-10

I ask this question in continuation of this question.

Lately I've been trying to figure out exactly what makes a good answer on Stack Overflow.

If I write carefully, using proper punctuation, grammar, and structure, my answer will sometimes be flagged or commented on as one generated by large language models, but if I write more carelessly or skip some corrections, my post will be at risk of being downvoted due to poor writing quality.

I sometimes use Google Translate, which automatically adds punctuation marks or even uses language models behind the scenes, but I write the text myself and it's a translation tool, I thought the problem with using language models is in the content they produce, not in the way they write the content.

I thought that using language models was forbidden because when someone asks a question on this site, they probably have already raised their problem with language models and not received a good answer and it may be unpleasant to keep getting the same answer on this site.

I think that the way a text is written should not be prohibited, even if it is generated by language models, but the content of that text is important. I have this theory that these days, artificial intelligence and language models are developing rapidly and can be felt in almost many tools. And the tools are needed by people and experts.

I used Google Translate to write this text, so don't think it was generated by a language model. But maybe Google Translate uses a language model behind the scenes. I'm not sure.

20
  • 17
    The problem comes from when you use a LLM based tool to do a translation or grammar correction that does it in the form of more or less rewriting/rephrasing what you said, it is often indistinguishable from completely generated content. Many of the same sentence structures and formatting LLM's are known for using will be added to your content. Commented Oct 21 at 20:30
  • 1
    You can post with bad grammar, and wait for someone else to fix it. I would edit your posts, but for translation, there are other languages supported by SO. Commented Oct 21 at 20:40
  • 9
    Yes, i understand the frustration and how ridiculous things have become... but that's the world we're in now. If it can be thrown at an LLM it is thrown at an LLM, regardless of the result. It's unfortunate that it is so abused that it gets in the way of legitimate uses. But what are we to do about that? Commented Oct 21 at 20:41
  • 1
    @TwineeeTheZeldaWizard Unfortunately, as far as I know, it does not support Persian, and my experience has been that English-language resources are usually better. That's why I've been learning English for about a year. Although it's a bit difficult for me, I think it's necessary. Commented Oct 21 at 20:48
  • 6
    English language quality is mostly a "red herring" here, meaning it's distracting you and us from something more important. When you were told your other post looked AI generated, that post would look a little similar to AI generated answers even if it were in Persian. Commented Oct 21 at 21:16
  • 5
    You cannot separate content and style as good as you think. If you use an LLM it will not be your content, if you continue learning English, you won't need an LLM. I personally don't mind if you write something, then improve the style of it, but don't fool yourself, using LLMs very likely means that you do not know the content, that the content of the answer isn't from you. Maybe at some point in the future, should LLMs be so ubiquitous that you cannot be without them, maybe then the rules here will change. So far they haven't. Sorry, you have to write every single word by yourself. Commented Oct 21 at 21:18
  • 3
    Why not learn English first until you get a "decent" command of the language...? Programming requires mastering English up to some certain level (all the documentation is in English, any Tech Site (Forum, or 'Stack Overflow') is in English, as soon as you start working, you'll have non-local colleagues and English will be the language of communication)... And you will never become "senior level" in any digital area if you don't master English "a bit decently", spoken + written... (Everything is a question of motivation...!) Commented Oct 21 at 21:35
  • 3
    @chivracq Python tutorial in Spanish: docs.python.org/es/3/tutorial. It's also available in Greek, French, ... Commented Oct 21 at 22:20
  • 3
    @mehdi_ahmadi - If your content is being flagged as AI generated then it’s not because your using tools designed to fix grammatical errors. I run every single one of my 1,500 answers through Grammarly and I have never had an answer flagged as AI generated. If you are using LLM to rewrite what you write that’s not allowed. If you’re using it to detect grammatical errors that’s fine. LLM rewriting your content is a good explanation for what you’re describing. Stop doing that if that’s the case. Commented Oct 21 at 22:28
  • 3
    @Wicket Yeah, a tutorial will get you "started", sure, you can also follow some physical course in your nearby city (that will be in that language), but you still need to be able to read the documentation (in English), then any other examples/tutorials of some a bit advanced level will be in English, Tech Sites will still be in English, etc... Commented Oct 21 at 23:07
  • 1
    Please research reasonably before considering posting a question. This post doesn't reflect the main points of Policy: Generative AI (e.g., ChatGPT) is banned. Commented Oct 21 at 23:45
  • 2
    @SecurityHound That's the point. I don't use it for posts on this site at all, but some users think so and because of the many downvotes they give, my account is being restricted and this is annoying to me. Commented Oct 22 at 4:07
  • 4
    @mehdi_ahmadi - Downvoted answers contribute to answer bans more than question bans. Most users complain about question bans more than answer bans hence the reason I assumed a question ban was applicable. Unfortunately if you are currently answer banned, there is no rate limit, and you are stuck trying to improve past contributions (answers) to get enough upvotes to not be answer banned. If you don’t mean your under an answer ban, then explaining what restrictions you actually are under, would probably help. Commented Oct 22 at 4:59
  • 2
    Like I always say whenever something like this comes up: SO is a programming site. There is just no way you can function as a programmer without knowing basic English. Period. So this is not even an issue, since basic English is a prerequisite before you can begin to learn programming. Nobody expects fluent English though. Spelling and grammar can have mistakes, those are just minor problems that will get fixed by editing. Commented Oct 22 at 6:37
  • 11
    @mehdi_ahmadi "I don't think you can introduce a professional translator who doesn't use language models" On the contrary, I don't think I would consider a translator a professional if they couldn't translate without using the internet, but especially not if they decided to use an LLM-based tool. That is definitely amateur hour. A professional translator should be fluent in the languages they are translating between. Commented Oct 22 at 13:12

1 Answer 1

Reset to default
11

If I write carefully, using proper punctuation, grammar, and structure, my answer will sometimes be flagged or commented on as one generated by large language models, but if I write more carelessly or skip some corrections, my post will be at risk of being downvoted due to poor writing quality.

Wrong - the only thing we ultimately care about in an answer is its correctness. And in your previous Meta question it was pointed out that your answer was fundamentally incorrect, hence the downvotes. So ultimately, whether or not you used an LLM to generate or assist with that answer is irrelevant because the answer itself is wrong - that wrongness is why it was downvoted.

Now, the fact that the answer is wrong is bad enough, but it also appears to have been created using an LLM in some shape or form, and we have a strong stance against the use of genAI on Stack Overflow. Note that this policy does not preclude using an LLM to format or otherwise improve your answer, e.g. by fixing up poor grammar or spelling - but when your answer is already wrong, and it looks like you used an LLM on top of that, people are going to be more harsh than they would be normally - including possibly assuming that you did just use an LLM to generate the whole thing.

(I'm not going to explain how or why I believe your answer to have been LLM-assisted, as that might help others to post answers that violate the above policy.)

In short, stop worrying about spelling and/or grammar and worry about correctness because if your answer is right, the rest doesn't really matter. We're software developers, code is the primary language by which we communicate, and in my own personal experience of over 20+ years in this industry the best way to overcome language barriers and get everyone on the same page is a simple "show me the code".

On Stack Overflow itself, I've come across some questions and answers with truly abysmal, tending to almost unintelligible, spelling and grammar. But the code in those posts has been sufficient to allow me to understand what's going on and whether they're correct or not, and I've then used my editing privilege and my own command of English to improve the original writer's words. So even if your own spelling and grammar aren't up to scratch, someone else will probably help with that anyway.

Get the fundamentals correct, and the rest will follow.

2
  • 1
    Agreed, but yet Stack Overflow is pretty adamant about targeting the English language. So you are expected to have at least a base level of English, and answers must be more than code. If texts are translated wholesale, I can't really imagine that not being flagged as an AI-generated post. It'd be asking too much of people to try and see a difference there. People will pattern match, "does this look AI generated"? Yes/no. "Yes, but" won't be a consideration. Commented Oct 23 at 13:38
  • 3
    @Gimby - Translation of human written text in one language to English isn’t what’s being caught or excused of being AI generated, AI generated content, or perhaps AI enhanced content is being caught. Commented Oct 24 at 6:58

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.