« first day (1185 days earlier)      last day (917 days later) » 

8:00 PM
@Undo should be aware of this ^
 
Horribly low effort and off topic, but an actual question
 
How'd it come to your attention that it was false? System keep track of votes it casts and monitor the eventual outcome of the posts automatically?
 
Manual feedback
 
@Pops 99% of times, there's people in this room
We manual feedback on every single report
 
@Pops the feedback on it (those two red crosses) were put there by humans, who also happened to ping me about it
 
8:01 PM
Every report gets at least one qualified feedback, either through this room or metasmoke review
 
i.e. I knew about that having happened within about two minutes of it happening
 
Was on a call, reading transcript
heh, that's awesome
Gotta remember to leave it at one on that oddball site where they cranked down the 6-flag threshold
 
@Undo neither of us ever built a UI for per-site max flags, so you'll need to enable it manually with a console. Beware of Workplace and ELU.
 
Are those the only two?
 
AFAIK
 
8:08 PM
Morning everyone
 
Okay, set to 3 on all sites except ELL, Workplace, and meta sites
 
And hi @Pops, great to see that you've popped in again
 
So we should be only constrained by the global setting now
 
@angussidney evening :)
 
Thanks Pops, great to make progress on this! :)
@ArtOfCode Whatdya think of defining a simple formula for max flags for a post? Something as simple as floor(reason_weight/100) might be plenty
 
8:10 PM
Hi angussidney, Undo, Ferrybig.
 
@Undo ya know, I was thinking this earlier. I'd be happy with a generic round instead of floor, actually
 
@Undo imo not worth it
 
@ArtOfCode Only reason I floored it is that 180-type things aren't as awesome, overall
where's my pretty graph
 
Our FP rates are so low that whether we 3 flag 99,75% or 99,98% posts is largely same in effect
 
@Undo I'm thinking that given the obviousness of our disclaimers, I'm happy pushing the max up without making everyone set themselves up again. Might do a mass ping, though. Thoughts?
 
8:12 PM
@Undo If we have a metric thats unreliable enough to not want to trust it with max_flags then we shouldn't have it autoflagged at all
 
@ArtOfCode Could be done, we have chat IDs for everyone anyway
 
@Undo was just gonna look at the list of names and ping 'em all manually, actually :)
 
yeah, that'd be the hard way
 
(Imo)
 
ap User.where(:flags_enabled => true).pluck(:stackexchange_chat_id).map { |i| "@@#{i}" }.join(" ") would be the easy way
 
8:14 PM
@Undo there's only 22 of 'em :) but if you're happy with not making everyone set up again, then I say we do it
@Undo but that would be sensible
 
Good to push the global max up, definitely
 
@Undo Yes, just those two, both of which are set to nuke at three instead of six.
 
@Undo python hax
 
@Pops Great, thanks. We'll stay at 1 on those.
We should be able to stay on top of future changes like that, too, if they're announced on meta somewhere with the [spam] tag which they usually are
 
no doubt
 
8:16 PM
@ArtOfCode See graph above, the one that takes a little reading. There's a pretty significant accuracy bump at 200 and at 250
Maybe floor((reason_weight - 150)/50)?
I think that gets what we want
 
I love how your graph bottoms out at 99%
 
@Undo 180 results in 0
I'd put a min 1 on it
 
oh, right
floor((reason_weight - 100)/50)
 
maybe ceil instead of floor?
 
I question why we're mathing out like this at all
Whats the purpose behind it
 
8:18 PM
@NobodyNada same effect, yeah
 
autoflagged FP will require our attention in any case
 
@Undo works
 
Whats the difference if it's 1 auto flag or 3?
 
@Magisch we've gotta ping 3 people to retract flags instead of 1
 
@Magisch For now? It's probably not necessary. But it'd be nice to have the system in place, and if we're going to build it we might as well test it.
 
8:19 PM
@Undo It results in vastly fewer flags cast though
So it's partially getting rid of the system
 
@Undo should be EL&U, not ELL.
 
Your current calculation would make a whole bunch of confirmed spam stay at 1 autoflag
For what?
 
@Glorfindel yeah, english.SE. Might have mispoke there, but it's ELU in the database
@Magisch metasmoke.erwaysoftware.com/flagging/logs refutes that, mostly
 
@Undo FR: Make it not pick the same user twice to flag
Like it did just now
 
8:20 PM
@Magisch huh?
that shouldn't ever happen ever because math
 
It tried to use 2 of MAR's flags on the last post
:p
 
those are different posts
 
different posts with the same title and the same reason weights
I'm a failure
:/
 
yeah, threw me for a loop too.
 
@Undo How? I have my weight at 180
 
8:21 PM
(afk for a while)
 
Should probably let mods know this is happening...
yeah, gonna post a thing in TL
 
@ArtOfCode I'm in - no need to ping me :)
 
New metasmoke user 'caffeineaddiction' created
 
That doesn't look good
 
8:30 PM
that's a fun one
 
I broke all the things
 
I thought we weren't doing that anymore
Ohhh
@CaffeineAddiction Happen to not have an mSE chat account?
 

 Tavern on the Meta

MY GOD, IT'S FULL OF STARS AND BACKLOG! General friendly chit-...
 
@ArtOfCode Meta post when?
 
8:31 PM
I have a meta account on security.se
 
@Glorfindel @Andy @NobodyNada @Kyll @rene @Floern @angus @Magisch @tripleee @Avery @Ashish @M.A.R. @Mith @Cerb @ThomasWard @Yvette @doru @KazWolfe @Byte @Mego @JF @Ferry: FYI, max flags is now 3 per post. Heads-up, in case you want to opt out at this point.
@Magisch not meta post yet, just telling mods.
 
you did that manually, didn't you? :P
 
@Undo yep, I don't have a console :P
 
"max flags"?
 
@ArtOfCode should probably start composing it soon, though
 
8:32 PM
@Magisch why?
 
@Kyll we can now cast up to 3 flags per post automatically
 
at some point the public deserves to know we're doing this
 
Woaah
Nice
 
ok, so is my account active at this point?
or is it broke?
 
If we have CM approval, I really don't have a reason to go fight the meta war
deploy pending
 
8:32 PM
@CaffeineAddiction "active" as in what
it'll let you into review
 
@Undo Not a war
We just need to inform people
 
I dont know, i just got that error ...
 
I mean we're doing this with sites where the people are the moderation force
 
@ArtOfCode so i can use the ! commands for SD?
 
@Magisch It wouldn't end that way. For now, we don't need to go to meta
 
8:33 PM
@Undo You think it'd be a meta war?
 
@CaffeineAddiction if you want Smokey commands, we need to add you to the privilege list
 
oh, what does that req?
 
which we can do, if you like
 
@ArtOfCode I noticed it and am still in. So Metasmoke is now randomly picking 3 users to let each cast a flag and the SmokeDetector account does not get used, correct?
 
@ByteCommander correct
 
8:34 PM
Right
 
cool, cool.
 
@Undo I'd be for doing a meta post at some point soonish. I think it's worth doing so that we tell people upfront, and we control the dialogue, instead of someone finding out in a couple months and saying "why have these people been half-nuking posts without telling us about it"?
 
For caffeine
deny/approve at your leisure
 
got it
 
@CaffeineAddiction Should be fixed now. Try metasmoke.erwaysoftware.com/authentication/status again
 
8:37 PM
cool, worked
 
Yep, worked that time.
 
!!/pull
 
@Magisch CI build is still pending, wait until the build has finished and then pull again.
 
really
CI is slow today
 
CI on 409af59 succeeded.
 
8:38 PM
logging into GH, it's becoming more and more common to look at my notifications and then spend half an hour reading them because there's 10 or 11
 
!!/pull
Now, work
 
[ SmokeDetector ] SmokeDetector started at rev 409af59 (ArtOfCode: Merge pull request #525 from magisch/patch-3) (running on teward/Aurora)
Restart: API quota is 9074.
 
@ArtOfCode amateur :P
 
Pff
You're a mod on SO
You have more flags coming in faster then you can handle them at any given moment
:P
 
8:40 PM
Not recently, actually. SOCVR's envoy to the group has added enough to keep them down
 
tpu- by Undo
 
@SmokeDetector hey, it worked
 
@SmokeDetector autoflagged
 
@Magisch That post was automatically flagged, using flags from: Yvette, Mithrandir, Magisch.
 
\o/
 
8:40 PM
\o/
nvm
 
works on reload, websocket parsing seems to be borked anyway
 
good to see the system is performing
for now
 
I don't have any reason to think it wouldn't work any different with 3 than with 1 - it was built with this in mind anyway
 
me neither
 
@Undo it's always nice to see it happening, though :)
 
8:44 PM
but I'm notoriously wary of software
Now the spam is getting the boot harder and more fervently then ever
 
@ArtOfCode Probably a good point. Ought to set up a google doc or something and draft something together
 
Art is better at phrasing then me
mumbles project managers ....
 
@Undo mostly for the control of dialogue than anything else - I'd rather just be able to state our case, than have to fight off angry MJW's before we can even start stating it
 
aye
 
@Magisch Annoyingly calm, collected and communicative sometimes :)
 
8:47 PM
Just read the transcript. Nice!
 
@ArtOfCode Worst case scenario would be someone going to MSE going "Who are these charcoal guys and why aren't the CMs banning them for their obvious sockpuppetry?"
 
@ArtOfCode For some reason, I got no ping and no inbox notification there.
 
@Mithrandir huh, I wonder if @Mith conflicted with something else
 
@Magisch And then one of saying 'This is what we do; here's a CMs approval'.
 
@Mithrandir but in this case we'd have to fight off the MJW's before they start listening, which is why I want to get there first
 
8:49 PM
@Mithrandir You know how it goes
It goes into a circlejerk long before we get a chance to explain ourselves
 
Well, the more API keys you are using ... the harder it will for them to determine that its sock puppet-ing
 
Controlling the conversation upfront is better. That way, people read our (hopefully rational and sensible) points before reading conspiratorial rants
 
Consideration: When making such a post, should we allow folks to sign up? It'd be a way to get a heck of a lot of IDs fast.
 
The CMs know what we're doing
We're not hiding it
 
@Undo sign up to... MS, or to flagging?
 
8:50 PM
flagging
 
@Undo Pls no
 
Both, why not?
 
The latter, no. I like that being restricted to people who deserve some free flags.
 
The last thing we need is 100 people bandwagoning onto MS using the lowest possible weight to farm free flags
 
The former, go ahead, though we might want to cull reviewer roles on some of 'em
 
8:50 PM
@ArtOfCode That's easy enough to control, though
 
@Undo it is?
 
@Undo do we need more? we've got 22, times 10 divided by 3 gives us at least 73 posts per site
 
It's a non-issue
there is no site that can exhaust 22 flaggers worth of flags, even with 5. And even if it could, that'd be self-correcting, from getting more the more helpfuls you have
 
@ArtOfCode If we wanted to keep flags to ourselves, suppose we had a group of 'core' users who always got 1 or 2 of the three slots, for example
 
It would be a good way to gather a lot of potential participants for emergencies - and that's a good thing - but I would like to see day-to-day flagging done at least mostly by our own people
 
8:52 PM
Honestly, it's less about getting more API tokens and more about getting people invested in the idea
 
@Undo and emergency capacity doesn't hurt.
 
What do you want with 500 api tokens
 
@Undo this would work - another role?
 
Folks who can sign up to reap rewards of work they didn't do are less likely to yell at you for creating said work.
@ArtOfCode That'd be one way, yeah
 
Prepare for spamsday when drupal.SE gets 2000 spam posts in a day
 
8:52 PM
Drupal's been getting less
It's more apple and graphic design
 
It's Apple that's likely to get it
 
@ArtOfCode if there is an emergency of that magnitude im not sure SD is the correct solution tbh
 
@Undo Not a bad idea
 
@Magisch What don't we want with those tokens is the question. Remember, we're God over the system, we can move flag allocations around however the heck we want.
 
And by getting more people signed up, we'd also be getting more people to really support the idea
 
8:53 PM
aye
 
So future meta wars wouldn't be one sided
 
If anything, we should keep autoflagging to Chq regulars just to guarantee that we can reliably ping them if something goes wrong
2
 
Whats the worst case if a fp happens and the flagger cant be pinged
 
mods can superping
 
a declined or disputed flag
 
8:54 PM
@Magisch declined flag
 
@angussidney We always have chat IDs on all three domains for everyone, and we have plenty of diamonds to go around
 
Thats a non-issue
 
and we tell people enough about the potential for declined flags, so... yeah, okay
 
Honestly, though, at >99% I'm less concerned about that
 
@NobodyNada but I'm an admin on MS, but not a mod, and I might need to do this
 
8:54 PM
I'd be for allowing signups, with the core role implemented
 
@angussidney define 'need', though
 
If someone said to me "You can get free flags but 0,02% of the time they'll be declined" i'd be like, "Where do I sign, this is more accurate then I am"
 
@angussidney but we've got plenty of mods in here
 
@angussidney actually, we're not obligated to ping people if that happens
 
True
 
8:55 PM
People flag completely legit stuff as spam all the time. It wouldn't even raise an SO mod's eyebrow to get one or two flags on something legit
 
If I got a declined flag from autoflagging, then so what
 
we've done it as a courtesy, but declined flags are a real possibility that people accept when they sign up
 
I get declined flags from time to time
It happens
And if I get 300 good flags and 1 declined one
I mean well
 
like literally, if you watched my eyebrows while handling spam flags they'd stay perfectly level.
 
punishment sets in when you get 25% of your last 20(?) declined
statistically impossible
 
8:56 PM
Here's my proposal: we don't talk about how to sign up in the meta post at all, to reduce the flood of signups. If people come in here asking, we can let them know how to then
 
good idea
that way we get only really interested people
 
but still get people
 
4 mins ago, by Undo
Folks who can sign up to reap rewards of work they didn't do are less likely to yell at you for creating said work.
 
I like that a lot.
 
8:57 PM
^ My counterargument for that
 
@Undo yeah, that does negate the effect
 
@Undo it does that
We'll still get a bunch of people
 
Not near the volume
like, probably two orders of magnitude less
 
What do we gain by obscuring how to sign up, given that our control over this is total and indisputable?
 
The question is
 
8:57 PM
Random newbie question: what does this message indicate? Was this flag applied automatically somehow?
 
Do you want and can you build the capacity to accomodate a potentially very popular, cm endorsed meta thread worth of people signing up
 
@Nathaniel He gave feedback with FDSC
 
@Nathaniel No, it means that he flagged as Not An Answer
 
@Nathaniel that was me manually flagging the post it referred to as NAA. I have a userscript that sent that feedback to our web dashboard.
 
and are you comfortable with hosting so many api keys
 
8:58 PM
@Magisch yes, it's EC2
 
@Nathaniel It's feedback
Means ArtOfCode judged that post to be NAA
 
@Magisch I'm comfortable hosting my API key, which could do a heck of a lot more damage than anyone else's that would sign up for this. So yes.
 
You rang?
 
@Undo so go for it
 
@rene ding ding, scroll up
 
8:59 PM
Whatever we do, I think that our previous standard of at least 50 feedbacks required to sign up for autoflagging should be kept in place.
 
@angussidney for core, yes, absolutely
 
@ArtOfCode Ah okay, makes sense – it was different from the replies to SD I see in here.
 
@ArtOfCode I like that as a compromise
 
@Nathaniel yeah, it just means it came through the metasmoke API instead of through chat
 
@Nathaniel It's because it comes from FDSC
 
8:59 PM
@angussidney 50 correct feedbacks :)
 
it's a userscript which automates feedback when we flag
 
@ArtOfCode what do you mean by 'for core'
 
So art flagged NAA and the userscript used that opportunity to feed back NAA on the post
 

« first day (1185 days earlier)      last day (917 days later) »