« first day (1570 days earlier)      last day (2236 days later) » 
00:00 - 06:0006:00 - 09:0009:00 - 11:0011:00 - 12:0012:00 - 13:0013:00 - 15:0015:00 - 17:0017:00 - 18:0018:00 - 21:0021:00 - 00:00

Undo
Undo
9:00 PM
@Seth Unlikely. But if it's manual at all, they might notice.
Everyone wants money; everyone is optimizing for more money. Whether they're technical about it or not.
 
Seth
Seth
What would be really interesting would be to see what happens if you squeeze SO really hard
 
Undo
Undo
"really hard"
Like six flags hard?
 
Seth
Seth
The moderator in me revolts at that.. but the scientist wants to try.
 
Undo
Undo
I could probably stay up late a couple nights to handle any false positives.
 
Mithrandir
Mithrandir
At 400+ weight cast 6?
 
K.Dᴀᴠɪs
K.Dᴀᴠɪs
9:03 PM
That would be fun to watch
time for some popcorn.
 
Undo
Undo
@Mithrandir heck, 280 would be good enough
If I'm watching it.
 
K.Dᴀᴠɪs
K.Dᴀᴠɪs
you can certainly do on SO since you mod it @Undo
higher on others
 
Undo
Undo
One of twenty. Would ask first, doubt there would be too much pushback.
also might have to start coffee.
 
K.Dᴀᴠɪs
K.Dᴀᴠɪs
Spammer posts > insta-delete
 
Mithrandir
Mithrandir
Tag team with Bhargav
 
Undo
Undo
9:05 PM
Yeah, could find someone in a good timezone
 
Mithrandir
Mithrandir
Madara is in my timezone, I know that much
 
Undo
Undo
We could have a decent chance of outpacing even an automatic verification system. Last I checked, we could get flags on posts within three seconds
can barely get photons from India to the US in three seconds
 
WELZ
WELZ
I like it!
 
Mithrandir
Mithrandir
Should probably check with Shog first though :P
 
John Dvorak
John Dvorak
@Undo that's if you use a 0.1c medium
0.05
 
Undo
Undo
9:10 PM
The holy grail would be to delete fast enough that the red banner shows on the page before the spammer hits 'ask question' again. That'd be... discouraging.
 
SmokeDetector
SmokeDetector
[ SmokeDetector | MS ] Bad NS for domain in answer: How to have my PHP Send mail? by Tushar Soni on askubuntu.com
tpu- by Jarko Dubbeldam
New metasmoke user 'Codingale' created
 
M.A.R.
M.A.R.
@SmokeDetector K probably
 
Undo
Undo
@M.A.R. Sure? Link looks legit
 
JAD
JAD
hmm...
btw, is there any need for renaming my MS account to reflect (changed) SE username?
 
Undo
Undo
yeah, they're just using namecheap.
@JAD It'd probably be a good idea
 
JAD
JAD
9:14 PM
done
 
Mithrandir
Mithrandir
@SmokeDetector f
 
SmokeDetector
SmokeDetector
Conflicting feedback on How to have my PHP Send mail?.
 
M.A.R.
M.A.R.
@Undo I have no expertise in anything close to what's being discussed there, but everything has code except the answer, and the answer seems useless without the link
 
JAD
JAD
@SmokeDetector f
 
SmokeDetector
SmokeDetector
Conflicting feedback on How to have my PHP Send mail?.
 
Mithrandir
Mithrandir
9:14 PM
Old, leave it
 
SmokeDetector
SmokeDetector
Conflicting feedback on How to have my PHP Send mail?.
Conflicting feedback on How to have my PHP Send mail?.
 
M.A.R.
M.A.R.
K
 
WELZ
WELZ
boom bam boom beem
 
M.A.R.
M.A.R.
@SmokeDetector F
 
JAD
JAD
@Mithrandir yeah, my initial tpu feedback missed the post date
 
WELZ
WELZ
9:14 PM
@SmokeDetector autoflagged
 
SmokeDetector
SmokeDetector
@WELZ That post was not automatically flagged by metasmoke.
 
WELZ
WELZ
don't forget to retract your flags
 
M.A.R.
M.A.R.
@JAD all those little details
 
JAD
JAD
yep, did
@M.A.R. kinda surprised FIRE doesn't show post date tbh
 
K.Dᴀᴠɪs
K.Dᴀᴠɪs
@WELZ Some mods do not like that. All over meta right now xD
 
doppelgreener
doppelgreener
9:15 PM
any reason i should retract my flag? i still think that post is spam
 
WELZ
WELZ
meh
 
JAD
JAD
@K.Dᴀᴠɪs uh, I don't think retracting flags is what they have issue with
 
M.A.R.
M.A.R.
Should flag for mod: "This answer is wrong. Please kindly do the needful."
 
WELZ
WELZ
I would say naa, though the main part of the answer is still there
 
JAD
JAD
@M.A.R. that's no mod's job though
 
doppelgreener
doppelgreener
9:16 PM
there's no answer
it's a link and a teaser to visit the link
 
WELZ
WELZ
you're right
 
doppelgreener
doppelgreener
"all you need is THREE SIMPLE STEPS visit this link now to find out more"
 
Undo
Undo
askubuntu.com/a/751771/200337 links to the same site, same author
 
M.A.R.
M.A.R.
@doppelgreener right
 
WELZ
WELZ
very generic and maybe even plagiarized
 
SmokeDetector
SmokeDetector
9:17 PM
tpu- by doppelgreener
Conflicting feedback on How to have my PHP Send mail?.
Conflicting feedback on How to have my PHP Send mail?.
 
WELZ
WELZ
oooh lmao
 
JAD
JAD
@Undo that one looks more like an answer tbh
 
Andy
Andy
I'm not in a position to answer rightnow. Can someone get this?
How are posts being evaluated spam/not spam for calculating accuracy? And how is 99.99% meaningful? .01% is less than the margin of error. How correlated are human false positives? What impact does reputation have on posts being marked as spam? — Acccumulation 20 mins ago
 
doppelgreener
doppelgreener
@Undo that one's fine
@Andy "less than the margin of error" sounds like they're applying the science litmus test of 5% which isn't relevant here?
 
M.A.R.
M.A.R.
I am a pacifist. Just flagged as NAA
 
doppelgreener
doppelgreener
9:19 PM
 
JAD
JAD
@doppelgreener the margin of error
also, the 5% thing is not a margin of error
but ugh, people and understanding statistics
 
doppelgreener
doppelgreener
yeah i just, i am not sure what thing they're referring to
 
Undo
Undo
@Andy Something like "As you say, 1 in 30,000 (or even less) is a tiny percentage, so we don't have data on human false positives on these posts. They just don't happen often enough. Reputation is already taken into account on the 100.00% number; higher reputation (anything above 1) is a strong indicator of a post not being spam."
 
doppelgreener
doppelgreener
The Grand Unifying Margin of Error has not yet been discovered by the Priests of Statistic, hallowed be her name
 
M.A.R.
M.A.R.
Yep. Statistics does look like a religion
Albeit a subtle one
Have you noticed the similarities between an atheist and Christian wordboxing and a Trump and Hillary candidate fan fighting over numbers?
 
SmokeDetector
SmokeDetector
9:25 PM
fp- by doppelgreener
 
M.A.R.
M.A.R.
@SmokeDetector F
 
Undo
Undo
@Accumulation As you say, 1 in 30,000 (or even less) is a tiny percentage, so data on human false positives on posts in this level of accuracy just doesn't exist. Reputation is already taken into account on the 100.00% number; higher reputation (anything above 1) is a strong indicator of a post not being spam and excludes it from many checks. All told, only 25 posts were autoflagged from users with >= 2 rep — Undo 11 secs ago
 
SmokeDetector
SmokeDetector
[ SmokeDetector | MS ] Phone number detected in title: turbotax customer service+18448002457turbotax phone number by dharma production on stackoverflow.com
 
doppelgreener
doppelgreener
@M.A.R. i try to notice all of those things as little as possible
 
SmokeDetector
SmokeDetector
tpu- by WELZ
 
WELZ
WELZ
9:26 PM
 
SmokeDetector
SmokeDetector
@WELZ Cannot find data for this post in the API. It may have already been deleted.
 
M.A.R.
M.A.R.
@SmokeDetector K it's that time of the year
Hold on. It's not that time of the year. It's probably not Christmas
 
Rand al'Thor
Rand al'Thor
Wow, looks like Smokey is doing really badly on Vegetarianism as well as Bioinformatics: metasmoke.erwaysoftware.com/sites/…
I guess there must be something about those sites that generates FPs. Nutrients being on-topic and whatnot.
@Riker You guys going to ask to change anything with Smokey over there too?
 
SmokeDetector
SmokeDetector
New metasmoke user 'mevans' created
 
Undo
Undo
@Randal'Thor Not 'badly' - just the incoming signal isn't as good. Autoflagging only got one FP there.
 
M.A.R.
M.A.R.
9:29 PM
@Randal'Thor nutrients, and weight loss
 
JAD
JAD
and working out
 
Undo
Undo
SmokeDetector goes for 0% false negatives, autoflagging goes for 0% false positives. It's easy to read that page wrong.
 
M.A.R.
M.A.R.
@SmokeDetector so is that M.Evans or MeVans?
 
Rand al'Thor
Rand al'Thor
@Undo I know that :-)
 
JAD
JAD
@M.A.R. or MevAns
 
M.A.R.
M.A.R.
9:30 PM
@Undo that FP is really bloody though
 
Rand al'Thor
Rand al'Thor
But "autoflagging only got one FP" is already a really bad thing, IMO.
@M.A.R. Aren't blacklisted words customised per site though?
 
Undo
Undo
... also, that one FP wouldn't be flagged now. System has evolved in the last four months.
(click 'eligible flaggers' on metasmoke.erwaysoftware.com/post/94047/flag_logs, there are none)
 
M.A.R.
M.A.R.
@JAD Or M.E.Vans. OMG
 
doppelgreener
doppelgreener
Bioinformaticcs data: metasmoke.erwaysoftware.com/sites/… (Past 6 months: 5 posts, 5 FPs, nothing else)
 
JAD
JAD
@M.A.R. endless possibilities
 
M.A.R.
M.A.R.
9:32 PM
Someone salvage this urgent situation now
 
Undo
Undo
@doppelgreener None flagged, valuable data for knowing what not to do.
 
M.A.R.
M.A.R.
@Randal'Thor someone should do the customizing
If you don't add the rule, it's not customized
 
SmokeDetector
SmokeDetector
[ SmokeDetector | MS ] Potentially bad keyword in answer: Kernel Data Inpage Error by Abigail Vaughan on superuser.com
 
M.A.R.
M.A.R.
@SmokeDetector fishy
Very fishy
Can someone confirm it's spam?
 
Rand al'Thor
Rand al'Thor
@doppelgreener Isn't "past 6 months" equal to "all time"?
It's a very young site.
 
M.A.R.
M.A.R.
9:35 PM
@SmokeDetector why
 
SmokeDetector
SmokeDetector
@M.A.R. Body - Position 318-337: errorsolutions.tech, Position 409-428: errorsolutions.tech
fp- by doppelgreener
 
doppelgreener
doppelgreener
@Randal'Thor don't know
 
K.Dᴀᴠɪs
K.Dᴀᴠɪs
@doppelgreener The point is that SD attempts to catch all possible spam, which include FPs.... But only autoflags all true spam. The "possible" spam goes in this chat for quick review
 
WELZ
WELZ
we've had 2 reports with that link (same user)
 
doppelgreener
doppelgreener
@K.Dᴀᴠɪs yeah the situation there is fine
 
M.A.R.
M.A.R.
9:35 PM
@WELZ thank you
@SmokeDetector K
 
SmokeDetector
SmokeDetector
Conflicting feedback on Kernel Data Inpage Error.
 
doppelgreener
doppelgreener
@WELZ hm, that's more fishy
@SmokeDetector k
 
WELZ
WELZ
@M.A.R. but it says in "our" guides.
 
doppelgreener
doppelgreener
"i found a website..." in one of those posts
 
M.A.R.
M.A.R.
None of those really help
@doppelgreener more fishy
 
doppelgreener
doppelgreener
9:36 PM
@WELZ i'm K'ing it on the basis it's not a good quality answer: without the link it's not an adequate answer at all
 
M.A.R.
M.A.R.
Dang, so much seafood
Phosphorus is good for everyone else and not for me.
 
WELZ
WELZ
I'm just not gonna feedback, but flag as naa
 
doppelgreener
doppelgreener
the spam guidelines: (a) post your link as part of an otherwise good quality answer, and (b) disclose affiliation. people are expected to pass both bars.
 
M.A.R.
M.A.R.
And seafood lacks Vitamin C
 
doppelgreener
doppelgreener
@SmokeDetector naa-
 
SmokeDetector
SmokeDetector
9:37 PM
New metasmoke user 'jhpratt' created
 
M.A.R.
M.A.R.
So is it Vitamin C or vitamin C?
vITAMIN c
 
rene
rene
@M.A.R. C
 
M.A.R.
M.A.R.
@rene that solves everything
 
Rand al'Thor
Rand al'Thor
Vitamin C++
 
JAD
JAD
Vitamin C#, way more modern
 
M.A.R.
M.A.R.
9:39 PM
You might have answered the P N(P) problem, rene
 
JAD
JAD
Imma call that jinx
 
Rand al'Thor
Rand al'Thor
Yep.
 
rene
rene
@M.A.R. sure. yw.
 
M.A.R.
M.A.R.
@rene I get 20 percent of the money you get for solving it, BTW
 
Rand al'Thor
Rand al'Thor
That's a lotta money.
 
rene
rene
9:47 PM
@M.A.R. You can get 20 percent of my profit/loss balance.
 
M.A.R.
M.A.R.
@Randal'Thor fine, I'll give you 20 grand to stay silent
 
SmokeDetector
SmokeDetector
New metasmoke user 'mootrichard' created
 
M.A.R.
M.A.R.
@SmokeDetector sounds moot
Bah dum tiss
 
Rand al'Thor
Rand al'Thor
Treebeard once made a moot point at the Entmoot.
 
M.A.R.
M.A.R.
@Randal'Thor mootopia
 
SmokeDetector
SmokeDetector
9:50 PM
[ SmokeDetector | MS ] Shortened URL in answer: HPLIP Drive in Elementary by William on elementaryos.SE
 
Mithrandir
Mithrandir
@Randal'Thor I doubt that.
 
Rand al'Thor
Rand al'Thor
@SmokeDetector f
 
doppelgreener
doppelgreener
@SmokeDetector i've suggested a revision
 
M.A.R.
M.A.R.
@SmokeDetector F
Elementar yos
 
Rand al'Thor
Rand al'Thor
L.M.N. Taryos
 
quartata
quartata
9:52 PM
@tripleee you use Bottle, right?
 
thesecretmaster
thesecretmaster
@doppelgreener Mind if I steal this wording for the site dash?
 
Rand al'Thor
Rand al'Thor
L.M.N. t'R.E. OS
 
doppelgreener
doppelgreener
@thesecretmaster that was my intention
 
thesecretmaster
thesecretmaster
Ok. If someone else isn't already doing it, I will.
 
doppelgreener
doppelgreener
as captain jack would say: raid, pillage, plunder and otherwise pilfer your weasely black guts out.
 
Mithrandir
Mithrandir
9:54 PM
@Randal'Thor What're these Random letters?
 
WELZ
WELZ
> savvy?
 
M.A.R.
M.A.R.
Element tree ooze
 
Rand al'Thor
Rand al'Thor
@doppelgreener As Captain Jack would say: "Excellent bottom. ... What, can't I say hello to anyone?"
Oh, wrong Captain Jack.
 
doppelgreener
doppelgreener
@Randal'Thor hahaha
 
doppelgreener
doppelgreener
10:08 PM
inkscape will do, i've been fiddling with it a bit. but i'll need to do the rest of planning out the infographic tomorrow
 
Rand al'Thor
Rand al'Thor
inkographic?
 
doppelgreener
doppelgreener
yes that
context, also:
3 hours ago, by doppelgreener
I think y'all would benefit from an infographic showing a bunch of posts arriving, being pulled out to smokedetector, being given to human reviewers which train algorithms, and those algorithms prompt when flags should get a few automatic flags to start with.
i'm gonna work out how to lay it out then work out how to make it look pretty. it'll be focused on tall and narrow, so that it can fit in stack posts.
(it won't be very tall, but it'll be taller than wide)
 
SmokeDetector
SmokeDetector
fp- by doppelgreener
 
Rand al'Thor
Rand al'Thor
@SmokeDetector f
@SmokeDetector why
 
SmokeDetector
SmokeDetector
@Randal'Thor Post - faucetastic.fun NS suspicious dns1.namecheaphosting.com.,dns2.namecheaphosting.com.
 
Rand al'Thor
Rand al'Thor
10:15 PM
NS?
Neta Smoke?
 
CalvT븃
CalvT븃
@Randal'Thor Nameserver
 
SmokeDetector
SmokeDetector
New metasmoke user 'rollstuhlfahrer' created
 
doppelgreener
doppelgreener
i.e. the service that hosts the domain name
namecheaphosting seems to be connected to a lot of spammy sites
 
SmokeDetector
SmokeDetector
New metasmoke user 'Etienne' created
 
Shog9
Shog9
10:40 PM
More trivia than anything else, but... I got curious
% of spam flags that actually result in the flagged post being spam-deleted, by site: 
                Site                 PctSpamFlagsEffective
------------------------------------ ---------------------
Stack Overflow                                     62.88 %
Super User                                         96.12 %
Server Fault                                        85.6 %
Stack Apps                                         95.27 %
Seasoned Advice                                     77.2 %
Home Improvement                                    90.5 %
Game Development                                   76.47 %
(that's markedly different from the % of flags that are marked "helpful" or which are attached to deleted posts in some way)
 
quartata
quartata
Whoa... Ask Patents at the top??
Of course Drupal second, not much of a surprise
 
Rand al'Thor
Rand al'Thor
Wow, what's with Stack Overflow en español?
 
WELZ
WELZ
@Shog9 graphic design 98.97% wow
 
Wrzlprmft
Wrzlprmft
@Shog9 Actually, the strong variation is quite impressive and it would make sense to think about the reasons for some sites.
@Randal'Thor I don’t speak more than a few words Spanish, but I would take a look at the wording of the flag.
 
quartata
quartata
It and Portuguese tend to have spam stick around for a long time. Doesn't help that Smokey has ridiculously huge queue sizes on it (not sure why)
 
Shog9
Shog9
10:46 PM
take these with a grain of salt. They don't differentiate between the first flag on a post, the only flag on a post, and a post that got 6 flags but was later disputed by a moderator.
actually... Lemme regenerate this, there's an edge-case involving that last thing
waffles
...too many windows open
 
Rand al'Thor
Rand al'Thor
@Wrzlprmft I'd guess a lot of lower percentages are due to site mods who don't know about the effect of the mod red-flag and simply delete spam-flagged posts.
It's pretty counter-intuitive for a mod to be raising flags instead of handling them.
 
Wrzlprmft
Wrzlprmft
@Randal'Thor Right, that could be.
@WELZ I am not too surprised. We get lots of blatant spam, little borderline spam, and we don’t have idiots who flag everything as spam that they don’t like.
@Shog9 On how much time is this based? The percentages on some sites suggest that there were only a handful of flagged posts in total.
 
Shog9
Shog9
@Wrzlprmft oh, yeah - this is just the last 365 days.
 
WELZ
WELZ
@quartata isn't GD top?
 
quartata
quartata
yes it is I misread
followed by workplace and AU, then Ask Patents
 
Wrzlprmft
Wrzlprmft
10:57 PM
@WELZ Astrononomy is top with 99.00 % followed by Graphic Design with 98.97 %. (Gotta find that idiot who raised that one flag that made the difference and suspend them …)
 
SmokeDetector
SmokeDetector
New metasmoke user 'Anko' created
 
WELZ
WELZ
@Wrzlprmft Lmao
 
quartata
quartata
Astronomy I suspect gets so much spam because it's first in alphabetic order of the non 1.0 sites
(probably skip 3D printing because they don't expect numbers in the subdomains)
...wait no, academia is first
huh
beats me then
 
Wrzlprmft
Wrzlprmft
My bet is that some spammers were successful in the past with their spam staying hours or days and they remember.
 
Rand al'Thor
Rand al'Thor
@Undo Is there a particular reason why an out-of-context "please do ignore SFF" is pinned to the top of the star-board?
 
quartata
quartata
11:06 PM
Shog wants us to focus on the MSE post first
(plus it was starting to get under our skin a bit and we needed someone to tell us off)
 
WELZ
WELZ
Made @Shog9 's list ordered by highest percentage to lowest bit.ly/2I8FhW6
 
SmokeDetector
SmokeDetector
New metasmoke user 'Christian Dean' created
 
doppelgreener
doppelgreener
11:23 PM
@Randal'Thor context: people kept mentioning SFF's meta. right now it's not the issue we need to be focused on, is the idea. plus you need space to work out whatever you're going to work out, don't need us crowding your meta again.
 
Rand al'Thor
Rand al'Thor
@doppelgreener Gotcha, thanks. Out of context it seemed kinda dismissive of our site, but that makes more sense.
 
Undo
Undo
That.
 
Rand al'Thor
Rand al'Thor
(although there have also been some less-than-polite comments about SFF and its community in here - I've even deleted one or two from the transcript)
 
doppelgreener
doppelgreener
@Randal'Thor and that is also why we need to heed that advice right now
 
Undo
Undo
@Randal'Thor Please... don't do that. Ping me or Art instead; that's a good way to ignite something no one wants to deal with.
 
Rand al'Thor
Rand al'Thor
11:29 PM
(I get that some users were being pretty annoying for you, and that this room can be a tough place to moderate, but ... well, I've written about this elsewhere, and have tried to enforce it in other rooms in the past)
@Undo Ping you here or in TL? Silent deletion seemed to me to be the best way not to start an argument out of it, but of course you know this room better than I.
 
ArtOfCode
ArtOfCode
1359 new messages whistles
 
SmokeDetector
SmokeDetector
CI on c78054b succeeded. Message contains 'autopull', pulling...
[ SmokeDetector ] SmokeDetector started at rev 782fad3 (SmokeDetector: Auto watch of go4trial\.com by iBug --autopull) (running on teward/Solar Flare)
Restart: API quota is 7626.
[ SmokeDetector | MS ] Offensive answer detected: Difference between earlier, recently, and lately by user70776 on ell.SE
 
Undo
Undo
@Randal'Thor Probably, but deletion from me is going to be received better than someone 'not involved'. I trust your judgement; I'm just in a better place to shut down dissent.
 
Rand al'Thor
Rand al'Thor
@SmokeDetector n
 
Undo
Undo
(basically, 90% chance of not starting an argument by silently deleting... But that 10% is going to be really ugly.)
 
SmokeDetector
SmokeDetector
11:39 PM
tpu- by K.Dᴀᴠɪs
tpu- by K.Dᴀᴠɪs
 
WELZ
WELZ
 
SmokeDetector
SmokeDetector
@WELZ Post Difference between earlier, recently, and lately does not look like spam.
 
Shog9
Shog9
@Randal'Thor sorry about that. I said this on MSFF yesterday, but... I really didn't want the folks here doing this on every meta site. The concerns that've been raised thus far are almost universally applicable to every site; it's a much better use of everyone's time if they're discussed and addressed once.
 
Rand al'Thor
Rand al'Thor
@Undo Fair. There's also the risk of me being seen as biased because I'm from SFF. I just ... really couldn't face another exhausting discussion, after already having had to handle those comment threads on meta.SFF, about the whys and hows of the appropriateness of things said in here. Especially since some of the people concerned were mods, so even pinging you in TL wouldn't necessarily avoid that debate.
 
Undo
Undo
Understand, no harm done
 
Shog9
Shog9
11:43 PM
I can't stop y'all from discussing this on MSFF, but... As much as is possible, that should not be the focus of the SD team, this room, or the general strategy going forward - it's just not workable.
 
Rand al'Thor
Rand al'Thor
@Shog9 Sure. I've tried to steer the debate on SFF meta towards "what do we want / not want Smokey doing on this site specifically" rather than a more general "are bots and autoflags bad"-type discussion.
 
Shog9
Shog9
@Randal'Thor which is helpful, thanks.
 
Undo
Undo
Yep, that was really good. The "bots are bad" thing wasn't going to do any favors.
 
Shog9
Shog9
But even with that, please do keep in mind that any concerns that have merit are probably not isolated to the folks on SFF. Turning it off on SFF doesn't placate users on SU, or SA, or AU, or EL&U, or...
 
Rand al'Thor
Rand al'Thor
And I looked at site-specific concerns in my answer there, although a lot of it ended up concluding that there aren't really any SFF-specific issues.
 
Shog9
Shog9
11:47 PM
The entire point of having a massive featured discussion on MSE - which we/I have to coordinate and moderate - is to get all this out in the open; all the objections and fears and misinformation and unanticipated concern and trolling and whatever - and try as much as possible to address it once rather than doing it once for every site ever created now or in the future.
 
Rand al'Thor
Rand al'Thor
@Shog9 I do think that medium-sized sites in general - big enough to have many active users and flaggers, not SO-scale so that new content disappears in moments - don't get much benefit from autoflagging. But trying to have that debate on main meta brings in so many different factions with different views.
You have to worry about the whole network. Most of us just worry about our own individual sites ;-)
 
Undo
Undo
@Shog9 I pulled some numbers on SO a while back for 18 autoflagged false positives, looking at how many of them received bad manual flags. Any reason I can't share that single number replying to meta.stackexchange.com/questions/307585/…? It'd help show an order of magnitude, at least
 
Shog9
Shog9
@Randal'Thor see... That's a concern that's been raised a number of times. And it would be helpful to address it.
 
Undo
Undo
I'd just do it but it's information obtained through manual mod-timeline snooping.
 
Shog9
Shog9
@Undo no reason
 
Catija
Catija
11:49 PM
@Randal'Thor Where do you get that info, though? Lots of it disappears in moments because of the autoflagging and the users in here manually flagging it.
 
Undo
Undo
k, thanks
 
quartata
quartata
some medium sized sites (like PPCG) are bad at handling spam because the votes get split across spam flags/delete votes. It's useful there
 
Undo
Undo
So... for anyone interested here, the number is two.
 
Shog9
Shog9
that is a boring number
 
Undo
Undo
2/18 false positives on SO were autoflagged and later received manual flags.
 
quartata
quartata
11:49 PM
What's the reason weights like
 
Undo
Undo
low
 
Shog9
Shog9
@Randal'Thor y'know... I made a passing comparison yesterday to the vaccination debate for a reason...
 
Undo
Undo
+/- 1 because it was a manual process and I'm a dog and don't have hands.
 
Rand al'Thor
Rand al'Thor
@Catija It's more of a gut feeling than a hard statistic. Worth bearing in mind, though, that it can be hard to tell how quickly a site's active users would get rid of spam when they never get a chance because Smokey/Charcoal are so efficient.
Think of ... a site where mods closehammer every off-topic question.
 
ArtOfCode
ArtOfCode
Hardware Recs, yeah
 
Rand al'Thor
Rand al'Thor
11:51 PM
The community has the ability to close these questions themselves, but they get lazy because the mods are always doing it.
 
Catija
Catija
@Randal'Thor I'm really sorry but I don't understand why you even want your regular users to have to deal with this in the first place... They should be doing actual site stuff, not looking at links to penis enlargements and creepy apps for stalking your girlfriend.
 
Undo
Undo
@Randal'Thor The autoflagging experiment gives you some insight into that - look at the 1 bar.
 
Rand al'Thor
Rand al'Thor
It's impossible to prove that the community would close them by themselves, because they never have a chance due to the swiftness of the mods.
 
quartata
quartata
Closing is a whole different story
 
Undo
Undo
(which, by the way, I need to regenerate sometime with new data)
 
Rand al'Thor
Rand al'Thor
11:52 PM
@quartata It's an analogy :-)
 
quartata
quartata
@Undo you got my stats right?
 
Undo
Undo
ummmm
Email or something?
 
Catija
Catija
I've seen even mods just delete spam instead of spam flagging it...
 
ArtOfCode
ArtOfCode
@Undo wanna deploy MS so the next dump has the right data in it?
 
Rand al'Thor
Rand al'Thor
@Catija Did I say that I want regular users to have to deal with this?
 
Undo
Undo
11:53 PM
@ArtOfCode ... the right data?
@Randal'Thor Some people here make the jump.
 
quartata
quartata
@Undo no I pinged you (I think)
 
ArtOfCode
ArtOfCode
@Undo right format. I'm a migration or three ahead of prod at the moment.
 
Undo
Undo
oh
Laptop is ded, I'll just re-run the dump after I deploy.
 
quartata
quartata
tl;dr 4 vs 5 flags p = 0.005, 3 vs 4 flags p = 0.03
 
Rand al'Thor
Rand al'Thor
Some regular users do, though. Some people feel that part of the moderation of their site is being taken away from them.
Personally I don't much care who flags spam; I'm more worried about practical things like the possibility of mistaken flags.
 
Catija
Catija
11:55 PM
@Randal'Thor Your ambivalence to autoflagging at all sort of implies that you'd rather your own users handle it... sure.
 
Rand al'Thor
Rand al'Thor
@Catija Autoflagging =/= CHQ flagging.
 
Undo
Undo
@Randal'Thor Hypothetically, suppose you could pick a number of false positives per timeframe you were happy with, only constraint being > 0. What would it be?
 
Shog9
Shog9
Look, this is why it's best to discuss the value to the network not any one site. Nobody needs tool-assisted flagging until they get slammed, and then it's kinda too late to discuss.
 
Undo
Undo
On a "n/year scale". Like 0.5/year.
 
Rand al'Thor
Rand al'Thor
If a bunch of 101-rep people who know a lot about spam help to keep my site free of it, fine. (As long as there's shitloads of oversight and accountability and whatnot, which CHQ has.)
 
thesecretmaster
thesecretmaster
11:58 PM
@Undo Wait! If you're deplying, please have a look at my PR first :)
 
Catija
Catija
@Randal'Thor Getting rid of Autoflagging means that more flaggers have to actually be in here to flag. So, in this case, turning it off leaves more work for your site's users.
 
Rand al'Thor
Rand al'Thor
@Catija Aren't there always people in here to flag? It's one of the busiest rooms on the entire network.
 
quartata
quartata
no. We don't just do auto flagging as a convenience thing
 
Undo
Undo
@Randal'Thor Getting 6 has always been a challenge, even pre-autoflagging. Stuff used to get missed and live hours.
 
Catija
Catija
@Randal'Thor No, there aren't. The number of posts that end up in The Tavern after five minutes of being unhandled pretty much tells me that quite often there aren't even three people in here, let alone six.
 
00:00 - 06:0006:00 - 09:0009:00 - 11:0011:00 - 12:0012:00 - 13:0013:00 - 15:0015:00 - 17:0017:00 - 18:0018:00 - 21:0021:00 - 00:00

« first day (1570 days earlier)      last day (2236 days later) »