@Undo At the same time, please don't underestimate the potential value smaller sites assign to new users. I don't know terdon's site well, but cooking has... seven autoflags ever? We would absolutely be willing to handle all of those ourselves in order to get one new user.
Searching this room for cooking.stackexchange.com returns over 300 results... not all of them are reports about low-scoring potential spam posts ... and many are probably false... but I'm guessing that many are true.
@Mithrandir I see. Still, if there are somewhere between 7 and 300 posts that might get automatic comments from someone, and those comments are less likely to work out for a new user, there is a point at which we would rather handle the spam ourselves.
'Automatic' comments are still manual. Some folks leave templated comments when they feel like it - this is something to be pursued with those people, not with the project as a whole.
While I understand that individual users are ultimately going to do what they want, I think that if you're handing them a list of things to maybe act on, the project as a whole should still consider what the results of those actions are.
Review queues are perhaps a decent comparison: there's a reputation threshold, there are potentially audits, and users can be banned from them.
Anyways, I appreciate the idea of detection and letting users manually decide about flagging/commenting etc, just interested in seeing a little investment in making sure that human element works out as reliably as possible.
Probably the best I can do it throw up pinned messages occasionally about comments. It's hard to justify that one for a single case, especially when I can't enforce it, but we'll think about it.
For whatever it's worth: I agree with Andy on this one...
this is not about Smokey and spamdetection, this is about an isolated incident with a "pro-forma" comment that was formulated stronger than one moderator liked it
@Undo At some level, we ultimately answer to ourselves - we could, technically, move to Slack and keep doing what we do without any oversight - but I don't think anyone wants to separate from the community like that.
@Undo Well, when you offer a service, your customers are who you answer to... In some sense, the mods are the voice of those customers - the network sites - and, in so much, they have the right to ask for changes to be made.
That said, even when we request things from the CMs, we have to go through something of a meta process... meaning there's discussion, looking at data, and a decision is made. It's not made on the spot by a single person.
@Undo 1 out of 20+ mods can't reasonably be said to speak for the site. It wouldn't be the best thing to do for the site to kill it for everyone because one moderator said to do it.
@Catija In this case, though, the 'customers' have their own voice. At what point do we make a call that a moderator isn't the final authority on a site?
I think it's fair to say "we can do that but it will take us some work, so we'd like for you to talk to your community and we can discuss it and show you what might be the result of this being turned off for your community".
If a moderator comes to us and says "hey, we're a new health site, probably be a good idea to add it to the big list of health sites for the viagra filters" - awesome! But I'd take that advice from anyone, and evaluate it on its own whether it's from a moderator or not.
@Undo multiple of the above. We ourselves are the authority on our own systems (i.e. how they work, what we can/will or can't/won't do) - nobody else can be.
Right... which is probably where I think that some rule codifying is necessary... in the case that there's a stop work request... I think it's better to go with meta post > shutoff...
But at some level, we have to respect communities. At that level, we don't answer to mods - we answer to communities. Mods are the representatives of communities, but they're not dictators - coming to us and saying "do this thing" has no more weight than any individual community member as far as I'm concerned.
@Catija In the instance above, though, do you think we'd get a post with a fair representation of what we do if it was a "do this first"?
It'd be easy to throw a post with a giant red "1" and a black "0" and say "the one is the number of false positives, the 0 is the number of times it helped". You'd get a pretty predictable reaction from that.
It'd also be easy to throw out one with network stats, explanation of basic statistics for "sample size of one is not representative", etc. It's a complex topic.
And of course we answer to CMs. There's also no question about that. We remain the authority on nuanced things about the system (like whether we can, say, pin a reason weight for a specific site), but CMs are the final on whether we run overall or not.
The complexity of the topic means it's easy to lead people on meta to whatever you want them to think. We have all the context and can recognize when someone is doing that. Average meta people don't.
@Catija I don't think there will be any of these cases that aren't complicated.
I think that a quick, unequivocal "one bad autocomment almost cost us a valuable member of the community" is a training problem... and that starts making you think of how to better train the people responding to smokey reports.
@Undo Absolutely not. The comments could have been utterly unrelated to Charcoal at all... anyone could have posted that comment. But we do, on the charcoal site link to that specific autocomment... so if we're supporting it and it's causing problems, we should rewrite it.
Can we vary the number of auti flags per site? 5 on the big spam sites, 3 on the rest? or put a meta post giving communities the choice to opt out of 5 auto flags?
@Catija This... is it. Maybe we do need to think slightly more about 'training'. Charcoal is bigger than it used to be, and it's expanding faster than it used to. Our current guides are more designed for a new person who comes along and can spend a week talking to people and working out the system. That's still kinda possible, but maybe we should have things that are more comprehensive
@Undo Charcoal wasn't mentioned in the comments. The user and the site have no indication that we were involved at all. It was a comment saying that they need to disclose their own stuff. We were pointed as the reason because the post was picked up by smokey and a user that frequents this room posted the comment.
