Moderation Strike: Stack Overflow, Inc. cannot consistently ignore, mistreat, and malign its volunteers

Question

Update

On August 2nd, 2023, negotiations between community representatives and representatives of the company concluded with an agreement being reached. The details of the agreement can be found at Moderation strike: Results of negotiations.

On August 7th, 2023, based on the result of several polls held by various sections of the community, the coordinated call to strike concluded. Further details can be found at Moderation strike: Conclusion and the way forward.

---

Introduction

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.¹ We have posted an open letter addressed to Stack Exchange, Inc. The letter details which actions are being avoided, the main concerns of the signees, and the concrete actions that Stack Exchange, Inc. needs to take to begin to resolve the situation. Striking community members will refrain from moderating and curating content, including casting flags, and critical community-driven anti-spam and quality control infrastructure will be shut down.

However, the letter itself cannot contain all of our concerns, and we felt it was important to share some of the background and details that were not included in the letter in the interest of brevity. We also wanted to touch upon several points at the same time that are related to Stack Exchange, Inc.’s recent behavior.

Background

A history of the Artificial Intelligence policy

On December 5th, 2022, Stack Overflow moderators instituted a “temporary policy” banning the use of ChatGPT in particular on the site. This was instituted due to the general inaccuracy of the answers, as well as that such posts violate the referencing requirements of Stack Overflow. The moderator team kept an eye on community feedback to guide it, and support welled beneath this policy. Similar policies were enacted across the network.

Within the next several days, thousands of posts were removed and hundreds of users were suspended for violating this policy.

Over the next few months, Stack Exchange, Inc. staff assisted in the enforcement of this policy. This included adding a site banner announcing the ban on these posts as well as editing and adding Help Center articles to mention this policy. Moderators were also explicitly given permission to suspend for 30 days directly in such cases, skipping the escalation process that is generally encouraged.

On May 29th, 2023 (a major holiday for moderators in the US, CA, UK, and possibly other locations), a post was made by a CM on the private Stack Moderators Team² (now publicly published). This post, with a title mentioning “GPT detectors”, focused on the rate of inaccuracy experienced by automated detectors aiming to identify AI- and specifically GPT-generated content - something that moderators were already well aware of and taking into account.

This post then went on to require 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. It was received extremely poorly by the moderators, with many concerns being raised about the harm it would do.

On May 30th, 2023, a version of this policy was posted to Meta Stack Exchange and tagged mod-agreement-policy, making this a binding moderator policy according to the Moderator Agreement. The policy on Meta Stack Exchange differs substantially from the version issued in private to the moderators. In particular, the public version of the policy conspicuously excludes the “requirements” made in private to immediately cease practically all moderation of AI-generated content.

The problem with the new policy on AI-generated content

The new policy, establishing that AI-generated content is de facto allowed on the network, is harmful in both what it allows on the platform and in how it was implemented.

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. Moderators are expected to enforce the policy as it is written in private, while simultaneously being unable to share the specifics of this policy as it differs from the public version.

In addition to these issues in how Stack Exchange, Inc. went about implementing this policy, this change has direct, harmful ramifications for the platform, with many people firmly believing that allowing such AI-generated content masquerading as user-generated content will, over time, drive the value of the sites to zero.

A serious failure to communicate

Throughout the process of creating, announcing, and implementing this new policy, there has been a consistent failure to communicate on the part of Stack Exchange, Inc. 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.

An offer by Philippe, the Vice President of Community, to hold a discussion in the Teachers’ Lounge moderator-only chatroom took days to be realized. During that conversation, certain concerns were addressed³, but the difficult questions remained unanswered – particularly about the lack of communication ahead of time.

The problem with AI-generated content

This issue has been talked about endlessly, both all around the Stack Exchange network and around the world, but we feel it’s important to highlight a few reasons why several communities, not just Stack Overflow, decided to ban AI-generated content. These reasons serve as the backbone not only for our moderation stance against AI-generated content, but also why we feel confused and betrayed by Stack Exchange, Inc.’s sudden decision to halt our efforts to enforce our community-supported decision to ban it.

To reference Stack Overflow moderator Machavity, AI chatbots are like parrots. ChatGPT, for example, doesn’t understand the responses it gives you; it simply associates a given prompt with information it has access to and regurgitates plausible-sounding sentences. It has no way to verify that the responses it’s providing you with are accurate. ChatGPT is not a writer, a programmer, a scientist, a physicist, or any other kind of expert our network of sites is dependent upon for high-value content. When prompted, it’s just stringing together words based on the information it was trained with. It does not understand what it’s saying. That lack of understanding yields unverified information presented in a way that sounds smart or citations that may not support the claims, if the citations aren’t wholly fictitious. Furthermore, the ease with which a user can simply copy and paste an AI-generated response simply moves the metaphorical “parrot” from the chatbot to the user. They don’t really understand what they’ve just copied and presented as an answer to a question.

Content posted without innate domain understanding, but written in a “smart” way, is dangerous to the integrity of the Stack Exchange network’s goal: To be a repository of high-quality question-and-answer content.

AI-generated responses also represent a serious honesty issue. Submitting AI-generated content without attribution to the source of the content, as is common in such a scenario, is plagiarism. This makes AI-generated content eligible for deletion per the Stack Exchange Code of Conduct and rules on referencing. However, in order for moderators to act upon that, they must identify it as AI-generated content, which the private AI-generated content policy limits to extremely narrow circumstances which happen in only a very low percentage of AI-generated content that is posted to the sites.

This isn’t just about the new AI policy

While a primary focus of the strike is the potential for the total loss of usefulness of the Stack Exchange platform caused by allowing AI-generated content to be posted by users, the strike is also in large part about a pattern of behavior recently exhibited by Stack Exchange, Inc.

The company has once again ignored the needs and established consensus of its community, instead focusing on business pivots at the expense of its own Community Managers, with many community requests for improved tooling and improving the user experience being left on the back burner. As an example, chat, one of the most essential tools for moderators and curators, is desperately out of date, with two-minute, high improvement changes being ignored for years.

Furthermore, the company has repeatedly announced changes that moderators deem would cause direct harm to the goals of the platform, this policy on AI-generated content among them. The community, including moderators and the general contributor base, was not consulted nor asked for input at any point before these changes were announced, phrased in a manner that indicated that there was no possibility of retraction or even a trial period.

Some of these planned changes have been temporarily held off due to controversy, this strike influencing those decisions, but that does not change the recent tendency of Stack Exchange, Inc. to make decisions affecting the core purpose of the site without consulting those most affected.

The events of the last few weeks seem like history repeating itself. Stack Exchange, Inc. ventures into a new pursuit, this time, generative AI, in contrast with the community’s interests, makes a decision at odds with all feedback available to them, ceases communications with us, and we go on strike. This is similar to what happened last time the community moderators prepared to go on strike.

How we resolve this

Even though the strike may end, many community members are not comfortable with returning to the status quo before the AI policy itself, if nothing else changes. The strike’s focus on the AI policy is not downplaying the significance of SE’s other actions. We deserve much more than just retracting the AI policy. Stack Exchange already made promises after the 2019 debacle that they have since failed to meet. We are worried that Stack Exchange will continue down the same path once the situation calms down.

While the recent actions by Stack Exchange, Inc. are in conflict with the community and take a significant step backward in terms of the relationship between the company and the community, we do not think that our relationship is beyond repair. We do however worry that we are nearing the point at which it cannot be repaired anymore.

While it certainly may be true that the company wants to meet our needs, and wants to care for us, the reality is that this is not happening. It is time to wake up and realize what must be done. Stack Exchange, Inc. is not acting in our interest. It is time to do so.

What the striking users want

For the strike to end, the following conditions must be met:

• The AI policy change retracted and subsequently changed to a degree that addresses the expressed concerns and empowers moderators to enforce the established policy of forbidding generated content on the platform.
• Reveal to the community the internal AI policy given directly to moderators. The fact that you have made one point in private, and one in public, which differ so significantly has put the moderators in an impossible situation, and made them targets for being accused of being unreasonable, and exaggerating the effect of the new policy. Stack Exchange, Inc. has done the moderators harm by the way this was handled. The company needs to admit to their mistake and be open about this.
• Clear and open communication from Stack Exchange, Inc. regarding establishing and changing policies or major components of the platform with extensive and meaningful public discussion beforehand.
• Honest and clear communication from Stack Exchange, Inc. about the way forward.
• Collaborate with the community, instead of fighting it.
• Stop being dishonest about the company’s relationship with the community.

A change in leadership philosophy toward the community

We need business leadership to meet minds with the community members and community managers, because currently, it appears that leadership ignores them.

Immediate financial concerns appear to drive feature development. The community also has feature development wants and needs, but no substantial consideration is given to those needs, let alone resource allocation. The lack of merit leadership gives to the community and CMs even leads to its own business decisions being reckless and harmful, like the AI policy.

Leadership needs a change in philosophy to one that treats the community as more than a product, and values its needs and expertise. Such a philosophy is evidently currently missing, and leadership takes the expertise of its own product for granted. Leadership needs to represent this philosophy in actually allocating resources based on community needs as well as its own, and informing its feature development using the expertise of the community. Development can be guided by both business and community needs!

In conclusion

The sites on the Stack Exchange network are kept running smoothly by countless hours of unpaid volunteer work, and, in some cases, projects paid for out of pocket by community members. Stack Exchange, Inc. needs to remember that neglecting and mistreating these volunteers can only lead to a decrease in the goodwill and motivation of those contributing to the platform.

A general moderation strike is being held until the concerns laid out in the open letter and this post are addressed. Moderators, curators, contributors, and users, you are welcome to join in by signing your name in the strike letter.

If you would like to sign the open strike letter, but do not have a Stack Overflow account, please reach out to @mousetail (start a new room).

Updates

This post contains strike updates (representatives elected, reactions to the GPT data analysis posted by Philippe, conditions for ending the strike).

---

¹ While we’re aware that the legal name of the company is “Stack Exchange, Inc.”, the name “Stack Overflow” is more recognizable, and thus used in the open letter. The “Inc.” serves to demonstrate that our concerns lie with the corporate entity, and not the site itself, its moderators, or individual employees.

² Stack Exchange, Inc. provides a free Stack Overflow for Teams instance for Stack Exchange moderators, allowing moderators to store and share private information, bug reports, documentation, and communication with SO staff.

³ This includes another planned change to the foundational systems of the platform that has the potential to facilitate unprecedented levels of abuse. (This was referred to as “the second shoe” during the planning stages of the letter and the strike, as in “waiting for the other shoe to drop”.) This has been delayed indefinitely while parts of the plan are reconsidered.

Answer 1

Background from a striking SO mod who handled 10000+ GPT flags

I’m a Stack Overflow moderator coming here to give more details on what management has said that’s helped caused me to decide to join the strike, and to give details from my own personal experience handling several thousand of the warnings and suspensions we’ve done at Stack Overflow for ChatGPT-copy-paste abuse.

And I’m framing this as a response to a comment elsewhere from company management way back in December that now looks to have been an unintentional “tell” from management that they may have already started musing way back then about plans to frame it all as a misguided “witch hunt”.

So, to be very clear: The massive effort and hours the curators/flaggers and elected moderators spent together over the last 6 months very successfully stemming the ChatGPT copy-paste tide and shielding our community’s users from all that junk never in, any way shape or form, came anywhere close to actually resembling anything like a “witch hunt” or any other kind of malicious or hysterical misguided analogy that management might want to try to smear us with now.

The reality is instead this:

We’ve been very accurately identifying the ChatGPT posts with an extremely low number of false positives. And I say that as someone who handled a very very large number of the flags.

For support of the assertion above, please really do take time to read on here.

(Non)Use of any detection tools

I want to take a moment here now to respond specifically to the continuing nonsense insinuations (smears) from management (repeated to the media even today) that we have been over-relying on various detection tools that “have an alarmingly high rate of false positives” and their “decision to require that moderators stop using that tool [sic]”.

So, let me make clear:

In the thousands of cases I handled, I never ever used any of the detection tools.

That is, not only did I not “over rely” on detection tools, I never relied on them at all.

And in cases where the flaggers had included scores from any detection tools in their flag comments, I even completely ignored whatever score information the flaggers provided — to the point that I would actually stop reading the flag comments at the point where they were providing score information, because I found no use at all for it.

So not only could it be said of me that I am one of the moderators whose “own analyses were in use as well” — I can say that my “own analysis” was exclusively in use, with zero reliance on any detector information.

Numbers and assertions

To help clear up any doubts on whether I know whereof I speak, here are some numbers:

• Conservatively speaking, for the last 6 months (since the beginning of December 2022) every single day I have been spending 40 minutes a day just on moderating ChatGPT flags. That works out to 120 hours in total that I’ve personally spent on them so far.

• And in those hours I’ve spent on this week after week, I’ve looked at multiple thousands of ChatGPT-flagged posts: in the order of 10000 or even 15000 at this point.

And so I do now assert that in the many hours I've spent actually looking at the ChatGPT-flagged posts, and from the many thousands of those posts that I have actually closely and carefully scrutinized in detail, I have in fact learned to recognize the ChatGPT copy-paste cases (including many of the ones whose content the posters had edited before pasting in, to intentionally obscure the provenance) and I can in fact say this:

I assert: I can, with a very high level of confidence and a very low level of false positives, very effectively identify answers whose provenance is ChatGPT or other AI.

And all that leads me to finish with an invitation…

An invitation to management

To anyone in management who claims the elected moderators have been getting it wrong for the last 6 months, with an unacceptably high rate of false positives, and who wants to challenge my personal assertions about how accurately I can identify the ChatGPT cases: I invite you to work together with me over several days looking at least some several dozens of ChatGPT flags as they come in.

Alternatively, rather than looking at new flags as they come in, I invite anyone to go together through the logs of my moderation activities, and examine some significantly-sized random sample at least of the thousands of ChatGPT flags I have handled.

Either way, let’s spend time scrutinizing the actual content of the flagged posts together, and actually honestly talking together in good faith about which posts we think we can agree are highly likely not the poster’s own original work.

Until anybody who questions how well we’ve been handling the ChatGPT flags actually does something concrete like what’s described above, waving vague suggestions about possible “witch hunts” and other slanders around under the noses of the elected moderators and flaggers is completely irresponsible at best — but in reality, in combination with other indirect accusations and aggressions that key company reps driving company messaging around this have slung at the elected moderators, it is completely unconscionable and completely unprincipled.

Answer 2

Here's a list of per-site meta discussions related to the strike (largely gathered from searching "is:q created:2023-05-30.. strike [discussion]" on stackexchange.com, sorted based on site traffic, and updated on a best-effort-basis. ^{_{stackexchange.com search seems bugged and shows fewer items than the count it returns. I've seen enough bugs now to just not be surprised.}}):

• Meta Stack Exchange:

  • This Q&A
  • Moderation Strike update: Data dumps, choosing representatives, GPT data, and where we’re holding
  • Would the striking moderators welcome a user strike? Should we, the users, go on strike?
  • Does the strike letter oppose AI?
  • What is needed for users to trust the Stack Exchange company?
  • Are users being suspended for being on strike?
  • Does the company acknowledge misusing the press, and why won't it agree on a no-comment policy?
  • Does the company claim that every article the press made on ongoing issues wasn't requested by them and was instead an unsolicited initiative?
  • Mod strike: Is it realistic to require a no-comment press policy?
  • Moderation strike: Results of negotiations
  • Does the Charcoal team plan to run SmokeDetector again?
  • What do you feel like is still missing to end the strike?
  • What kind of percentages are reasonable for moderator consensus on the company violating the Mod Agreement?
  • Moderation strike: Conclusion and the way forward

• Stack Overflow:

  • (mirror of the above post): Moderation Strike: Stack Overflow, Inc. cannot consistently ignore, mistreat, and malign its volunteers
  • How is Stack Overflow Inc. prepared for a possible moderation strike by community elected moderators on June 5?
  • What has happened to lead moderators to consider striking?
  • What is the impact of the moderation strike?
  • What can "ordinary" Stack Overflow users do to support the moderator strike?
  • Could SE management conduct moderator elections with reduced requirements in response to the ongoing moderation strike?
  • Moderation strike: how is the discussion going between the moderators and Stack Overflow, Inc?
  • How can we handle troublesome contributors during the moderation strike?
  • How do our moderators respond to the result of negotiations?
  • Is Stack Overflow back to normal moderation?

• Unix & Linux: General moderators' strike

• Ask Ubuntu:

  • Stack Exchange general moderation strike: the effects on Ask Ubuntu
  • I'm going on a general moderation strike due to the new AI-content policy forced upon us by SE
  • Homepage spammed
  • Decline in AU quality
  • Are we there yet? 8 weeks on strike

• Super User: General moderators' strike

• Mathematics:

  • Moderator Strike
  • General Moderation Strike: Mathematics StackExchange moderators are participating
  • Moderation strike update, and flags
  • It is possible to add a banner indicating there is currently a moderation strike?
  • Moderation Strike Resolution

• Server Fault: Labor action notice

• Stack Overflow en español: Huelga de moderación en la red de Stack Exchange

• Database Administrators: Ongoing Moderation Strike

• Cross Validated: June 5 General Moderation Strike: Recap, Possible roles, consequences and other relevant info

• Academia:

  • Academia.SE Moderation Strike
  • Moderation strike (posted to the main site)
  • What is the impact of the moderation strike on Academia.SE?
  • What is the best way for high rep users to support the moderation strike?
  • After a month of moderator strike, is there any update on the situation?
  • Strike stopgap -- will this work, and is it OK?
  • Moderation Strike is Over

• Arqade:

  • Community Moderator Strike
  • Solidarity: what does it mean to us?
  • Community Moderation Strike - The Return of the Arqade Mods

• MathOverflow:

  • Stack Exchange moderation strike
  • Moderation strike update 1
  • Moderation strike update 2
  • Moderation strike update 3
  • Moderation strike update 4
  • Moderation strike update 5 : the results of negotiations
  • Moderation strike update 6 and last : back to work

• Role-Playing Games:

  • Stack Exchange general moderation and curation strike begins today
  • How do our strike-participating community members feel about the moderation-strike negotiation results?
  • What was the impact of the moderation strike on site activity?

• Web Applications: Stack Overflow and Stack Exchange Moderators Strike

• Science Fiction & Fantasy: Sci-Fi moderators on AI content strike

• Aviation: How much is Aviation Stack Exchange involved in the moderation strike?

• Bicycles:

  • Stack Exchange Moderation Strike
  • Moderation Strike Update

• Workplace: No need for drama

• Cryptography: Moderator strike due to SE forcing AI-moderation policy

• Seasoned Advice: Are our mods taking part in the walkout?

• Music: Practice & Theory: General Moderation Strike!

• Astronomy: June 5 Strike - Who, What, Why, and how to join

• Bioinformatics: Notification of Mod action on SE

• Bitcoin: Notice: I’m participating in the Moderation Strike

• Computational Science: Moderation & curation strike: Mistrust, mistreat, and harmful imposed policies

• Retrocomputing: The Stack Exchange data dump has been switched off

• Personal Finance & Money: I will be joining the moderator strike

• Biology: Moderators and community curators are on strike - how will it affect the site?

• Chess: June 5 Strike - Who, What, When, Why, and How to Join

• History: Where is HSE regarding the current strike?

• Writing: June 2023 Writing Challenge: Voices of Protest

• Physical Fitness: The Stack Moderator Strike

• Literature: Special thanks to Mithical for their role in helping to resolve the recent moderation strike

• Medical Sciences: MedicalSciences.SE Moderation Strike

• Arts & Crafts: Community Moderator Strike

• Sustainable Living: Community Moderator Strike

• Programming Language Design & implementation: How is the strike going to impact PLDI?

Honourable mentions:

• Community Building main site: What could be the way out when interests of content curators and company drastically clash over ChatGPT content?
• Law main site: What 'specific legal meaning' does the word "strike" have?
• Puzzling main site:
  • Moderation strike: a tired lunatic (4)
  • Solidarity with the Moderation Team (A Cryptic Division Puzzle)
• Blog posts by users I find notable ^{_{(yes, I know these aren't meta discussions, but this is my post, so too bad hehe)}}:
  • Monica Cellio: Stack Overflow is alienating its community again
  • Jon Ericson: Stack Exchange moderator strike
  • Jon Ericson: Why I'm signing the Stack Overflow strike letter
  • Jon Ericson: How's the Stack Overflow strike going?
  • Jon Ericson: What Stack Overflow is telling employees about the strike
  • Jon Ericson: The data dump diversion
  • Jon Ericson: Why AI is a problem on Stack Overflow
  • Jon Ericson: Stack Overflow's CEO doesn't understand Stack Overflow
  • Jon Ericson: Stack Overflow's strike is over (But its problems persist)

Answer 3

So far, this is the closest thing that we have to an official response from the company, found in an article on Dev Class.

Speaking for myself, this is not a good start.

The trigger for the current crisis was an instruction on Monday last week (a public holiday) to Stack Overflow moderators in an official but private forum, “Moderators were informed, via pinned chat messages in various moderator rooms (not a normal method), to view a post in the Moderator Team that instructed all moderators to stop using AI detectors (as outlined above) in taking moderation actions,” said a post. The details of the instruction are not public. VP of Community Philippe Beaudette posted that “AI-generated content is not being properly identified across the network,” that “the potential for false positives is very high,” and “internal evidence strongly suggests that the overapplication of suspensions for AI-generated content may be turning away a large number of legitimate contributors to the site.” He said moderators had been asked to “apply a very strict standard of evidence to determining whether a post is AI-authored when deciding to suspend a user.” However, the moderators claim that a description of the policy posted by Beaudette “differs greatly from the Teams guidance … which we’re not allowed to publicly share.”

There is no evidence that this is true.

Although there is evidence that some detectors have false positives, this shouldn't be news to moderators and is something that has been discussed. It's why we don't rely exclusively on the detectors, but on other moderator tooling as well as our experience and expertise with the content on each of our communities.

I also don't see evidence that the people posting generated content are legitimate contributors. Of the people that I personally suspended, 1 had previous positive contributions. When someone is suspended, they can also appeal by responding to the moderator message. No one did. I can't speak for all moderators on all sites, but my understanding is that the number of accounts with prior positive contributions and accounts that responded to suspensions is low.

In a statement sent to Dev Class, Stack Overflow’s CEO Prashanth Chandrasekar told us:

“A small number of moderators (11%) across the Stack Overflow network have stopped engaging in several activities, including moderating content. The primary reason for this action is dissatisfaction with our position on detection tools regarding AI-generated content.

“Stack Overflow ran an analysis and the ChatGPT detection tools that moderators were previously using have an alarmingly high rate of false positives. Usage of these tools correlated to a dramatic upswing in suspensions of users with little or no prior content contributions; people with original questions and answers were summarily suspended from participating on the platform. These unnecessary suspensions and their outsize impact on new users run counter to our mission and have a negative impact on our community.

“We stand by our decision to require that moderators stop using the tools previously used. We will continue to look for alternatives and are committed to rapid testing of those tools.

“Our moderators have served this community for many years, and we appreciate their collective decades of service. We are confident that we will find a path forward. We regret that actions have progressed to this point, and the Community Management team is evaluating the current situation as we work hard to stabilize things in the short term,” he added.

The number of moderators may be small, but includes the most engaged moderators across some of the most active sites on the network. It also discounts the impact of deactivating various notification tools and anti-spam bots that users run on new posts to mitigate their impact. The statement trivializes who is participating and what this participation will look like to visitors to the platform.

I don't think that the position on detection tools is the issue. Moderators have known, since the very beginning, about the limitations in the tools used to detect algorithmically-generated content. The tipping point was, from my perspective, how the policy was unveiled. Per the Moderator Agreement, policies are supposed to be reviewed by moderators on the Moderator Team prior to being made public. Although the agreement doesn't say how long the review period is, a day of review that starts on a holiday in the US, Canada, UK, and other countries doesn't seem to be consistent with the spirit of the agreement. In addition, the post on the Moderator Team was a decree or edict and not an opportunity to give feedback. However, this is just one in a trend of announcing fundamental platform and policy changes without appropriate feedback.

Although I don't doubt that, across the network, there were a large number of accounts with few (or no) contributions suspended, this policy is far more forgiving than our policies to deal with spammers. I moderate two smaller sites on the network, where we had less than 10 accounts suspended for posting algorithmically-generated content. Only 1 had meaningful contributions previously. If they had posted spam, the majority of the accounts would have been destroyed, which would have also fed into anti-spam measures. If there was a human behind these accounts, there's a (small) possibility that they have learned a lesson and contributed in the future. However, my suspicion is that these accounts were created only to post generated content and there was minimal loss by suspending these accounts. If the problem was indeed the suspensions, then I know that I would have agreed to end the policy of immediately jumping to 30-day suspensions (which was something promoted by the staff), and I suspect other moderators would have as well.

Personally, it seems like leadership at the company doesn't understand where we're coming from or what we want. And that is the first fundamental step to take.

Answer 4

Power to the striking moderators! 100% support from us users.

A Strike song for some inspiration...

Having said that, I think the demands do not go nearly far enough. Like the opening post here said, this is much larger than the AI decision; it's about the entire relationship of the company to the community and network of sites.

I call on the striking moderators to demand the following:

• Formal declaration by SE Inc. recognizing the Stack Exchange network as a public resource, irrespective of its private ownership of servers, code, databases, copyrights, etc. And of SE Inc.'s relation to the network being foremost, though not exclusively, that of a trustee. (Yes, this should have legal ramifications.)
• Acceptance of a community veto power on network policy changes, the details of which the moderators should flesh out either as part of their demand or to be worked out in negotiations.
• A third-party ombudsman for information disclosure will be appointed, agreed upon by the company and the moderators (mechanism for such agreement to be worked out), who will have full and unrestrained access to all company documents and information, excluding employees' personal affairs and correspondence, and will have the authority and the obligation to disclose all such information which is deemed relevant to the network.
• Accountability for SE Inc. management vis-à-vis the community (collectively or individually)—a formal obligation, via SE Inc. company bylaws/charter or a binding agreement with the delegates of the moderators, to publicly answer, on Meta.SE, collective queries from the moderators, with full and complete answers (with exact mechanism to be worked out by moderators vis-a-vis management).

Whether these demands are a condition for ending your strike or not, that's entirely up to you. But I believe that is the kind of relationship change we need to see on this network (and I may not have even gone far enough).

Answer 5

This is my personal opinion and not necessarily representative of all of the moderators, curators, and users on strike.

The first company response on Meta Stack Exchange was incredibly underwhelming.

As you may be aware, a number of moderators (on Stack Overflow/across the network) have decided to stop engaging in several activities that they had taken on, including moderating content - in fact, almost all moderation tasks. The primary reason for this action is dissatisfaction with our position on detection tools regarding AI-generated content, and discontent with how that was rolled out.

This is an accurate assessment. Personally, I'm more discontented with how the position/policy was rolled out. Having an appropriate discussion with moderators and then the community, as it's laid out in the Moderator Agreement, would have been better, even if the result was the same policy.

We ran an analysis and the ChatGPT detection tools have an alarmingly highly high rate of false positives, which is correlated to a dramatic upswing in suspensions of users with little or no prior content contributions.

As I, and other moderators, have said before, we do know that these tools aren't perfect. When the policy was handed down to moderators, there was an assertion of testing done on Stack Exchange data regarding the false positives in these detectors. However, no specific information was shared, although the moderators are expecting it to be provided in the near future. The tools were only a single tool used to detect algorithmically-generated post and I'm not aware of any moderators who relied on a tool to tell them that a post was generated.

When it comes to users with little or no prior content and posting content that is not permitted, suspension is letting them off easy. When garbage or spam is posted by accounts with no other valuable contributions, we destroy those accounts. In fact, there are destruction reasons for posting "spam or nonsense" without any other "positive participation" and for cases where the vast majority of content violates the terms of service. When we take these actions, not only is the account destroyed, but it feeds into

People with original questions and answers were summarily suspended from participating on the platform. We stand by our decision to require that moderators stop using that tool. We will note that it appears to be very rare, however, for mods to use ONLY the ChatGPT detection tool, and frequently their own analyses were in use as well. We will continue to look for other, more reasonable tools and are committed to rapid testing of those tools and any suggested heuristic indicators.

There is a public admission that moderators used their own analysis. We also used other moderator tools and, in some cases, engaged with moderators from across the network, prior to taking action. Yet the response is to remove a valuable tool from the toolbox. Although there is a false positive rate, the analysis that was shared puts the false positive rate into what I would suspect would be a very tolerable number when combined with other tools, including other detectors.

The moderators who are engaged in this action served this community collectively for many years on the platform. Personally, I consider a number of them friends, and anytime friendship is tested like this, it’s difficult. I would like to say to them clearly that I hope they know how much I, and the whole staff and community, appreciate their collective decades of service to this community, and I hope that we are able to come to a path forward. I regret that actions have progressed to this point. The Community Management team is evaluating the current situation and we’re working hard to stabilize things in the short term. We’ll be working collaboratively with the community on the long-term solution.

This is not the first time that we've been here. But lessons from the past don't appear to have been learned.

I’ll be honest, the next few days and weeks might be a bit bumpy. Both sides share a deep and unchanged commitment to quality on this platform. Additionally, it’s important that we guarantee that anyone who has meaningful contributions to make has the opportunity to do so. As we have updates I and the team will be sharing them here.

I - and the moderators that I'm in touch with - are also very committed to the quality of content on the network. We do hope this resolves quickly, but at the same time, want the resolution that's best for our communities.

Answer 6

Mainly focussing on the title, with the statement that "Stack Overflow, Inc. cannot consistently ignore, mistreat, and malign its volunteers":

That's wrong. Of course they can. Obviously.

The most active sites already have been flooded with an unmanageable amount of low-quality content for years. Now, there's a new source of low-quality content. And... why exactly should they care about that?

They don't.

There will always be new people to join the site. Some of them will become volunteers who work off the crap-queues, for a few internet points and badges (Gamification 101). Some of them will become moderators, because ... oh, what a privilege and honor. Some of them will be disappointed with the increasing workload, the decreasing quality, and the ... $%!$§ ... behavior of the company. Some of them will complain. Some of them will leave. Some of them may initiate a strike *shrugs* - that's just noise. New users will come. New volunteers will come. New moderators will come.

They don't care.

I'm aware that this does not sound constructive. But I think that it is important to know that the company does not care about you at all. People who joined after the previous management debacles may not be aware of that. They may have joined with the same idealism and drive for altruism that I had ~10 years ago. And they should be prepared to be deeply, deeply disappointed.

They will sit that strike out.

Answer 7

Reveal to the community the internal AI policy given directly to moderators. The fact that you have made one point in private, and one in public, which differ so significantly has put the moderators in an impossible situation, and made them targets for being accused of being unreasonable, and exaggerating the effect of the new policy. Stack Exchange, Inc. has done the moderators harm by the way this was handled. The company needs to admit to their mistake and be open about this.

I agree that it would absolutely be in the community's interests to be able to read this secret policy. It is completely unreasonable for us to be subject to new rules which can't even be said in public.

If Stack Exchange, Inc. won't allow the policy to be published here, there are other venues which might consider publishing it. (These outlets have written about Stack Overflow's policy on AI content before, so they might run further stories on the same topic.)

To me, it seems inevitable that the secret policy will become available one way or another, so it's just better for everyone if Stack Exchange, Inc. publishes it themselves.

Answer 8

This is peripheral to the question, but pertinent:

I'm perplexed as to why the company is seeking to destroy itself in this manner. While a useful amount of AI output is reasonably accurate and useful, a significant portion is either very misleading or entirely fabricated.

ChatGPT will:

• Produce totally fake references,
• Produce wholly non-existent web links.
• Make comments which, when challenged, it will disclaim and say something entirely different.
• In some cases contradict itself in the same answer.

The only way to deal with these issues is to either have an expert knowledge greater than the level in the answer, or to use it only as a source for cross referencing.

• By allowing essentially uncontrolled use the company will destroy the integrity of its core asset.

A relevant example: I am not a Quora member, but I read a significant number of their answers in my (relatively wide) areas of interest. I have been surprised to see some Quora members, obviously using ChatGPT for essentially their whole answers, producing many factually incorrect and hardly relevant answers, but receiving numerous upvotes from other members.

While Quora still contains a very large percentage of genuine and highly useful human-generated answers, it is obvious that they have not chosen to actively oppose the posting of complete rubbish to their site. The likely outcome seems clear. Why Stack Exchange inc would wish to follow the same path is puzzling.

---

Added:

The company may be inadvertently "throwing the baby out with the bathwater". The proposed use of AI to improve asked questions has SOME prospect of being useful.

Experienced users can ignore the advice (as long as it is only advice) and many beginners would benefit from good advice.

Worst case, it may make questions worse in some cases (but possibly not too many) and bad questions can be and presently are treated badly. This would be a negative for newcomers, but it could be ameliorated by thoughtful community action if members are prepared to actively help newcomers. Based on present practices, this seems unlikely; I perceive the site and the majority of member actions creating a substantial barrier for new members. Improving this is hard.

Overall, AI may quite possibly improve questions on average. This impression may affect the company's perception of AI answer improvement, which is far less likely to be net-positive overall.

One area where I have reservations is the suggested use of AI generated code to improve questions. This is very much a "double edged sword" as AI code is often good, but may contain subtle errors leading to 'rubbish out'. If new user code related questions suddenly sprout AI-generated code examples, the quality would need to be extremely high on average to not produce negative reactions overall from the community.

Answer 9

I wrote a user style to hide the review queue, downvote, VTC, VTD, delete, undelete, and flag buttons/links.

Answer 10

The purpose of this answer is to provide an indication of the financial motivation for SE's decisions. In all humility I would suggest that one cannot assign motive for any particular actions without following the money. In that vein, a good starting point may be to see who the investors in SE were in the latest Series D startup financial round. A Series D startup is generally either one which has not achieved its objectives in rounds A, B and C, or one that is designed to prepare a company for going public. Said information is available crunchbase's stack-exchange financial base. These include and can be followed up for further investigation by following the link above and links to those investors. More complete information is on pitchbook SO.

It may have made more sense to offer SE's users and moderators the ability to invest in SE in a funding round than to do so seeking investment from those with little or no understanding of what they purchased, and who come with intellectual baggage or conflicted interests and may or may not wish to dismantle SE or transform it to be unrecognizable. Such is the behaviour of the nouveau riche who for good reason are oft regarded as uncouth.

Rather than lamenting an opportunity lost, perhaps investigation of a means of having SE's users buy out SE should be explored, as management currently is inept enough that that may be its most realistic option.

Unfortunately SO (and SE) was bought out by Prosus for $1.8 billion. As Prosus is an ESG activist investment firm, it has the power to distort market forces by propagandizing, censoring, and proselytizing Neo-Marxist ideation. For example, it is a major shareholder in Tencent, a Chinese information firm.

Edit BTW, I'm poisoned, I have almost completely stopped reviewing, posting, voting, etc. Stuff like this doesn't help.

Answer 11

As someone who has been working on programing and data analysis on computers since 1981, I am very concerned about allowing people to submit answers that are not a product of their own minds but a product of an artificial intelligence that has simply collected a series individual 'facts', made available by persons unknown, which leads to a conclusion which may (or may not) be plausible.

The same applies to an artificial intelligence (AI) tools checking whether artificial intelligence has created an answer.

That the moderators are not using these tools is, to me at least, obvious, based on their actions that I have observed in the last 6 months.

I myself have flagged/voted many answers that due the style of writing, lack of sources and often very weird constellations of 'facts' making it obvious to me who has the required background knowledge in the topic of the question.

The last 2 weeks the user Serg Z

• 2023-07-22: User spamming chatbot answers - Mathematics Meta Stack Exchange

on Law and History Stack Exchange, has flooded many existing questions with junk answers.

How should a later reader determine which answer is based on knowledge and facts and which are simply a collection of random facts pasted together by a AI and submitted by a user as if it was their answer?

---

Below is a email I sent to three friends today.

• a Professor of Philosophy from Vancouver, Canada
  • who is sympathetic to the idea of a central source of knowledge
• a US programmer since the mid 1960's for supply automation in the car industry and later debit bookings
• the Italian author of a well known PC based Spatial software who I have worked with for many years

I have added this mainly for the benefit of the business leadership of Stack Exchange, Inc. so that can, hopefully, better understand from the viewpoint of a data analysis on computers of many decades that the AI policy should be to treat such answers as a form of plagiarism (and therefore be deleted), since the answer is not a product of the author that submitted the answer.

The ChatGPT experience was from December 2022, where it became clear that the system (at least then) was designed to give an answer at all costs not caring if the result was nonsense or not. This is copy and paste knowledge at its worst.

---

I was taught that one 'had to learn how to learn'.

One lesson was that one had to have enough background knowledge to determine if any found/returned result was plausible (part of the 'learning' process).

Next, check the sources from preferably multiple -independent- reliable sources.

Finally assume that any result might not be conclusive and that multiple possibilities may exist (often dependent on certain situations).

In 1981, long before the first MS-DOS for PC's were available here, the first lesson in the programming course was: A program will always be as stupid as the person that programmed it (you can't program a bookkeeping program if you know nothing about bookkeeping). Lesson #2: the main difference between a stupid program and a stupid person is that the stupid program is more capable in making more mistakes faster.

---

Tell this to some people today and you can see the question in their eyes: 'From what alternative universe (far, far, away) does this person come from?'.

This is what is meant with 'a critical lack of tech skills' in the article below.

In the report the tech giant [Google] also highlights what it calls "a critical lack of tech skills in the UK", which - if unaddressed - will "remain a stubborn barrier to equitable nation-wide growth, especially as demand for AI and other tech expertise soars".

• 2023-07-28: BBC News - Google: Don't trust what our chatbot Bard tells you

A few years ago, I saw a quote:

With the internet making information available to everyone will make them smarter

• well, we got that one wrong didn't we?

As far as I am concerned, for the likes of ChatGPT and Bard, the first two lessons from 1981 still apply.

---

On History Stack Exchange forum there was a question about a ChatGPT result that the person couldn't make sense of:

• 2022-12-25: Was there a Moresnet Conference of 1867?

So I tried the ChatGPT system for the first (and only) time to see if I could figure out how ChatGPT came to the result that the OP of the question reported.

My conclusion was that ChatGPT was designed never to say 'I don't know' or 'insufficient results to come to a (sane) conclusion'.

It collected 3 facts that were true for a 3 month period in 1867:

• 1815-12: Dutch and Prussian representatives created Neutral Moresnet
  • since 1830 Belgium instead of the Netherlands
  • existed until 1914
• 1867-04/06 : International Monetary Conference of 1867 in Paris (Latin Monetary Union)
• 1867-03-23 / 1867-05-11 Luxembourg Crisis

Thus the 'Moresnet Conference of 1867' (with the participation of The Netherlands, Prussia and Luxembourg - who never belonged to the Latin Monetary Union - and Luxembourg that had nothing to do with Neutral Moresnet) was simply 'created' as a fact, with no supporting sources, hoping that nobody would notice that the answer is nonsense.

For the gullible, ChatGPT and Co. are a dangerous source of information.

Answer 12

If I visit the open letter website it states that the strike is finally over:

I want to know if this is an official statement or not.

Thanks beforehand.

Answer 13

It would make sense for a company promoting and selling new technology to buy up a company selling old technology and then mothball its old-technology operations.

Meanwhile it would make sense for a company promoting and selling old technology to get with it and maximise its selling price with the above in mind.

"Artificial intelligence" is about solving issues - perhaps even issues you didn't know existed - by running computer programs. OpenAI is a leader in the field.

SE/SO is the world's leading site featuring expert questions and answers that are written by human intelligences. Go figure.

Edit: it seems from comments that Prosus, the company that already owns SO/SE, already knows a fair bit about AI, so perhaps SO/SE is not about to be sold. As user @leanne writes, it seems the writing is on the wall. Online expert question and answer provision based on human intelligence is about to get cut back drastically.

*Another edit: OK, downvoters, the decision is nothing to do with profit and AI-isation but derives merely from ignorance at the top of the company of what the company's unpaid volunteers know. Seriously how likely does that sound?

Answer 14

From a longtime SO/SE user: I do not support this strike.

A company owns this collection of sites. A corporation has the responsibility, and right, to make their own decisions about their properties' functions and appearance. They are not beholden to their volunteers regarding their decisions.

I have seen complaints about "firing" a moderator, the appearance of voting buttons on the sites, and now the disallowance of removal (except in very narrow circumstances) of perceived AI-related posts.

As with any company, I'm sure that Stack Exchange Inc's management has made decisions based on input and findings by their employees as to the best way to proceed.

Just because some volunteers don't like some decisions, and/or the way they were communicated, does not mean the company has to do what the volunteers want done. The company must do what they think is best for their bottom line, including what they think is best for their users.

I personally, and most people I know, have stopped using SO for any new questions because of the combative way we have been treated by some volunteers who believe they know exactly what's right and wrong in every case; who downvote and close questions even when they don't have experience with what's being asked. They're so professional and knowledgeable that they can somehow magically determine whether a question or answer is valid without any experience in the topic.

There are a bunch of companies out in the world, including the creators of ChatGPT, who are trying to find a way to determine whether something was created by AI or a human being. If all of those people can't even determine 100% if something was created by AI, then what extraordinary ability do some SE volunteers have to determine, without a doubt, that a post's information was provided by AI?

OMG: instead of volunteering, you should be getting paid the big bucks for this incredible skill!!!

---

Update:

After doing some research on Prosus and checking out the SO blog entry from April 2023, Is this the AI renaissance? (Ep. 564): I can say that

• Prosus has known about AI and its workings for years.

• They have their own AI team.

• They have purchased Stack Exchange, Udemy, Codecademy, and other learning sites.

Based on this information, I posit that "the writing's on the wall": Prosus is going to be using these sites' data for its own search/learning system that they can possibly hugely monetize. Soon, no more "volunteers" needed...

And, if international corporations, like US corporations, are beholden to their shareholders... well, that's why they seem to not care about their volunteers.

---

Update to clarify my update:

For those who might not understand what I'm saying about shareholders - Prosus, owner of the SE/SO collection of sites, is a public company. "Public" means that a company, such as Prosus, provides stocks, aka, a portion of the company that can be purchased by the public or provided to their employees as compensation. The people who own these stocks are called "stockholders" or "shareholders".

Although these companies are not actually required by law to maximize profit for their shareholders, they often do exactly that - which is why I say "beholden to their shareholders".

What I'm saying with the earlier update is that Prosus can, and likely will, find the optimal way to increase their profits in order to satisfy their shareholders. That may include using the purchased assets, such as SE/SO data, to train AI to teach people to code and to answer their questions - which may or may not make the SE/SO (SO, particularly) sites OBE (no longer necessary).

This has nothing whatsoever to do with my non-support of this strike - the reasons for which are in the portion of my post prior to the updates. The updates are merely further information around why the company might not be interested in yielding to the strikers' demands.