Skip to main content
Meta Stack Exchange

Return to Answer

added 2042 characters in body
Source Link
Thomas Owens
Thomas Owens
  • 58.4k
  • 19
  • 109
  • 201

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.

Speaking for myself, this is not a good start.

Speaking for myself, this is not a good start.

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.

Source Link
Thomas Owens
Thomas Owens
  • 58.4k
  • 19
  • 109
  • 201

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

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.

Speaking for myself, this is not a good start.

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.