14

I recently flagged an answer because generated by AI.

The flag was declined because they found no evidence of it: enter image description here

But there is evidence, because at the beginning of the answer it is written it is got from an AI:

enter image description here

and in the comments it is reaffirmed:

enter image description here

Which other evidence is missing?

9
  • 1
    I assume it was declined because Stackexchange failed to make a network wide ban for AI posts and leaves it to every site to decide themselves, see meta.stackexchange.com/questions/384922/… for a list of sites which already ban it. I'd be very much in favour of instating such a ban on tex.se. For previous AI posts it was easier to get them deleted as their code was just non-working gibberish and they could be deleted as not being an answer. Apr 6, 2023 at 8:09
  • In this case the code technically worked, so I think the sites needs an explicit ban for AI to deal with such cases. (it was interesting to see that the AI code was actually closer to semantic TikZ code than the one produced by mathcha.io) Apr 6, 2023 at 8:09
  • @samcarter_is_at_topanswers.xyz I thought that AI answers where banned anywhere on Stack Exchange sites. Should I change my post in "Should AI answers be banned?"
    – CarLaTeX
    Apr 6, 2023 at 8:21
  • 1
    Either that or make a new post for discussing an AI ban? Then mods could answer this post with the actual reason why they declined the flag, my comment was just an assumption which might be wrong. Anyway, I think it is important to have a clear policy how to handle AI posts. Apr 6, 2023 at 8:29
  • 1
    The original motivation of the ban on AI is because it allows people to generate a lot of low-quality hard-to-check-for-correctness answers quickly. In this case the OP manually edited the answer and checked that it draws the expected shape (although there are probably a lot of redundant code) so it shouldn't really count as AI-generated, just "written with the help of AI for research".
    – user202729
    Apr 8, 2023 at 3:45
  • 2
    @user202729 That‘s reasonable, but I don't think the OP manually edited the code, what they say in their comment is they refined the question to the AI. (I'm waiting for an official answer by moderators but still missing.)
    – CarLaTeX
    Apr 8, 2023 at 4:50
  • 1
    @user202729 the poster may have checked the output, but that's a very low bar. By their own confession they don't know TikZ well, so this seems clearly AI-generated. It may be a skill to craft a series of questions to get "good" output from an AI but that's not building on tex knowledge.
    – Dai Bowen
    Apr 8, 2023 at 13:36
  • My point is not on "building TeX knowledge", the point is that the reason we apply such a drastic measure on AI-generated answers (instant deletion as spam, -100 reputation) is because of the low-quality spam made possible by ChatGPT. In this case it doesn't need to apply.
    – user202729
    Apr 8, 2023 at 14:29
  • 1
    I asked a specific question on AI: tex.meta.stackexchange.com/questions/10049/…
    – CarLaTeX
    Apr 10, 2023 at 8:59

1 Answer 1

9

I was the one who declined the flag; here are some of the considerations I had:

  • We (TeX.SE) don't have any policy on AI-related posts.

  • The answer didn't seem off-target nor was it wrong.

  • The answer came from the OP (who mentioned through comments that it works for them).

Based on the above I made a choice from the following list of options when declining a flag:

  • Flags should not be used to indicate technical inaccuracies, or an altogether wrong answer

  • A moderator reviewed your flag, but found no evidence to support it

  • Flags should only be used to make moderators aware of content that requires their intervention

  • Using standard flags helps us prioritize problems and resolve them faster. Please familiarize yourself with the list of standard flags: see What is Flagging?

  • Other: (custom comment)

The second option seemed appropriate at the time: I reviewed it, didn't think the use of AI was a problem, nor did we have something in place to prohibit or even discourage it. So, the flag itself wasn't supported, nor the statement in it... Nothing against you.

In answer to your question of whether there is a way to reply to a declined flag: No, there's no official way. Options include chat or the child meta; you followed the correct route.


I, personally, had/have no objection with AI answers supporting the needs of the community. Based on the assumption that answers are often tested, at worst only by the OP, but at best by the rest of the community, there will be some form of communal vetting of whether or not the answer is valid... and these types of things can therefore be dealt with by the community (through, for example, flagging content as Low Quality so it ends up in a review queue where others can provide their opinion; alternatively, through voting).

While the answer in question explicitly mentioned the use of AI in the generation of some of the code, this may not be the case in general. So, identification of AI-generated content and its approval/disapproval should be handled by the community as a whole. And the best way to achieve this is through voting.

Here are some different thoughts on the process:

  • If the question was answered without mention of AI involvement (whether or not there is some policy in place), do you think you would have handled the situation differently?

  • There are often questions on this site asking to create something (an image, or a table, or some structure, often in a do-this-for-me fashion), which is then answered - literally - with "Like this?", followed by some code... could those also be coming from AI-generated prompts?

3
  • 4
    Hi Werner, I flagged it because I thought such answers were forbidden, but samcarter told me they are not banned network-wide, but it depends on each site (I've already opened another Meta-question on it). However, the reason "there is no evidence of it" is completely misleading. There is evidence the answer is generated by AI! The correct reason should have been "AI answers are allowed, at least at the moment". That's why I complained about it.
    – CarLaTeX
    Apr 10, 2023 at 7:01
  • 3
    @CarLaTeX: Sure. I responded to the flag statement - Answers generated by AI are not allowed - which I found no evidence supporting it.
    – Werner Mod
    Apr 10, 2023 at 14:55
  • 2
    Ah, that's the misunderstanding :D It's clear now
    – CarLaTeX
    Apr 10, 2023 at 15:15

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .