5 ms·

> If I teach English as a foreign language, am familiar with the kind of mistakes non native speakers make when writing in English, and can easily identify text

> If I teach English as a foreign language, am familiar with the kind of mistakes non native speakers make when writing in English, and can easily identify texts written by non native speakers - that’s not confirmation bias. And if you immersed yourself in the writings of non-native English speakers for a significant amount of time, then a paper written by a native English speaker was presented without you knowing who the author was, would you look for the same kind of mistakes in it that you have been finding?

gyomu 1y ago
I wouldn’t look for anything. As I’m reading a text, something in my brain would immediately go “this is probably written by a non native speaker” when coming across certain awkward turns of phrase.
- AdieuToLogic 1y ago
  >> And if you immersed yourself in the writings of non-native English speakers for a significant amount of time, then a paper written by a native English speaker was presented without you knowing who the author was, would you look for the same kind of mistakes in it that you have been finding? > I wouldn’t look for anything. As I’m reading a text, something in my brain would immediately go "this is probably written by a non native speaker" when coming across certain awkward turns of phrase. The problem is that the more a person is looking for a pattern, the more they are likely to believe they have found an instance of one. For example, if a police officer is regularly tasked with identifying drivers leaving bars who might be intoxicated, they will naturally use a form of "pattern recognition" to determine which vehicles to stop. Common indicators are a vehicle that is driving too slow or remains stationary after a traffic light turns green. So what happens when a sober designated driver is tired and drives overly cautious? What will the cop do if the designated driver falls asleep at a traffic light? Would the cop testify along the lines of: I wasn't looking for anything. As I observed the vehicle, something in my brain immediately said "this is probably a drunk driver" when coming across those driving patterns. This is by definition a form of confirmation bias.
anal_reactor 1y ago
Just like in speech, people also have accents in writing. For example, despite having used English on daily basis for years, I can't tell when to use "a" and "the" even if my life depended on it. Native speakers pretty much never make this mistake, but they're likely to make other mistakes, like confuse "there", "they're" and "their", which never happens to me. Not to mention the overall sentence structure even if my text is grammatically correct. I'm likely to build sentences in a way that mirrors my native language. Sure, after being exposed to tons of text in English I got rather good at it, but if you ask me to speak Spanish, I'll produce sentences that are grammatically correct, but a native would never say them because they prefer other structures to express the same ideas.
- AdieuToLogic 1y ago
  > Just like in speech, people also have accents in writing. For example, despite having used English on daily basis for years, I can't tell when to use "a" and "the" even if my life depended on it. "A" and "an" are indefinite articles[0] used to identify a single entity (noun) without specificity. "The" is a definite article[1] used to identify a single entity (noun) specifically. For example: A car is parked on the street. The car is parked on the street. The former sentence indicates a vehicle of unknown origin exists "on the street", whereas the latter indicates a vehicle known to the speaker exists "on the street." Now, to the topic at hand. Part of the original post to which I replied is: > The piece was deeply thought-provoking, but I struggled to get through it sensing how much AI was used to write it. > I’ve been drafting a manuscript for a novel lately, trying to see how well llms can help. > I recognize this prose immediately as OpenAI gpt 5. To the best of my knowledge, this was not posted by the person who wrote the article nor someone with direct interaction with same. Additionally, the quoted text self-identifies as being intimate with LLM generated content in work important to the poster as well as unequivocally identifying the article as "OpenAI gpt 5." My premise is only that having this level of intimacy with LLM generated text may create a bias toward "this was made by a LLM" when reading prose published by an unknown person. 0 - https://www.merriam-webster.com/grammar/is-it-a-or-an https://www.merriam-webster.com/grammar/is-it-a-or-an 1 - https://www.merriam-webster.com/dictionary/the https://www.merriam-webster.com/dictionary/the

gyomu 1y ago

I wouldn’t look for anything. As I’m reading a text, something in my brain would immediately go “this is probably written by a non native speaker” when coming across certain awkward turns of phrase.

AdieuToLogic 1y ago
>> And if you immersed yourself in the writings of non-native English speakers for a significant amount of time, then a paper written by a native English speaker was presented without you knowing who the author was, would you look for the same kind of mistakes in it that you have been finding? > I wouldn’t look for anything. As I’m reading a text, something in my brain would immediately go "this is probably written by a non native speaker" when coming across certain awkward turns of phrase. The problem is that the more a person is looking for a pattern, the more they are likely to believe they have found an instance of one. For example, if a police officer is regularly tasked with identifying drivers leaving bars who might be intoxicated, they will naturally use a form of "pattern recognition" to determine which vehicles to stop. Common indicators are a vehicle that is driving too slow or remains stationary after a traffic light turns green. So what happens when a sober designated driver is tired and drives overly cautious? What will the cop do if the designated driver falls asleep at a traffic light? Would the cop testify along the lines of: I wasn't looking for anything. As I observed the vehicle, something in my brain immediately said "this is probably a drunk driver" when coming across those driving patterns. This is by definition a form of confirmation bias.

anal_reactor 1y ago

Just like in speech, people also have accents in writing. For example, despite having used English on daily basis for years, I can't tell when to use "a" and "the" even if my life depended on it. Native speakers pretty much never make this mistake, but they're likely to make other mistakes, like confuse "there", "they're" and "their", which never happens to me. Not to mention the overall sentence structure even if my text is grammatically correct. I'm likely to build sentences in a way that mirrors my native language. Sure, after being exposed to tons of text in English I got rather good at it, but if you ask me to speak Spanish, I'll produce sentences that are grammatically correct, but a native would never say them because they prefer other structures to express the same ideas.

AdieuToLogic 1y ago
> Just like in speech, people also have accents in writing. For example, despite having used English on daily basis for years, I can't tell when to use "a" and "the" even if my life depended on it. "A" and "an" are indefinite articles[0] used to identify a single entity (noun) without specificity. "The" is a definite article[1] used to identify a single entity (noun) specifically. For example: A car is parked on the street. The car is parked on the street. The former sentence indicates a vehicle of unknown origin exists "on the street", whereas the latter indicates a vehicle known to the speaker exists "on the street." Now, to the topic at hand. Part of the original post to which I replied is: > The piece was deeply thought-provoking, but I struggled to get through it sensing how much AI was used to write it. > I’ve been drafting a manuscript for a novel lately, trying to see how well llms can help. > I recognize this prose immediately as OpenAI gpt 5. To the best of my knowledge, this was not posted by the person who wrote the article nor someone with direct interaction with same. Additionally, the quoted text self-identifies as being intimate with LLM generated content in work important to the poster as well as unequivocally identifying the article as "OpenAI gpt 5." My premise is only that having this level of intimacy with LLM generated text may create a bias toward "this was made by a LLM" when reading prose published by an unknown person. 0 - https://www.merriam-webster.com/grammar/is-it-a-or-an https://www.merriam-webster.com/grammar/is-it-a-or-an 1 - https://www.merriam-webster.com/dictionary/the https://www.merriam-webster.com/dictionary/the