiffit.net 4 points 1 year ago* (last edited 1 year ago)

As a non native am keenly aware of every mistake i make. Sometimes, i fix the same text 4 or five times in row.

As a non-native, I am keenly aware of every mistake I make. Sometimes, I fix the same text 4 or five times in row.

FTFY.

[-] TimewornTraveler@lemm.ee 3 points 1 year ago

hey just to give some validation, I'm an esl teacher and this doesn't stick out as non-native at all. They're all just taking the piss, correcting anything they can find for the joke of it.

maybe saying "am" instead of "I am" but that's kinda just meme speech right

[-] Diabolo96@lemmy.dbzer0.com 1 points 1 year ago

That's really nice of you. Thanks !

[-] cloudless@feddit.uk 1 points 1 year ago

I have been using ChatGPT to proofread most of my formal written communications.

[-] tal@kbin.social 2 points 1 year ago

"This guy is definitely AI."

[-] FunderPants@lemmy.ca 2 points 1 year ago* (last edited 1 year ago)

I'm a college instructor and encourage my students to use ChatGPT this way.

[-] dethb0y@lemmy.world 20 points 1 year ago

That's actually very interesting, i would not have expected that.

[-] FlyingSquid@lemmy.world 20 points 1 year ago

I feel like I should have expected that. The fact that they write in a non-standard way would be a red flag to detection software.

[-] merde@sh.itjust.works 6 points 1 year ago

in non-standard ways, we multilinguals do not write

may the farce be with you

[-] luthis@lemmy.nz 5 points 1 year ago

It actually makes complete sense when this interesting point has attention drawn to it.

[-] db2@sopuli.xyz 5 points 1 year ago

Kindly do the needful. Thank.

this post was submitted on 04 Sep 2023
278 points (98.6% liked)

Technology

58073 readers
3106 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related content.
  3. Be excellent to each another!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, to ask if your bot can be added please contact us.
  9. Check for duplicates before posting, duplicates may be removed

Approved Bots


founded 1 year ago
MODERATORS