66

At the moment AI content policy specifically mentions answers https://stackoverflow.com/help/ai-policy as content that is not allowed to be AI generated.

However, answers are not the only content that will be used as a part of the knowledgebase where having AI generated content is absolutely not admissible for all the reasons AI generated answers are not allowed and AI policy should explicitly list all such content.

Using AI for writing Articles should also be prohibited and explicitly listed in the AI content policy.

I am leaving whether we need to add other kinds of content to the policy, like Discussions, open for discussion.

16
  • 15
    It shouldn't just extent to articles, or should be any content. Questions, answers, comments, articles, anything.
    – Thom A
    Commented Oct 22, 2023 at 11:01
  • 4
    @ThomA I am leaving that open for discussion, because questions are a bit separate category from the rest. You don't expect that question will contain accurate content. Also I cannot exactly say that I am for banning AI in questions, but at the same time I am not really against it.
    – Dalija Prasnikar Mod
    Commented Oct 22, 2023 at 12:02
  • 18
    The "Temporary policy: Generative AI (e.g., ChatGPT) is banned" already applies to Articles. The ban very explicitly applies to all user-generated content on Stack Overflow, "except each user's profile content (e.g., your "About me" text)."
    – Makyen Mod
    Commented Oct 22, 2023 at 15:59
  • 4
    @Makyen True, but the policy text in help center only explicitly mentions answers stackoverflow.com/help/gpt-policy
    – Dalija Prasnikar Mod
    Commented Oct 22, 2023 at 16:19
  • 8
    @Makyen The additional problem is that Articles need to be approved and apparently nobody told staff that it might be prudent to look for AI heuristics before approving them.
    – Dalija Prasnikar Mod
    Commented Oct 22, 2023 at 16:27
  • 5
    Notwithstanding whether the current policy bans it (Makyen covered that), I don't really see that much of a problem with AI content in questions. Factual accuracy is far less important in questions, which tend to have misconceptions simply due to the nature of asking about something you don't know. I've seen AI questions used to aid in the violation of other rules (mostly spam, a little bit of sockpuppetry), but overall they're not noticeably worse than human-written questions. The main issue is accurately copying code into the question (which humans also fail at surprisingly often...).
    – Ryan M Mod
    Commented Oct 22, 2023 at 20:04
  • 4
    beautiful that the first two articles in the new Mobile Development collective 1 2 are AI-generated and got approved by SE staff.
    – starball
    Commented Oct 23, 2023 at 0:29
  • 3
    I didn't see it initially (IDK if it appears later), but it might be worth requesting the addition of a "No LLM content" notice when submitting an article (similar to the one for answers)
    – cocomac
    Commented Oct 23, 2023 at 2:37
  • Slate Commented "I like that post, though. With a little refinement (or if folks think it's ready to go), flag a mod to slap [status-review] on it, and if they think it's good to go as well, it'll enter the CM workflow." (I have flagged this post to have [[tag:[status-review]] added.)
    – Thom A
    Commented Oct 24, 2023 at 15:58
  • See also Can a "No LLM Usage" banner be added for the Collectives Article box? by me
    – cocomac
    Commented Oct 24, 2023 at 16:29
  • 1
    As noted in comments here and in the answer, the policy applies to all user-generated content. Since there is no change, I am marking thisstatus-bydesign
    – Berthold StaffMod
    Commented Nov 28, 2023 at 22:39
  • 3
    @Berthold stackoverflow.com/help/ai-policy is written about answers. If you want to resolve it like that, then the Help Center should be updated to reflect that.
    – starball
    Commented Nov 29, 2023 at 2:55
  • 1
    @starball This is the policy and it is further affirmed by a previous company answer. The Help Center provides guidance about the most common scenario of AI usage in answering
    – Berthold StaffMod
    Commented Nov 29, 2023 at 18:28
  • 2
    @Berthold The help center is the most "official" way to inform the users about the policy. Updating help center policy is not for "us" Meta users who know about AI policy and this request is not meant as clarification of some sort whether policy applies on other content or not. It is meant for newcomers, where explicitly stating only single type of content (answers) may confuse them to think that using AI in other parts is acceptable. Another confusion may stem from the fact that Meta post also includes words "Temporary" which may also give wrong impression on the nature of the ban.
    – Dalija Prasnikar Mod
    Commented Nov 29, 2023 at 19:18
  • 1
    @Berthold Possible solution for the Meta post would be replacing "Temporary" with "Current".
    – Dalija Prasnikar Mod
    Commented Nov 29, 2023 at 19:19

1 Answer 1

Reset to default
26

Tl;dr all AI-generated content is banned under the Temporary Policy and uncited AI-generated content is a CoC violation.

As Makyen said (in a comment), it's also banned under the Temporary Policy under

NOTE: [...] the ban applies to all content on Stack Overflow, except each user's profile content (e.g., your "About me" text).

Additionally, it's also a CoC violation.

Assuming that Collectives Articles need to follow the SE Code of Conduct, uncited AI-content is already prohibited.

According to Is uncited LLM usage considered a CoC violation under the Inauthentic Usage policy? (on MSE):

TL;DR: Yes, the current Inauthentic Usage policy prohibits posting AI-generated content without appropriate attribution.

Presumably, Articles must follow the CoC. I'd be quite suprised if SE staff would approve an article consisting of, say, attacks on another user, even if it somehow related to the Collective.

As a result, uncited LLM usage is a CoC violation, and can be flagged. I don't know what the state of flags is on Articles right now, but if it's possible, it'd be a good idea to state that you think it's LLM-generated. Also, if you can, including why would be helpful too (i.e., it contains "This article is likely LLM-generated because it contains 'As an AI language model ...'). This doesn't cover cited AI-content, but I'd expect that this would be mostly sufficient, as if it's cited clearly, it's clear to readers (and reviewers) that it's LLM-generated.

1
  • 2
    Yes articles can be flagged for mod attention.
    – Dalija Prasnikar Mod
    Commented Oct 22, 2023 at 19:37

You must log in to answer this question.

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