98

There is a user on Stack Overflow, who posted six answers within 12 (!) minutes: 1, 2, 3, 4, 5 and 6.

Those answers are quite clearly AI-generated: common style, extreme speed, extreme variety of covered topics. Seeing the time taken to post new questions, there was no answer quality check from the user: it is simply impossible.

Most of them as of now are deleted, but the user is active.

In light of the new policy regarding AI generated content, what will stop this user from posting a new pack of such generated answers?

And is it viewed as sustainable for communities and moderators to react to such answers? Because this is clearly one of the first users who does this. Once it become a public knowledge that this behavior is allowed, how are curators and moderators expected to sustain any reasonable level of content quality?

Improve this question
19
  • 30
    I'm not sure what you mean by "Because this is clearly one of the first users who does this.". This is a common thing, it's just new that mods can't do anything about it.
    – cigien
    Jun 2 at 9:15
  • 3
    @cigien, well, I meant "who does this in reality of the new policy"
    – markalex
    Jun 2 at 9:17
  • 16
    that moderator is likely risking their diamond with these actions. We are not allowed to delete those posts according to the new policy
    – Mad Scientist
    Jun 2 at 9:19
  • 3
    I'm not aware of new policy details, but I was lead to believe, that new policy doesn't allow to ban users, but not to delete low quality posts. Additionally, I think those posts where deleted by author due to downvotes. So I assume mods are safe for now.
    – markalex
    Jun 2 at 9:23
  • 10
    @MadScientist that's the crux here, I believe. The temporary policy to not allow Chat-GPT generated content on Stack Overflow has not been revoked (yet) and it is thus still not allowed to post AI generated content. However, under the new network wide policy, moderators may not act upon users violating said policy. So indeed, there seems to be nothing we can do except shaking our fist at them (post a comment and down vote). Basically, the new network policy gives users a free pass to flaunt the non-revoked do-not-post-AI-content rule.
    – Adriaan
    Jun 2 at 14:21
  • 1
    "Most of them as of now are deleted, but user is active." Just to play devils advocate: How can it be that the content was deleted in light of the new content policy?
    – NoDataDumpNoContribution
    Jun 2 at 14:51
  • @Trilarion, I believe it was deleted by author, once attention to there spree was brought in one of curation rooms, and this answers were dovnvoted.
    – markalex
    Jun 2 at 14:54
  • 4
    @markalex You may not see the message above these posts (?) but they say the answers were removed by some user with a diamond in its name.
    – NoDataDumpNoContribution
    Jun 2 at 14:58
  • @Trilarion, yeah, sorry: not enough rep to see deleted posts. I was lead to believe that answers were deleted by author (maybe this is the case of mods ensuring these answers wouldn't be restored later?)
    – markalex
    Jun 2 at 15:04
  • 10
    "it is simply impossible", no, it's possible if the used wrote those answer offline. I know at least two users that does this. They have several draft answers off line and post them once they believe they are ready.
    – Braiam
    Jun 2 at 17:35
  • 3
    @markalex The public-facing message to mods was “don’t suspend users”, but mods were instructed privately not to action those posts at all.
    – Thomas Markov
    Jun 2 at 19:25
  • 6
    "What will stop this user from posting new pack of such generated answers?" Nothing. Which must be the goal. At least that's the only logical conclusion we can draw from the CEO pushing this kind of change.
    – TylerH
    Jun 2 at 19:41
  • 5
    @TylerH Suspension is how we stop behavior we wish to discourage.
    – tchrist
    Jun 3 at 1:31
  • 1
    One could only hope, @mickmackusa. But since they postponed new incoming damaging policies, this might be a ray of light in the current very strange tunnel.
    – markalex
    Jun 3 at 11:18
  • 3
    A pattern often seen when users (who would normally only post questions) get into a question ban (for example, by deleting too many questions) is they suddenly start to post answers as a means to get out of the question ban. That could be the motive here. Nowadays, the method requiring the least amount of work is using ChatGPT.
    – This_is_NOT_a_forum
    Jun 4 at 3:04

4 Answers 4

Reset to default
106

As a devoted, volunteer content curator on Stack Overflow for the last 2299 consecutive days, a moderator of a community in the SE Network, AND as the user who first discovered the rapidly posted AI content in this question...

I am SUPER PISSED OFF!

This was just one user who was spraying content with mere seconds between posts. Can you imagine how quickly the Stack Exchange Network will be polluted with unvetted content when hundreds of users repeat this strategy? This will overwhelm the outnumbered curators with the time, privilege, and will to curate.

Those who care are now hamstrung.

Moderators cannot take action to impede their efforts.

I cannot bulk downvote answers from the solitary poster because then I will be guilty of targeted voting abuse. This will lead to suspension -- and of course, then there will be one less curator to pitch in.

Before AI-generated posts, it took curators less energy to handle unsavory content than it took other humans to create it. Now posting artificial content takes nearly no effort or time at all.

Don't even get me started on the impact of these hyper-FGITW posters multiplied by the "Upvote Fairies" who sprinkle upvotes on anything that vaguely smells like a reasonable answer.

I am having a hard time seeing a positive future for Stack Exchange if there are no protections put in place.

I also have a hard time seeing my future self devote so much of my free time to a network that allows itself to be aggressively polluted.

Improve this answer
7
  • 69
    If it's any reassurance, you are not the only one who is super pissed off.
    – Cody Gray - on strike
    Jun 2 at 14:21
  • 16
    Moderators have been telling them exactly this.
    – Bryan Krause
    Jun 2 at 15:32
  • 3
    (sorry, I didn't recognize you are also a moderator)
    – Bryan Krause
    Jun 2 at 15:39
  • "I cannot bulk downvote answers from the solitary poster because then I will be guilty of targeted voting abuse." I think you just need to vote on more things from enough other people. If you're voting on content, then you're within your rights.
    – starball
    Jun 2 at 18:48
  • 7
    @starball Nah, I got dinged for serial voting several times this week for downvoting rapid-fire VLQ answers from the same user.
    – Thomas Markov
    Jun 2 at 19:33
  • 6
    It looks like the footgun used this time is so powerful it will not only blow the feet clean off, but also leave a carcass that is not recognisable as a carcass. Anyway, it was an interesting experiment, all 15 years.
    – This_is_NOT_a_forum
    Jun 3 at 1:43
  • 10
    I think this marks a new chapter where the "repository of knowledge" needs to reward curation (more than) / (instead of) rewarding contributors who believe it is a transient help desk. A help desk where it is okay to answer the same question over and over or repeat what others have already said.
    – mickmackusa
    Jun 3 at 2:53
65
+50

Looking at those 6 posts, the new AI policy forbids moderators from deleting those posts or suspending this user. Exactly this kind of content is not possible to moderate anymore without violating the moderator agreement, specifically the additional guidance by SE we are obliged to follow.

Those are the consequences of the new policy. I think allowing obviously AI-generated content is a terrible policy and will harm the sites and the community.

Improve this answer
8
  • 2
    Yet they are now all deleted by a moderator. The way I read it, posts can still be deleted, the user just can't be suspended over it(?)
    – Cerbrus
    Jun 2 at 12:38
  • 32
    @Cerbrus the public policy contradicts the private guidance and private statements by SE staff.
    – Mad Scientist
    Jun 2 at 12:47
  • 10
    @MadScientist The moderator agreement does not require us to follow any policies not publicly posted by SE, or to follow any policies that were posted in contravention of SE's own obligations under that agreement. As such, feel free to ignore the policy and keep doing what is right for your community.
    – Chris
    Jun 2 at 17:05
  • 11
    @MadScientist When I agreed to the moderator agreement, it was with the explicit understanding that all these mandatory policies would be workshopped with the Moderator Council, given to mods as a preview, and then posted publicly in their entirety. Obviously, none of that has happened, in letter or in spirit. Agreements are a two-way street. One who holds their part in an agreement is honest. One who continues to hold their part in an agreement when the other side has already broken it, is continuing to break it, and shows no sign of following it in the future is a sucker.
    – Chris
    Jun 2 at 17:11
  • 1
    @Chris-RegenerateResponse: Quite right. And how on earth are we supposed to justify decisions without giving away what the secret policy imposed on us is?
    – Scortchi - Reinstate Monica
    Jun 2 at 17:15
  • 5
    @Scortchi-ReinstateMonica There's no way. I bet many users have already figured out the content of the secret policy just by the fact that there is a secret policy and that it's stricter on mods than the public policy.
    – Chris
    Jun 2 at 17:17
  • 5
    @Cerbrus Posts "can't" be deleted under the policy. Fortunately, the delete button still works. :p It's not like SE has the staff to start micromanaging every post deletion.
    – Chris
    Jun 2 at 17:46
  • 3
    It's equally outrageous that although the company have admitted breaking the agreement, they still haven't made any attempt to put things right. If the policy on Stack Moderators is the one they want, they need to leave it longer for review (a day is far too little time) & post it on Meta after any emendation - & remove the 'mod-agreement-policy' tag from the current policy on Meta immediately.
    – Scortchi - Reinstate Monica
    Jun 2 at 17:58
43

It's obviously unsustainable. I'm sure even some people at Stack Exchange realize that, so apparently someone up high in the chain is pushing this policy hard.

I applaud whoever deleted these posts, and call on other moderators to prioritize their communities over blindly following the ill-considered policies of Stack Exchange, Inc.

Improve this answer
1
  • 1
    Since CMs aren't seen fighting for this policy, I'm sure big chunk of staff realize that. But I don't if it calms me down, or makes me more angry, that they failed to convey to their management.
    – markalex
    Jun 2 at 17:49
15

It's funny, because according to SE's own analysis,

Based on the data, we would hazard a guess that Stack Overflow currently sees 10-15 GPT answers in the typical day

So we are supposed to believe that this one user posted about half of all GPT answers on Stack Overflow for that day. And Stack Overflow's community supposedly cannot survive if we suspend two (!) users per day.

Improve this answer
2
  • I believe you slightly misunderstood their analysis. SE believes, that what you wrote is true: 2-4 rogue users a day, who post chatGPT answers in this obvious streaks. But they see, that there is bigger number of suspensions, and don't like it. And based on their identification methodology, it seems like they don't have a clue about widespread of chatGPT answers outside of said streaks.
    – markalex
    Jun 12 at 5:36
  • @markalex But the new policy is not "limit the number of suspensions to what our metric predicts the number of GPT posts per day says", it's "no suspensions, even in obvious cases, unless the user admits to using GPT". At least, as far as us non-moderators know, that's the new policy; unfortunately it's a secret.
    – kaya3
    Jun 12 at 15:11

You must log in to answer this question.

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