4
$\begingroup$

Sites can now request to enable a banner to warn about their policy on AI-generated content

This is a policy introduction on the Meta of all Metas, and given how ChemSE has an anti-AI (as in, use AI to get your posts formatted but don't generate your answers using it) stance, when are we planning to introduce this? Also it would be great if we could tie up with an AI-checker and get answers / questions checked by the AI and prevent those questions whose chances of being generated by AI are more than 50%. What does the community have to say about this? I know that the AI issue was raised long back, but there were no warnings against users on this issue.


UPDATE:

A poll has been started. Users are requested to vote in large numbers.

$\endgroup$
10

2 Answers 2

7
$\begingroup$

update There was a further change in policy (see link in OPs question or Karsten's answer). Originally the choice was to opt in to display a banner, of which there are two styles (so 3 choices altogether counting also the no-banner option), but now there are two choices, either displaying the more drastic

**Reminder**: Answers generated by artificial intelligence tools are not allowed on [Site Name]."

or the default

**Reminder**: Answers generated by artificial intelligence tools must be cited on [Site Name]."

Officially the justification for the default is the site-wide proper attribution clause of the Inauthentic Usage policy, according to the accepted answer. According to the policy, the use of AI violates the expectation that content was produced by a human.

I think it's more complicated. Content on SE comes with no guarantees whatsoever, so the question of attribution is primarily one of integrity (honesty), akin to not plagiarizing content. Attributing output to AI is different from regular attribution in that AI output is anonymous and untraceable to a particular person or text source. Attribution at best serves as a warning to readers that what they are about to read comes from a black box and might be nonsense.

Use or not of AI does not alter the burden of ensuring quality, which requires screening a post for accuracy, including material sourced from another site. And quality is in the OPs best interest to retain reputation. Quality is also enforced by a voting system which assigns rep ("vetting by a community of human readers").

In fact, vetting by the community complicates matters, because candid use of AI through attribution generally gets a negative reception, whether AI is allowed or not, to the detriment of an honest user. Moreover, incontrovertible detection of AI can be difficult, and attribution is complicated when AI might be used to edit only parts of a post. Finally, it is hard to justify banning incorporation of AI content that improves a post. In fact, AI should increment quality if the OP can tell a correct from an incorrect answer and does a reasonable editorial job (ie an incorrect answer is still worse than no answer). Bayesian logic says that a post is like a trial wavefunction: any additional vetted information should improve quality.

We could opt for a "don't ask, don't tell" option which officially bans use of AI, but where we pretend there is no such thing as AI. This is in fact close to my current policy. I keep my eyes open but don't fuss over AI. But the need to choose a banner means we can't make this official policy.

Therefore I suggest officially banning AI use, since it reduces the likelihood of burdening the site, making exceptions for use in formatting, spell-checking, or assistance with translation. Quality remains the OPs responsibility.


General opinion on AI

AI is a tool, and like any its use should be judged based on intent and effect. Users should use it with caution, and in particular not post raw output from a chatbot such as ChatGPT.

Reasons include the purpose of answers on this site as expert-curated content. When you post an answer you should expect to vouch for accuracy by providing supporting statements (a logical exposition or supporting references). A chatbot does not provide references, is known to be often illogical, and cannot be held accountable for accuracy.

No one here can interfere with your use of AI to research an answer, or even to assist with writing text and formatting one. AI should however be used as a tool as might a web search engine. Ultimately the person posting should serve as a filter to ensure usefulness and accuracy.

$\endgroup$
17
  • $\begingroup$ I get your answer, and the second part of the question is done and dusted. Now looking forward to your action on the first part ! $\endgroup$ Commented Jan 26 at 11:20
  • 4
    $\begingroup$ @HarikrishnanM As moderators, we don't act on policy until we know what the community wants. You acted by posting the question. Now we await the community response. Patience you must have... (Star Wars quote). $\endgroup$
    – Karsten Mod
    Commented Jan 26 at 12:01
  • 1
    $\begingroup$ Do we actually have a policy on "AI"? Should we used this as the basis of discussion? I remember reading about these things here, but actually guidelines or the like don't come to mind. $\endgroup$ Commented Jan 26 at 15:33
  • 1
    $\begingroup$ @Martin-γƒžγƒΌγƒγƒ³ The policy as I understood it has been that posts generated by AI are not accepted. The problem is that there is no way to reliably detect AI-generated content, but I haven't kept up with the state of this. Either way, my answer here is more a short rationale for why unfiltered answers generated by AI are unacceptable. I don't pretend that it covers all possible reasons. Presenting an AI answer as ones own is dishonest, bordering on plagiarism. Spamming the site with low-quality answers might justify a ban. $\endgroup$
    – Buck Thorn Mod
    Commented Jan 27 at 17:19
  • $\begingroup$ Agreed. I also didn't follow closely, so I'm left to wonder if we need to, as a community, create a site policy. Or can we relay that to a network one? I guess we should at least create a faq or something, if we wanted to have that banner. $\endgroup$ Commented Jan 28 at 9:23
  • $\begingroup$ Another concern to all the mods: even if we check every answer for AI generation, ppl can go to websites aimed at paraphrasing the AI-generated answers and get them paraphrased. The final answers on checking will show that it is 100% human generated. $\endgroup$ Commented Jan 28 at 14:46
  • $\begingroup$ The question is simple but the means are not too straight forward. $\endgroup$ Commented Jan 28 at 14:47
  • 1
    $\begingroup$ @Harakrishnan using AI to write answers is similar to reputation farming. You are also likely to be posting a lot of junk. One way to get caught is by answering a large number of questions in a short amount of time, particularly if you copy-paste answers in the inimitably bland AI style. Once caught it'd be difficult to weasel yourself out of a bad reputation, for instance through edits, people will question what you post and are less likely to upvote your answers - just the opposite. $\endgroup$
    – Buck Thorn Mod
    Commented Jan 28 at 19:08
  • $\begingroup$ @HarikrishnanM meta.stackexchange.com/questions/395985/… $\endgroup$
    – Mithoron
    Commented Jan 31 at 13:32
  • $\begingroup$ Saw that @Mithoron, completely understand. This deepens the problem. But there are people who can translate.... is it not? I know English, Tamil, Hindi, French, Malayalam... there might be people who know German... $\endgroup$ Commented Jan 31 at 13:38
  • $\begingroup$ I remember that one person many days ago posted a question in Bahasa Melayu and was terribly downvoted $\endgroup$ Commented Jan 31 at 13:39
  • $\begingroup$ GTranslate, Bing translate also could be used. $\endgroup$ Commented Jan 31 at 13:45
  • 1
    $\begingroup$ AI is a tool. If its answers become consistently reliable there might not be much point to SE any longer, at least for those asking questions, but we are not there yet. For the time being AI is a tool you can use to accelerate the production rate or quality of questions or answers. How you use it determines how you should be "judged", or you and your posts. If you decide to spam the site with content of dubious quality you risk damage to your reputation or a ban, simple. Using AI to improve formatting, grammar, etc, is ok imo. It remains your responsibility to ensure quality. $\endgroup$
    – Buck Thorn Mod
    Commented Jan 31 at 16:32
  • $\begingroup$ Another perspective: using AI is like using SE to obtain an answer. It is a free tool. You could instead hire someone (a tutor maybe) to provide the answer. However AI is currently a rather unreliable worker. $\endgroup$
    – Buck Thorn Mod
    Commented Jan 31 at 16:34
  • $\begingroup$ @BuckThorn And for me so called generative AIs are pretty much tools for stealing everything on the internet to make money. Definitely aren't "free". $\endgroup$
    – Mithoron
    Commented Feb 1 at 14:16
2
$\begingroup$

The question is almost moot now, as the banner will be the default on all sites, and the help pages will be updated to give guidance on AI policies: Sites can now request to enable a banner to warn about their policy on AI-generated content (not the best title for the updated content of the mother meta post).

$\endgroup$
5
  • $\begingroup$ I updated my answer. I think a proper ban is best for the site. $\endgroup$
    – Buck Thorn Mod
    Commented Feb 1 at 10:42
  • $\begingroup$ When is the default being updated? When can I expect to see the banner on the answer? $\endgroup$ Commented Feb 1 at 11:22
  • $\begingroup$ Also should I edit the tag and say tag[status-completed]? $\endgroup$ Commented Feb 1 at 11:22
  • $\begingroup$ It might be worth starting a fresh post since Harikrishnan's was created before the change in policy. The new post might serve explicitly to vote on the choice of banner with two answers, one for each banner, and visitors upvoting the one they prefer. $\endgroup$
    – Buck Thorn Mod
    Commented Feb 2 at 8:13
  • $\begingroup$ Of course! Please do start. $\endgroup$ Commented Feb 4 at 4:22

You must log in to answer this question.

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