25
$\begingroup$

I hardly ever visit Meta so forgive me if this has been addressed already. I just wanted to flag an issue:

The user Statistics Tutorial is a bot. So far I have deleted, along with other community members, two of its answers for being blatant GPT word-soup but it is persistent. At the time of writing it even has an accepted answer. Maybe this answer is fine, maybe it isn't - I haven't read it carefully. But people seem to be upvoting it (every post of its that I've seen has at least one upvote) unaware that it is a bot and of course this is a problem, since it has absolutely no understanding of what it is saying.

On the whole, the site would benefit from having this bot gone. At best, it posts the occasional reasonable elementary answer. At worst, it continues to write word soup that fools the less expert reader and misinform people. For example, it attempted to answer my question by unashamedly copying everything I'd already said and representing this information in bullet point form. Whether this is a real problem or just annoying is up to you. I have also seen at least one instance of an established user (with over $6000$ reputation and $9$ years experience) posting blatant bot posts.

My question is - what can we do to counter this? I can flag all of its answers, but there seems to be no obvious way to flag an entire user.

Of course, there is a risk of a witch-hunt, with bot accusations flying and accounts being unfairly penalised. So if there ever is such a feature, it should be restricted to high reputation users (in my opinion) and be subject to some kind of community review.

Inspecting their SO account, I think it is more likely this user is human but just testing out bot capabilities on MSE for a laugh. Either way, I find that problematic.

$\endgroup$
14
  • 10
    $\begingroup$ If you flag one answer for moderator attention, and you give your reason for flagging, that should be enough to draw the moderators' attention to the problem. $\endgroup$
    – Gerry Myerson
    Commented Sep 29, 2023 at 22:09
  • 5
    $\begingroup$ A recent contribution to the literature on the limitations of AI, and very relevant to sites such as math.stackexchange, is at bit-player.org/2023/ai-and-the-end-of-programming $\endgroup$
    – Gerry Myerson
    Commented Sep 29, 2023 at 22:13
  • $\begingroup$ Related: math.meta.stackexchange.com/questions/35651/… $\endgroup$
    – Shadow
    Commented Sep 30, 2023 at 12:05
  • 1
    $\begingroup$ A couple of the terms used in the title deserve some explication. Thinking over the post as a whole leads me to suspect that "no-platform" means some flavor of account suspension or deletion. Also "bot-users" might mean accounts run by bots or Community members who use "bots" (LLMs) to compose posts. A great deal of time was spent this year by moderators and corporate SE to negotiate around the issue of posting ChatGPT and like "AI" content, so there's a lot of catching up on your concern. $\endgroup$
    – hardmath
    Commented Sep 30, 2023 at 15:36
  • 2
    $\begingroup$ For the 6K user who seems to have gone rogue, keep in mind the possibility that their account got hacked. Of course the appropriate response doesn't change, but your moral indignation might. $\endgroup$
    – Nate Eldredge
    Commented Sep 30, 2023 at 20:05
  • 6
    $\begingroup$ Oh, I see, "no-platform" is a verb. I thought it was an adjective and got confused. "Deplatform" might be more familiar, but you could also just say "ban". $\endgroup$
    – Nate Eldredge
    Commented Sep 30, 2023 at 20:06
  • 1
    $\begingroup$ @hardmath no platform just means what it has always meant; restricting exposure on, access to and support from a certain social media platform. No platforming bot users wouldn't necessarily entail deleting their accounts (a lot could go wrong there) but it could mean somehow ensuring their posts got very minimal positive exposure (unless it was reviewed, etc. details shmetails) $\endgroup$
    – FShrike
    Commented Sep 30, 2023 at 22:50
  • 1
    $\begingroup$ @NateEldredge I don't necessarily mean something as strong as a ban. I was under the impression "no-platform" is a standard umbrella term on social media $\endgroup$
    – FShrike
    Commented Sep 30, 2023 at 22:51
  • 7
    $\begingroup$ "No-platforming" seems to be a "also known as" for deplatforming, which at least to this "olds" seems to be the more prevalent term in the U.S. $\endgroup$
    – Arturo Magidin
    Commented Oct 1, 2023 at 1:50
  • 2
    $\begingroup$ Am also an old, and "deplatform" is also the term I'm used to. $\endgroup$
    – JonathanZ
    Commented Oct 1, 2023 at 2:58
  • $\begingroup$ @FShrike "no platform just means what it has always meant;" $\quad$ In the UK, yes. $\endgroup$
    – ryang
    Commented Oct 1, 2023 at 3:10
  • 7
    $\begingroup$ When the same "established" bot-poster mentioned in your post answers your question with a bot-post... seriously, these are trolls, and trolls must be eliminated by grouping and crowding them out, period. A witch hunt is very justified when you're in fact hunting a witch. $\endgroup$
    – Sarvesh Ravichandran Iyer
    Commented Oct 3, 2023 at 12:12
  • 1
    $\begingroup$ @SarveshRavichandranIyer A witch hunt is not justified when you don't need to hunt witches. Diamond moderators have the tooling to identify whether someone's actually the same bad actor: without that, you risk getting innocent bystanders caught in the crossfire, which is exactly why we don't like witch hunts. Just flag, and raise it on meta if it gets particularly egregious – and escalate to staff if you think the moderators aren't doing their jobs properly. $\endgroup$
    – wizzwizz4
    Commented Oct 3, 2023 at 16:09
  • 1
    $\begingroup$ @wizzwizz4 I agree with you, but I seek a middle ground. I'll flag, but get mods to prioritize the flag in that case. The longer that kind of post stays the worse the impression it gives out, because to me there aren't much more blatant offenses than this. $\endgroup$
    – Sarvesh Ravichandran Iyer
    Commented Oct 4, 2023 at 12:27

2 Answers 2

Reset to default
10
$\begingroup$

One characteristic of GPT-generated answers is being confidently wrong. Additionally, Mx. Speaker, an existing answer to this question makes a few errors, some of which I believe warrant correction.

I can flag all of its answers, but there seems to be no obvious way to flag an entire user.

If you wish to raise a concern about a particular user, it is common practice to flag a single arbitrary post (question or answer) with the “In need of moderator intervention” reason. For completeness: there is no particular penalty for having your posts flagged in this way, so feel free to use it for things that aren't bad behaviour, but nonetheless require moderator attention.

If you were to escalate to staff, there would not be a back-and-forth (in this case), and they would probably not be providing guidance. Since it's not a user safety issue and it doesn't require particular discretion, the usual procedure would be to post on meta (as you have done), and have a moderator escalate the meta post to staff if staff intervention is required (e.g. to use the super duper mod tooling).

Of course, there is a risk of a witch-hunt, with bot accusations flying and accounts being unfairly penalised.

It's usually wise to talk in generalities on meta. (Regarding this answer's second sentence, do as I say, not as I do.) If you pick specific examples, use the posts rather than the users. Be aware that serial voting is forbidden on Stack Exchange, and witch hunts are generally frowned upon.

I have seen an unfair accusation or two, in various places on the network, but I've not seen an account be unfairly penalised for abusing text generation software. Though, there's not really a way of knowing for sure that that doesn't happen: there's an experiment in the works (of undisclosed nature) to double-check, which will probably yield results this year.

$\endgroup$
1
  • 2
    $\begingroup$ "Tag an entire user..." It is enough to tag one answer, and explain the situation. $\endgroup$
    – GEdgar
    Commented Oct 5, 2023 at 15:49
7
$\begingroup$

I suppose if everyone did their part and started flagging and reporting such post, then this is really a non issue. The real issue is that not enough people are concerned about moderator activites and upkeep of the forum.

$\endgroup$

You must log in to answer this question.

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