This update is to point to two posts by the company on the platform-wide meta site, Meta Stack Exchange.
- The post (Historical) Policy on the use of GPT Generators is a copy of the private policy on generative AI content that was given to moderators in moderator-only space. This policy no longer holds (so I'm posting the long disclaimer at the start). But the part that essentially triggered the strike was the part after the cut: no action could be taken by a moderator on content they judged to be AI-generated, unless the user volunteered the information "I used AI to make this" without prompting or coercion (or was silly enough to not cut out words like "Regenerate answer" that flagged it as ChatGPT output). To do so would have been to violate the moderator agreement that elected mods all sign. If a user had outright continued to lie and say they wrote it, even with the kind of stylistic flags that it was chatGPT (say), and the kind of non-human errors GPTs make, flags had to be ignored, and no action taken.
The policy below was posted on the Mod Team on May 29, 2023 to update moderators as to changes in how the GPT policy should be enforced. A day later, a shortened version of this policy was published on MSE. We normally do not release enforcement guidelines publicly. In recognition of the confusion caused by the differences between the public and private guidance in this case, we are now publishing the private text of this policy in full. However, we reserve the right to continue in the future to withhold specific enforcement guidelines from being published publicly in cases where doing so would make the jobs of moderators more challenging, and open up new avenues for abuse.
This is being released now so that it can be included in the public record. Since being published, this policy has been superseded, with the new policy available shortly. The historical policy begins with the line below this one.
May 29, 2023
[...]
Therefore, the only admissible evidence we can currently permit for GPT usage is self-admission by the author of the posts, freely given. Please note that “freely given” is important here: please do not, under any circumstances, try to trick users into admitting GPT usage, lure them into saying it, or otherwise coerce a response. Even a user saying they have used GPT in general may not count unless they specifically say they have used it here, or for this contribution.
The public position the company held (after the above) can be read in this massively downvoted post, where the top comment is
This post does not match the guidance given to moderators.
The omitted part in the big quote above outlines the company's justification, but it is based on what many have observed is flawed data analysis using a metric not verified as fit-for-purposes. Another part, dealing with the geographic distribution of genAI suspensions, is summarised by another commenter as
I can see why the moderators went on strike: this policy stops just short of openly accusing them of being racist robots blindly applying the results of GPT detectors.
To reiterate: mods were not just relying on so-called AI-detection software, but their own judgement, since it was clear early on that AI-detection software is about as accurate as generative AI is at answering questions. (this was largely a problem on the big coding/tech sites, we here on MO didn't even get to the point of even having to think about this, or dealing with scores of users flagging posts based on AI-detection scores of posts they tested)
- The company has openly committed to continuing the API, SEDE and the regular data dumps, in this post: The company's commitment to the data dumps, the API, and SEDE
The company is committed to the long-term (foreseeable future) survival of the data dumps, the API, and SEDE. We will continue to maintain them, and assure that community members have free access to them for legitimate usages that support the community, broadly construed (including for study in classrooms, for instance). We retain the right to place guardrails around them to ensure that companies constructing language models, etc, are charged for access, but community users, including the Charcoal anti-spam network and similar projects will continue to have free access.
This is good, and makes public what was until now a commitment, but not in this form, in writing, and public.
Added the company employee tasked with turning off the data dumps opened up and explained how it went down, giving direct quotes from the CEO's direct request to do so: https://meta.stackexchange.com/a/391640/226541
