Is it just me, or does everyone has trouble reading uppercase posts, I cannot understand posts that are uppercase, while the exact same "post" lowercase is perfectly fine.
cc @art I think the way to deal with smokey abuse is make everyone that gains SD privs accept explicitly and advise anyone else that they have to use their own judgement. Failing that will be harsh consequences for the user, beyond a declined spam flag
Ideas: - "Needs admin evaluation" SD command ([link](http://chat.stackexchange.com/transcript/message/28857404#28857404)) - Make purposely flagging posts without reading bannable ([link](http://chat.stackoverflow.com/transcript/message/29834183#29834183)) - Format SD reports as a "Review request" ([link](http://chat.stackoverflow.com/transcript/message/29834052#29834052))
The first in particular is something we can put into action now; it needs a single wiki page on Smokey's repo that we can point to when people ask for privs and ask them to agree before we add them.
Don't take it personally Undo. I have observed mob flagging in the room. I asked if an answer was spam and it was immediately manually reported and the user got -100. We had to get Jon involved. It wasn't a good day.
I also think the needs-admin command could be a good idea, as long as it isn't over-used. The borderline posts, and any errors, can just be sent to MS through Smokey for an admin to pick up on later.
@JAL We can't tell if people read the post or not, obviously, but if people are spam-flagging obvious not-spam, then that should be a bannable offence.
That's the proposal, anyway. I don't know what I think about it.
@ArtOfCode We can do that by requiring people to go through a meta smoke page designed to the the user, and in the case of a audit, we redirect different users to different posts, and then monitor the votes on those posts
@Cerbrus Let me be clear: I don't plan on having any "audits" for Smokey. It's not really the best experience... and it starts turning the bot into a platform with meta-discussion and meta-meta-discussion and so on. :P
Another way would be to search and monitor what get spam nuked by flagging from users (in short time frame), potential errors, to really understand if something bad is going on. (since I continue to not worry about a single wrong flag, it will get denied by mod and users learns)
For example if we have list of stuff where post got spam nuked and one sd priv did sd f it, that would be interesting.
I think that goes for other things, too. Let's not get too hung up on meta stuff... the purpose is to identify spam, not to identify people who can't identify spam. ;)
The only ones that come up are people who had a fit of rage with nonsense answers (abusive tp's), false positives marked as tp or people who used excessive self promotion but were otherwise legit users
Actual spammers rarely post again on the same site, because they get blocked when the previous post is spam nuked. If a user was able to post again, then they are probably not a spammer.
So there is an observation bias here. Most are blacklisted correctly, but the effect is never observed.
MS-only currently; if we want a Smokey command that's doable but it probably won't be me that does it. I'm not familiar enough with Smokey's architecture.
@Undo didn't add tests, I just realised. Oh well, they can be in the next one.
Also, can somebody add me to the Smokey repo write list? I don't expect to do masses with it, but it'd be useful to be able to commit and pull to Smokey straight off for those bits I do.
"If it's something we would have wanted Smokey to catch were it was implemented as a block at the system level, it's a true positive" (just me thinking out loud)
as you will see he tell us that rule for abusive or spam flag is that user needs 1 rep otherwise we need to consider it an error (cat on keyboard) and only go for NAA
if it's a new user, it doesn't matter (and spam is plausible). If it's someone who has reasonable posts elsewhere, then that spam flag will be declined because it'd mess them up for what was probably an innocent mistake.
this is sane
except for the 'decline' part, depending on who gets to it.
@Undo after this discussion (and some denied spam flags) there was the meta post you answered (where the mods where some what disagreeing), but in the end you answered that we should flag abusive instead of spam, I guess you remember.
And then we're trying to get users to read instructions on how to post... when they don't even notice if the site is in Russian or English. Today had a post in Russian on SO and a post in English on ru.SO
@ArtOfCode use the SE API to login. You can use an empty scope. Then users don't have to remember a user/pass combo and you can easily tie information in MS to a user on SE
I would quite like to fix the issue that feedback from chat and feedback from review are considered two separate things, even if they're from the same user...
To see all my feedback, I have to look at the histories of two users, ArtOfCode and my email address.
