Man, what a mess.
I wasn't sure if making this public was necessary, but yesterday and today (of all times?) a relatively high amount of spam seems to have been introduced to the main site, so I want to clarify my current (dis)position.
Because of the transgressions of Stack Exchange Inc. outlined below, I have signed the open letter to Stack Exchange Inc. and am currently on strike (which, naturally, opens up its own can of worms) since last Monday.
This means I will not handle flags or go through Review Queues or delete spam.
AI-generated content is the network's Hot Topic of the Year. I will post a simplified timeline of how it has been dealt with this far, and link to the relevant posts on Meta.
At the end of last year, 2022, AI-generated (ChatGPT) content became highly problematic, especially on Stack Overflow, the largest site on the network, and got (temporarily) banned there. Hence this content was (and is, by the community, at least) very much frowned upon.
Nevertheless, SE decided to let all its separate communities deal with it in their own way: "while we evaluate, we hope that folks on network sites feel comfortable establishing per-site policies responsive to their communities’ needs" and "we encourage sites to create these policies as [the different degrees to which sites are impacted by the usage of ChatGPT] become an issue". And so many sites on the network created their own policy.
A few concepts on how to use AI to the network's benefit were posted on Meta:
About a week ago (May 29th), SE staff posted a new policy regarding the use of AI-generated content detectors on the private Stack Overflow Teams for Stack Moderators: it is forbidden for moderators to use "GPT detectors" or AI-detectors to, well, detect AI-generated content, and act on it. "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", as this post puts it neatly.
After a lot of negative feedback on that post, a similar (and similarly misleading), but fundamentally different policy change was posted on Meta here.
There were several things wrong with this approach:
- This policy was introduced privately.
- It failed to communicate a few of the details that had been posted privately before.
- This policy was introduced without prior introduction, which is a breach of their policy.
- The policy was accusatory of moderators across the platform, without any evidence to back it up.
It mentioned how there has been an "overapplication of suspensions for AI-generated content", and how moderators should be wary of suspending users.
I'm not entirely sure whether this is falsely implying that moderators have thrown the rules they signed out of the window whenever these detector's analyses have been positive (see also here and here), but the ChatGPT policy on Stack Overflow still mentions how "moderators are empowered (at their discretion) to issue immediate suspensions of up to 30 days to users who are copying and pasting GPT content onto the site, with or without prior notice or warning." So if moderators acted on these detectors this way, at least on SO they were (and are) completely in their right.
- It mentions analyses performed by SE Inc. that backs up their policy change, but none of the numbers or methods have been made public.
Indirectly related issues that have popped up:
- Something that has been an issue for a longer time, but resurfaced with this affair, is the apparent unwillingness of SE Inc. to listen to the very feedback they solicit (e.g. here).
- SE Inc. has unfeatured posts about the strike, which goes against SE policy.
- (Note also that the shadow of the Monica debacle is still to pass over, and the trust that SE Inc. gives us on paper has not at all been felt since before that whole affair.)
- Here is a response to the strike from SE staff.