25
$\begingroup$

I won't beat the bush here: for cognizance of the community, I would briefly state all the recent events that unfolded over the past week or so. And then (my personal opinion) there would be a message for the community.

What is happening?

The menace of AI-generated content is not a new thing. But how they were and would be handled is the bone of contention.

SE Inc. claimed (albeit without bothering to show the exact figures, methodologies and analyses):

We recently performed a set of analyses on the current approach to AI-generated content moderation. The conclusions of these analyses strongly indicate to us that AI-generated content is not being properly identified across the network, and that the potential for false-positives is very high.

Then they rather blatantly imposed on the moderators new policy to handle suspension emanating from such posts and dictated:

require[ing] an immediate cessation of issuing suspensions for AI-generated content and to stop moderating AI-generated content on that basis alone, affording only one exceptionally rare case in which it was permissible to delete or suspend for AI content.

Why did Mods go to strike?

As has been conspicuous in the actions taken by SE, of late, this time too it was marred by poor interaction with the stakeholders and rather stern standing against any form of constructive feedbacks.

More specifically:

The new policy overrode established community consensus and previous CM support, was not discussed with any community members, was presented misleadingly to moderators and then even more misleadingly in public, and is based on unsubstantiated claims derived from unreviewed and unreviewable data analysis. [...] There has been a lack of communication with moderators and a lack of communication with the community. When communication happened, it was one-sided, with Stack Exchange, Inc. being unwilling to receive critical feedback.

So

As of today, June 5th, 2023, a large number of moderators, curators, contributors, and users from around Stack Overflow and the Stack Exchange network are initiating a general moderation strike. This strike is in protest of recent and upcoming changes to policy and the platform that are being performed by Stack Exchange, Inc.

Moderators' response

Moderators are a vital cog in the machine. They countered the allegation made:

The decision to delete content or suspend users is always taken seriously, and experienced SE moderators are not foolish enough to rely blindly on ... any automated tool, really. Just like we don't automatically suspend everyone who shows up in "suspicious votes" tools, we don't automatically delete every post that automatic GPT detector tools claim is AI-generated, nor suspend every such poster.

To reiterate:

Just one tool saying one answer is AI is not enough to get a suspension. Moderators look for various heuristics, including posting patterns, answer speed, writing style, and others, in order to identify AI-generated content. This can include checking an automated detector, but this is not a primary method used, due to the false positive rate. While any one method can be unreliable, when an answer matches multiple heuristics, moderators can be relatively confident that the post is indeed AI-generated.

As Sycorax - On Strike aptly asserted:

The change in policy is poorly justified and communicated (and SE employees have said contradictory things in different channels in the past week), but the worse part of it is that SE is ordering us, as elected moderators, unilaterally on how to do our volunteer work. We were elected to apply our best judgment to maintain & improve our sites, using donated time and expertise, but SE Inc has decided that they Know Best, Actually (™️)... for secret reasons.

In solidarity with the mods

The recent rule imposition is an antithesis of what SE stands for. Our long goal is to nurture a high-quality repository of no-nonsense q&a. Botched implementation of such half baked rules and diktats would be nothing but an hindrance with respect to the spirit of SE with the pretext of attracting new users (which is also important but not in the name of loosening moderation over GPT generated garbage).

So, I urge the community to sign this open letter - yes, even users without diamonds can do that too - and show solidarity with our moderators. We are proud two of our moderators have signed in the letter (Sycorax - On Strike ♦, Scortchi - Reinstate Monica ♦).

As the letter suggests:

Even if you are not a moderator you can participate in the strike by:

$\bullet$ Not voting on posts

$\bullet$ Not submitting edits

$\bullet$ Not reviewing in the review queues

$\bullet $ Not commenting

$\bullet$ Not flagging posts

without resorting to any form of vandalism or other destructive tactics.

I would like to quote Scortchi - Reinstate Monica:

These rights hardly amount to democracy; but they do amount to the difference between enlightened despotism & mere despotism, between governance through promulgated standing laws & through extemporary arbitrary decrees.

All are welcome to act according to their own will - they can agree without even adhering to the suggestions altogether.


Update as of $26$th of June: link.

Update as of $4$th of August: Moderation strike: Results of negotiations. Also this.

$\endgroup$
5
  • 1
    $\begingroup$ I am not aware of the other mods, so apologize for not mentioning them. $\endgroup$ Jun 6, 2023 at 3:37
  • 8
    $\begingroup$ Thanks. I do hope "What side are you on?" isn't meant too seriously though - people with the best interests of our site at heart will still come to different decisions. $\endgroup$ Jun 6, 2023 at 9:19
  • 8
    $\begingroup$ And some good news at last: company analysts have been allowed to start talking to moderators about their findings & concerns. $\endgroup$ Jun 6, 2023 at 9:27
  • 1
    $\begingroup$ Thanks @Scortchi-ReinstateMonica for the update. Yes, in the long run, we are not making anyone antagonist(s). Disagreement may occur but what I loathe is the dearth of proper interaction and in the pretext of listening to feedbacks, ultimate imposition of the same. $\endgroup$ Jun 6, 2023 at 9:30
  • 2

3 Answers 3

19
$\begingroup$

I'm primarily participating in the strike because SE has

  1. Used the private, moderator-and-staff-only Teams platform to state a new policy for moderating AI-generated content. Mods are instructed not to share things posted to the private Teams platform, so I won't go into detail. The justification for this change in policy makes allusion to an analysis that SE conducted internally, but was not shared with the moderators at the time the policy was announced. (Some elements have been shared since then, but I haven't had time to review it.)
  2. Stack Exchange, Inc. seems poised to unilaterally change the moderator agreement, without soliciting our input.

Both announcements make factual claims about how AI content is moderated that I have trouble believing.

My objections are

  1. The two statements are not entirely consistent with each other. I can't follow a policy that is inconsistent. Moreover, SE staff have also stated different versions of the policy in mod-only chatrooms, so even among SE staff, there is not a coherent understanding of the policy.

  2. It strikes me as deeply troubling to have a secret policy (the one posted to the private Teams) that governs how users interact with the site. I think the site can only work and be fairly moderated of the site is transparent about that process.

  3. Diamond moderators and site curators want to be a part of the solution to moderating AI content. We want to find bad content and remove it. Stack Exchange has the stated purpose of creating a high-quality repository of question-and-answer content and this is a goal that we believe in; it's why we joined the site and volunteered our time. Catching us by surprise with a binding change to how we moderate content is not the way to improve how we do that task.

  4. Moderators have been asking for years to have better tools to help us detect low-quality content. Stack Exchange Inc. hasn't built them. But now that the network is facing a massive influx of AI-generated content, Stack Exchange has additionally announced that we are forbidden from using GPT detectors as a heuristic to aid in detection of low-quality content. Ok, fine, great, we can accept the loss of a tool that's prone to error, but if Stack Exchange not going to build any tools that do work, how are we supposed to do this job?

All of this is avoidable. Stack Exchange, Inc. has policies and processes for changing how the site works. There are also innumerable ways that they could solicit moderator input and feedback on how to achieve their goals. The sites themselves are home to hundreds, if not thousands, of experts in a variety of fields. Moreover, these people are all actively involved in using and moderating the sites. Surely there is some value to be found in tapping their expertise in curating content.

I recognize that my objections are primarily based in procedural issues, not the policy itself. It might be hard to understand why I feel so strongly about procedure, but imagine that you had volunteered to a particular role and then were subjected to arbitrary revisions to what that role is and the expectations for your performance. Moreover, you do not have the tools you need to do the job efficiently, so you improvise with some (relatively poor) tools... then they take away those tools. I think most people would object to that.

I think that a lot of the problems come down to poor communication. I am consistently baffled by how Stack Exchange staff interact with the volunteers (diamond moderators or otherwise) who do the grunt work of keeping garbage off the sites.

The coordinated action of a group of moderators working together to influence Stack Exchange, Inc. seems like the best way to effect change. This is why I decided to join the strike.


Whuber writes

My sense is we had developed a consensus that we should continue moderating posts on their intrinsic merits -- not on whether a chatbot (or for that matter, a computer algebra system or even a calculator) was employed to create them. (Personally--this might not be part of the consensus--I see potential advantages in the judicious use of chatbots to perform research and improve writing style, provided the poster takes responsibility for the accuracy of their work.)

As a philosophy of moderation, I agree: the reason to remove low-quality content is that it is low-quality content. If a responsive, factually correct and well-reasoned answer is submitted, we are all better off, regardless of the tools that produced it.

The problem is that the numbers are against us. AI is an automatic bullshit machine, so automation or persistence among bad actors might flood the site with useless content. The past few decades are replete with examples of cases where humans were overwhelmed with computer-generated bullshit. There was a period not too long ago when spam nearly made email unusable. To this day, various kinds of computer security issues persist due to bad actors using automated tooling to create malware and malicious websites.

This is especially close to home for me because I work in the computer security field. I've seen how this might end: without dramatic improvements to the tooling and automation around how content is curated, Stack Exchange will not be a useful resource because it will be overwhelmed with garbage, AI-generated answers.

So far, Cross Validated has been fortunate that we have not seen the same flood of poor answers that the high-traffic network sites have. But I don't know how long that will last.


Regardless of other site moderators participating or not participating in the strike, I hold them in the highest esteem.

$\endgroup$
2
  • $\begingroup$ the part you wrote on tools reminds me of Mithical's spoon analogy $\endgroup$ Jun 6, 2023 at 18:09
  • 1
    $\begingroup$ @starball Since the only development effort seems to be Twitter-driven, maybe the solution is to write a viral Twitter post. We'd only need to do it once, apparently. $\endgroup$
    – Sycorax Mod
    Jun 6, 2023 at 18:13
15
$\begingroup$

Thanks to all the moderators --- your great work is appreciated

I'm not a moderator, and I have only a thin knowledge of this controversy, based on reading some of the posts relating to this strike. My main purpose in this answer is to thank all our moderators for their work on this site and to note that I empathise with the difficulties they are experiencing dealing with an influx of AI-generated answers. This seems to me to be an inherently hard problem to deal with, so I can only hope that our moderators are given appropriate support. I'm a big fan of whuber's recommendation that moderators continue moderating posts on their intrinsic merits (once they are no longer on strike). I have confidence that our moderators can do a good job dealing with AI-generated answers if properly supported.

With that it mind, a general request from the community:

  • Do not post AI-generated answers without attribution to the relevant AI platform. If you want to post an answer that was generated using an AI language tool, make it clear in your answer that this is where it came from and cite the source appropriately (e.g., cite the tool you used and the query that generated the text). Let moderators deal with your post with the aid of candid attribution of your sources.

  • If you are the unfortunate victim of an incorrect attribution of AI-usage, and your question/answer is closed, deleted, etc., try to remain calm and understand that some false positives are an inevitable result of binary predictions applying to thousands of data points. (As statisticians, this should not come as a shock to us.) If you have been incorrectly flagged as using an AI language tool for an answer, stay calm and talk to the moderators about it. It is understandable to feel somewhat offended at having been flagged as AI if you have written your own work, but temper that feeling by acknowledgement that perfect prediction in all cases is an unrealistic expectation.

$\endgroup$
12
  • 4
    $\begingroup$ As a non-moderator too, and a fairly active longer-term member, I want to echo this strongly and add a few further comments. To me there is almost complete dissociation in practice between the fact that a company is behind SE and what I do here. I read and post here out of a mix of wanting to be helpful (which does include reacting to confusion or error as I see it), of wanting to learn from the site (which does include what I learn in writing answers), and of having fun. That all depends on SE being maintained, but the existence of the company is something I can usually ignore. $\endgroup$
    – Nick Cox
    Jun 7, 2023 at 10:08
  • 4
    $\begingroup$ (ctd) That said, it is enough that at least some moderators feel that the company is treating them badly to make me express sadness and sympathy about that. . But I don't think that withdrawing my labour would be noticed by anyone in power. I respect different positions on that detail. $\endgroup$
    – Nick Cox
    Jun 7, 2023 at 10:11
  • $\begingroup$ It is appreciable to have a post from a non-mod active user (Ben) from the community. Whatever maybe the mode, be it joining the strike or not, what it shows is we cherish the tiring and vital job the mods do and we unilaterally agree on the case of treatment of ai generated materials here. And shout out to @NickCox too with whom I agree. $\endgroup$ Jun 7, 2023 at 13:18
  • 1
    $\begingroup$ @NickCox "We're going to ignore it and see what happens" is the position the Physics mods have taken. It has some appeal, as a strategy. My assessment is that concerted effort is the better path forward. One regret I have is that, because the spark to this fire was a poorly-presented policy change, that's where a lot of focus has been. People are debating the finer points of the announcement itself. That's not where the real problems are -- as I see it, the larger problems are that SE hasn't built any good tools for mods to find & remove low-quality content, from any source. $\endgroup$
    – Sycorax Mod
    Jun 7, 2023 at 13:27
  • 1
    $\begingroup$ I work in a University. Some years ago a senior colleague remarked sarcastically "It seems that some of my colleagues think that the University is a democratic institution. I don't know where they got that impression." He was right and it's the same here. There is a company and they sometimes pay no more than lip-service to users despite those users creating most of the content. Your options are limited to protest or exit. $\endgroup$
    – Nick Cox
    Jun 7, 2023 at 13:32
  • $\begingroup$ Thanks @Ben. With regards to the 2nd point the advice is certainly sound, but I hope we have no false positives at all on this site. A practical advantage of only taking action on low quality posts is that the kinds of factual or logical errors ChatGPT can introduce are very distinctive, and leave no room for doubt that human thought was not involved in the writing of the post. (Sometime I can even reproduce the very same errors by feeding ChatGPT the relevant question.) $\endgroup$ Jun 7, 2023 at 13:39
  • 1
    $\begingroup$ Speaking of exit, Math Overflow mods are mulling over whether to secede completely from SO (though they have concluded it would be a bad spectre to do this right now). But this recent happening have again ignited the debate there. $\endgroup$ Jun 7, 2023 at 13:41
  • $\begingroup$ Goodness know how any group could take their content with them if they secede. Who would want to start all over again? $\endgroup$
    – Nick Cox
    Jun 7, 2023 at 13:55
  • 1
    $\begingroup$ @NickCox, see this post At what point should MO recognise it needs to leave Stack Exchange Inc?. $\endgroup$ Jun 7, 2023 at 13:58
  • $\begingroup$ Thanks; enough to note that MO is in a rather special situation. $\endgroup$
    – Nick Cox
    Jun 7, 2023 at 14:39
  • 1
    $\begingroup$ +1 As an interested non-mod myself, even thinking that for the moment such postings can be well managed based on intrinsic merits, regarding current events I can see a striking difference between the information available to me posted by mods supporting the strike and posted by the company, and this doesn't reflect positively on the company to say the least. $\endgroup$ Jun 7, 2023 at 14:40
  • $\begingroup$ @ChristianHennig, please also have a look at this: Dear SE, please stop disparaging the moderators in the press (if you haven't already) to see how the incident was presented to the outside world. $\endgroup$ Jun 7, 2023 at 14:52
12
$\begingroup$

Some context

This is news to me, so please don't read anything into what actions I might have taken or did not take. I devote my time to this site, not to SE generally; but I would think that in any real strike the organizers would (at the very least) have attempted to contact all diamond mods across the system.

About the letter

I have followed some of the Meta discussion about AI-generated content and am aware of SE's recent statements about it. However, I have not even read the entire open letter because its governing assumption, "Stack Overflow, Inc. has decreed a near-total prohibition on moderating AI-generated content," does not accord with the facts as I know them. That misconstrues the statements I have read and appears to me like an overblown reaction. Unless additional, contrary facts come to my attention, I don't think I will be signing that letter nor changing my interaction with the site.

Implications for us

As far as how CV moderates AI-generated content, your diamond mods have been discussing this privately for over half a year (since the public release of ChatGPT). My sense is we had developed a consensus that we should continue moderating posts on their intrinsic merits -- not on whether a chatbot (or for that matter, a computer algebra system or even a calculator) was employed to create them. (Personally--this might not be part of the consensus--I see potential advantages in the judicious use of chatbots to perform research and improve writing style, provided the poster takes responsibility for the accuracy of their work.)

What your mods have been doing

We are, however, closely monitoring the site for evidence of abuse of chatbots and we have been quick to close down activity that reflects such abuse, such as posts that spew statistical nonsense or users who appear to be spamming the site with nothing but chatbot output. But this doesn't require any change in our moderation policy at all and, as far as I can tell, is consistent with recent policy statements from SE.

$\endgroup$
11
  • 1
    $\begingroup$ Have you read this entire post on the Stack Moderators Team? That may account for the discrepency in the AI-generated policy being referred to. $\endgroup$
    – Mithical
    Jun 6, 2023 at 17:08
  • 4
    $\begingroup$ @Mithical I saw that when it first appeared. I upvoted it because I agreed with, and still agree with, it: namely, I have for the past half year advocated not relying on chatbot detectors to moderate our site and not suspending users solely for using chatbots. It summarizes my perspective at the end pretty well: "You may still naturally deal with content abuse in ways we always have: if someone is posting a flood of poor quality content, suspend them for low quality content. Just don’t make it about GPT or AI generated content - make it about the quality being poor on a repeated basis." $\endgroup$
    – whuber Mod
    Jun 6, 2023 at 17:47
  • 3
    $\begingroup$ (Continued) This is misrepresented by the initial thesis of the strike letter, which asserts "Stack Overflow, Inc. has decreed a near-total prohibition on moderating AI-generated content." That reads like a gross misrepresentation of the post you reference. $\endgroup$
    – whuber Mod
    Jun 6, 2023 at 17:48
  • 3
    $\begingroup$ I'm not a mod but I think part of the discrepancy between here and elsewhere is that many mods from SO say that they're no subject matter experts and cannot or do not judge posts by their quality, only by formal criteria. If they cannot apply the formal "suspected of being produced by an algorithm" reason, they don't know how else to moderate the content. They apply their judgement by having seen many AI generated texts and try to find similarities between these and new, suspect posts. They fear that otherwise high volume of such content is created. All this might be different for CV though. $\endgroup$ Jun 6, 2023 at 20:09
  • 8
    $\begingroup$ @Trilarion Thank you for those comments. We (the mods on CV) do appreciate differences between sites. None of us maintain that what works for us ought to work for them or vice versa. It clearly is a challenge to distinguish chatbot BS from legitimate answers. Fortunately, the nature of questions and answers here on CV is such that many, if not most, of our mods and most active users can usually spot the difference quickly. But as bad (or at least naive) chatbot users grow more sophisticated, the potential to be flooded with good-looking but wrong posts could become a more serious problem. $\endgroup$
    – whuber Mod
    Jun 6, 2023 at 20:48
  • 5
    $\begingroup$ Thank you. (1) S.E. users aren't unionized, & the organizers of this strike aren't shop stewards - or even a clearly defined group of people. I think they've done a good job, considering. I'm most to blame for your being taken unawares - I posted in the mod. chat about it before I'd decided to join, but neglected to ping anyone - sorry. (2) You make a very fair point about the letter - it's written from S.O. moderators' stand-point, as @Trilarion explains. $\endgroup$ Jun 7, 2023 at 7:55
  • 4
    $\begingroup$ (3) On C.V, we certainly have reached the consensus "that we should continue moderating posts on their intrinsic merits". (I personally agree with it too). $\endgroup$ Jun 7, 2023 at 7:56
  • 2
    $\begingroup$ My knowledge of this episode is limited to reading a few of the publicly available posts. I did have some sense from reading various posts that some phrasing seemed a little hyperbolic & that in some cases people may be overreacting. Without knowledge of things that are out of my sight, I can't fully judge that. It's an impression based on the style. Nonetheless, the thrust of what is said unfortunately matches my priors about SE based on previous events. As a result, I basically believe it. I signed the letter. At present, I don't regret doing so & don't intent to try to retract my signature. $\endgroup$ Jun 7, 2023 at 12:04
  • 7
    $\begingroup$ The other thing I would say is that I found the episode around Monica's mistreatment to be exhausting. Resigning and walking away was the best thing for me. I don't intend to be heavily involved in this episode or try to follow all the ins and outs. My estimation of SE will drop further, as will my activity here probably, but as I said before, I have no illusions that SE will notice or care. $\endgroup$ Jun 7, 2023 at 13:15
  • 1
    $\begingroup$ I was not familiar with the Monica episode that you mentioned in the chat @gung-ReinstateMonica; thanks for mentioning that (I have looked into few news articles and other SE posts). Now that I have seen, I can comprehend where you are coming from. To be fair, the trait of abruptly imposing a diktat without proper consultation seems to be utter baloney irrespective of what reason it was apparently done for and then what follows is a slow mode of indifference. $\endgroup$ Jun 7, 2023 at 13:22
  • 4
    $\begingroup$ It might be worth mentioning that I have a long history on Q&A sites. I founded several discussion groups in the late 90's (using listserver platforms), initially as a direct response to the patronizing, draconian approach taken by a major software vendor to moderating their official groups. I have seen what "despotism" in this arena looks like. SE is mild by comparison. Until they start actively interfering with our moderation here on CV, I see little to get upset about. $\endgroup$
    – whuber Mod
    Jun 7, 2023 at 13:29

You must log in to answer this question.

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