300+ weight is just a proxy for accuracy.

Ahh havent read it all

Been back and forth

if we were autoflagging all posts with 5 flags there would be suspensions up the ass

I stuck a CHQ link on the end of the post

I mean look at this right now

@ArtOfCode Oh boy

imagine getting 5 people to retract those flags

Ash would definitely track us down and skin us in real life

they're a cyborg

@ArtOfCode we're probably going to get that metasmoke flood we got last time lol

Well at 300 weight we're literally 100.00% accurate, sooo

@quartata :/ Huh?

@quartata there's a signup link in the post, so we'll get some. I predict less than last time, though.

@ArtOfCode yeah hence 300 weight

I'm getting reeealy nitpicky now, but maybe change "That's a chronological representation (left-right, top-bottom) of every post that could have been flagged" to "would have been flagged" to show that we'd intend to flag all of those green ones?

if we were autoflagging with 5 flags and the flagging conditions like only 2 reasons

even an FP a week would get people seriously mad at us

@Undo done

@Undo hmm that seems stronger

oh wait missing context

I was thinking more "could have been flagged, if we could ..." since it's more deferential

but those are just the ones that we already flagged that would get the additional flags right

aye

It's all of MS history

oh right

forgot reason weight is actually retroactive

Yeah, that's one I've wrestled with. I believe it's more honest to have retroactive reason weight because it shows the state of the world now, how we would act on those posts now.

There are good arguments for freezing reason weight with the post, too - but then you risk saying "this has a 99.9% accuracy" but actually your reasons are crap now

yeah yeah that's definitely worse

better to be conservative about it

huh... just got the SE API complaining about an invalid filter

that's... weird?

Invalid filter?

Websockets?

no when hitting up /comments

"Some posts already have 6 flags cast on them by Charcoal" s/Some/Many?

my filter includes fields other than comments. is that it?

Because I'm pretty sure it's "Many" and that's a smaller change

'!)k_3iHEfnikRI4A..IXrGIwdpOt2bdTX268heIMovWXJwZNk.BEPI*R(8AbEt)AkH&key=IAkbitm‌​ze4B8KpacUfLqkw(('

Holy mackerel

@Undo better make sure it's obvious this is manually

@Undo actually is "some", we pulled data on that

How close to "many" is "some"?

@Undo it actually has less fields than the Smokey one

I excluded a lot of stuff

If it's like 45%, we're probably a majority since we don't have all the data. But it's not something I care much about

"We do, however, recognise that it's possible that a site's community may want to handle its own spam; if that's the case, we're happy to turn the autoflagging system off on this site or to retain it at its current levels."

I like this just how it is - no reason to mention a procedure and get people debating that.

@Undo not very many at all

hold on, I had a graph

oh yeah. saw it. The rainbow looking one.

@Undo also the way it's set up is kinda absurdist: no one really wants to argue towards "handling their own spam" unless there's a serious problem

that is very colorful

o.O

Seems... oddly small. I thought the orange was 6 flags from us at first.

I feel like those shouldn't be stacked

obviously excludes manual flags not cast through the MS API, which is non-trivial, but it's not going to make a huge change

@quartata Bet you didn't want that key in there...

@thesecretmaster eh, it's not a graph we're using in the meta post

@Andy it's Smokey's

they're not secret

@Andy it's Smokey's key, which is already public

oh

Well...fine then.

you're right though

it shouldn't be in there

that's why it's erroring

wat

yeah, it's going like filter=<url encoded all that>

so the key got gobbled up and thrown in

I'm happy with it @ArtOfCode

Now... wanna talk about the PR strategy on this one?

Cool

Yeah, go for it

We're probably not going to overwhelm mSE, like we could on tinysite.meta.se

But it'd be good to have our message straight, and not look like... extremists, if you know what I mean ;)

Message being something along the lines of "we think this is a good idea and that the data backs that up, but we appreciate that it's partially a subjective issue and want to have a rational discussion"?

Probably

@ArtOfCode What are the numbers? Total posts?

yeah

One last comment on the post - we make it very clear what our goal for the post is in the TL;DR at the top. Then we say a few words and ask for an opinion. After those few words, though, users may have forgotten what we are asking for again. Can we reask the question at the bottom when we ask for opinions/feedback?

yup, makes sense

Question: What feedback might be truly useful to us?

Last time it was either "this is awesome" or "this is horrible" and then we convinced them it was awesome. Not really any technical feedback.

Might be able to channel the conversation to useful feedback, if we know what that might be.

@Andy what kinda of phrasing are you thinking?

"plz send teh words"

Tell us about any concerns or questions you have about this?

@Undo I'm going to guess we get the same type of feedback here too. The post is filled with the technical. Honestly, it's going to be hard to get technical feedback if you don't know the system anyway. I'm expecting feedback to range from "um...ok?" to "why isn't SE doing this?" to "how do I sign up" with a little bit of "automation is bad, m'kay"

So about like last time.

Yeah

Morning all!

Works for me. It worked last time, why change it this time

And most of the other "here's this automated thing that's running now"

@AJ Evening!

Do we want to cut the experiment when we post this, or leave it running and update the numbers every day or so?

watchin' Oscars right now. bbl

@ArtOfCode Pretty simple I think. In the last paragraph just add our goal at the start of the section:

> We want to increase the number of automated flags from 3 to 5 to reduce the time spam spends alive on the network. We'd like to hear your thoughts. We appreciate...

@Undo How long has it been running?

done

Are we at a week?

Just curious, what sort of response would make y'all decide not to do the 5 autoflags?

@Andy yup. A week almost to the minute right now.

@thesecretmaster An overwhelmingly negative one

A "meh" positive one would also make me ask questions, given the expected [feature]ing

@ArtOfCode Can you adjust the TL;DR to be just "5" too?

also done

Looks like the sidebar is clean. The HQ2 post has been around for a while; we won't crowd out anything important.

Go time?

\o morn

Marshal badge in Apple.SE \o/

@SurajRao Congratulations!

Thanks

I got a marshal on Hinduism.SE some time ago. \o/

Might be good to remind folks that if there are 5 autoflags, the 6th can't be retracted.

I suspect there will be plenty of reminding about that whether we want it or not ;)

@NogShine nice! congrats

@Andy you want the markdown here?

I can grab it from the gist

Might re-upload the images to stack.imgur so we don't get ourselves banned on GH

Are we happy with the title?

eeh

@Undo most of them are, I think

Heh. I'll be sure it uses Art's github links :)

Anyone got a better one?

Oh, here you go: At the bottom, add "What's a better title for this?"

That's a good constructive thing for people to discuss :P

Actually, we do need an attractive, short, sidebar-friendly thing.

I'll be sure to have to various cans of paint available too

Throwing out junk here:

"Can a machine (be taught to) flag spam faster?"

A machine can flag spam automatically. Can it do it better?

Like

Like

Go for it

[ SmokeDetector | MS ] Link at end of answer: Hiding secret from command line parameter on Unix by sambuca on stackoverflow.com

fp- by tripleee

I'm grabbing 355adf8c2afc595ea2726587dfeb4020 as the version unless anyone complains in the next minute or so

> Revisions (37)

yeah that's probably good enough

@Andy that's a permalink, not a commit SHA ;)

sigh I knew that

wanders into TL and prepares self

oh, and remember: We're diverting the TL conversations here. Not there.

You want a drumroll?

Big Fancy Meta Post

And off we go

I don't have issues with Charcoal folks participating there. We're network people too.

I ignored all the red underlines that Firefox had on the British spellings. I expect someone to come along later and fix those up

Someone ping Shog to feature

I'll do it

"you"? I'd think that the group name would be better?

Sorry... I didn't ever look at the draft...

@SurajRao Thanks!

So...bed time is in a few hours. I'm taking bets now on how many notifications my phone has when I wake up.

Couple dozen, probably. US is asleep, Shog is probably asleep. Insanity won't start in earnest until [featured] happens

thanks

[ SmokeDetector | MS ] Bad NS for domain in body, blacklisted website in body, link at end of body, pattern-matching website in body: Essencia Cream And consuming a small portion by fangynia12 on askubuntu.com

tpu- by Undo

Now... please don't have an unlucky bunch of 300-weight false positives in the next few hours.

... It's not about luck... (goes to post a bunch of answers that trip all of the flags from a sock)

Is it bad I thought the same thing?

Oh, something happened while I was gone :)

A thing happened!

Five upvotes. It's unanimous, let's pack it up.

@Undo 6*

!!/blacklist-keyword essencia

@tripleee Blacklisted essencia

CI on 4226b32 succeeded. Message contains 'autopull', pulling...

[ SmokeDetector ] SmokeDetector started at rev 4226b32 (SmokeDetector: Auto blacklist of essencia by tripleee --autopull) (running on teward/Solar Flare)

Restart: API quota is 18513.

fp by suraj on What is the best way to seal caulk tube? [MS]

fp by suraj

@SmokeDetector misclick sry

... each_with_index does it in the order item, index not index, item

that's not confusing at all

Oscars finished. And I am back here. What'd I miss?

Four comments, two votes

Everyone's sleeping, probably

IT LIVES

...although it's posting comments from stackoverflow for some reason

!!/report interpersonal.stackexchange.com/questions/11406/…

[ SmokeDetector | MS ] Manually reported question: How to know if your partner is cheating on you? by Jessica A Politron on interpersonal.SE

tpu- by tripleee

hahahahaha

yasss

maniacal cackle?

@SmokeDetector I think that question they want to answer was protected.

Odd... those are usually answers?

[ SmokeDetector | MS ] Link at end of answer: Would animals really mutate in the post-apocalypse? by Voglia Designers on worldbuilding.SE

tpu- by tripleee

@Catija The recent one was.

If you're also showing the site dashboard to mods at the same time, it would be helpful for the dashboard to also explicitly show numbers for 300 weight detentions. Right now it shows numbers for all autoflags. It's easy to not understand the distinction. When you don't understand the distinction, it looks like there are actually a considerable number of autoflags cast on false positives, which some people will equate to posts on which we're wanting to cast 5 autoflags.

Even just an additional line of text on the dashboard that gives # of autoflags on > 300 weight detections with # TP and # FP, would be helpful to make sure people are reminded that there are distinct categories (or at least that we're wanting to define distinct categories) within the autoflagged detentions.

@Undo @ArtOfCode when you get a chance

@quartata what about it?

would appreciate a review wrt the rails bits

[ SmokeDetector | MS ] URL-only title, bad NS for domain in body, bad NS for domain in title, bad keyword in body, bad keyword in title, +4 more: www.nutritionofhealth.ca/dsn-pre-workout/ by user802230 on askubuntu.com

tpu- by tripleee

New metasmoke user 'Nog Shine' created

You mentioned that JBuilder could probably be used to make BotController#update_from_json work, that's the ugliest bit @ArtOfCode

also I will need frontend help at some point (not quite there though)

[ SmokeDetector | MS ] Blacklisted website in body, link following arrow in body, pattern-matching website in body, potentially bad keyword in body, potentially bad keyword in title: Nitronemax then why not likewise do likewise by Pamelcarlin on graphicdesign.SE

@SmokeDetector k

tpu- by K.Dᴀᴠɪs

whoops already spotted a mistake

3-flag sites are Workplace, ELU?

aye

That's good to know, need to watch for that.

Just to clarify my understanding: We're wanting to increase to 5 autoflags, but only on a subset of detections. Is that correct? The way the post is worded it's not clear what is actually being proposed. It says things like "we could require", but it's not clear if you are proposing to go to 5 flags on every detection, or only on those which hit the (historical) 99.99% (it just says that you could do it that way, not that it's what's proposed).

It's carefully worded to not back us into a corner.

The only thing being proposed is to go to 5 flags on some subset of posts - our choosing. That freedom is important for us.

[ SmokeDetector | MS ] Blacklisted website in answer: How to get high resolution website logo (favicon) for a given URL by Evans Okyere Nyark on stackoverflow.com

tpu- by tripleee

[ SmokeDetector | MS ] Bad NS for domain in body: HtAccess Laravel by Sarwar Ahmed on stackoverflow.com

[ SmokeDetector | MS ] Blacklisted website in answer, pattern-matching website in answer: Kodi wont let me access Ubuntu or terminal by Sophia Joseph on askubuntu.com

tpu- by suraj

What's the warning users are shown when enabling autoflagging?

fp- by suraj on HtAccess Laravel [MS]

@Andy metasmoke.erwaysoftware.com/flagging/preferences

[ SmokeDetector | MS ] Blacklisted website in answer: Calling multiple instances of python scripts in matlab using java.lang.Runtime.getRuntime not working by Prasanthi 7 on stackoverflow.com

Just thought of that. It's a pretty big benefit for autoflagging when it catches the nastier stuff we get.

as long as skynet does not become self aware of course

@Undo Ok. I think it might have been better to go for a concrete plan rather than an open ended one, but I understand the desire not to have restraints. Another time, given that you are expecting, and getting, opposition, a strategy which requests the expansion of responsibility/freedom of action in smaller stages is more likely to be successful, or at least that's how most human dynamics work.

woops, forgot to append the token for polls

@Makyen We can always move that direction if necessary. It's harder to move back from that position if it goes better than expected.

[ SmokeDetector | MS ] Bad keyword in answer, blacklisted website in answer, pattern-matching website in answer: How to handle leadership saying, "we have been over this" by juliejohnson on interpersonal.SE

tpu- by K.Dᴀᴠɪs

Hmm, also need a post type-site relation for last scanned

> While getting rid of individual cases of spam is a worthwhile endeavor, it also means that mods are less likely to notice and nuke the accounts. It's not just about time to deletion.

gah I knew the "awarded deputy for flags I didn't raise" thing would be used against us

@SurajRao wrong wrong wrong wrong

just no

@Undo I disagree as to which is harder. IMO, it's much easier to ride a sequence of successes, even if that sequence is spread over a short period (e.g. restricted, then 3 months later ask for the restrictions to be lifted). It's, usually, harder to recover from a failure, because people remember that you failed, and are less likely to agree to future requests.

[ SmokeDetector | MS ] Bad keyword in body: What is Hot Shaper? by hotshaperbelt on stackoverflow.com

tpu- by K.Dᴀᴠɪs

@quartata cant mods be given account details from metasmoke?

They can be, and there's a page for it

I am not sure what part you are disagreeing with

It's not that though

!!/watch sweatslimbelt\.co\.in

@K.Dᴀᴠɪs You don't have code privileges, but I've created PR#1674 for you.

tpu by suraj on Calling multiple instances of python scripts in matlab using java.lang.Runtime.getRuntime not working [MS]

Ooookay, probably time to give JG some time to respond. All valid points, but we don't need to fight the "you're ganging up on us" battle too.

Yup, I needed to get that feature request in though. Just took me time to find it. Searching MSO didn't help there.

Merged SmokeDetector #1674.

[ SmokeDetector | MS ] Username similar to website in answer: What is the scope of questions regarding theoretical computer science rather than practical problems? by Tamahagane on meta.SE

I just meant that they have plenty of mod tools to find spammers that need to be deleted, the posts getting deleted faster isn't going to change their workflow

tpu- by K.Dᴀᴠɪs

CI on 58105da succeeded. Message contains 'autopull', pulling...

[ SmokeDetector ] SmokeDetector started at rev 58105da (metasmoke: Merge pull request #1674 from Charcoal-SE/auto-blacklist-1520227812.2381415 --autopull) (running on teward/Solar Flare)

Restart: API quota is 18276.

CI on ceffeb2 succeeded. Message contains 'autopull', pulling...

If their workflow really is to wait until they see the spam then destroy they're missing a lot. Metasmoke can actually help them in that regard...

[ SmokeDetector ] SmokeDetector started at rev 58105da (metasmoke: Merge pull request #1674 from Charcoal-SE/auto-blacklist-1520227812.2381415 --autopull) (running on teward/Solar Flare)

Restart: API quota is 18269.

That was weird

@ArtOfCode Why does this policy exist?

Oh boy

read transcript from March 1

@Undo What happened?

tl;dr a few overreactions in the wrong order

@ArtOfCode TL;DR

@gparyani someone commented on a thing; site wasn't happy with it because it almost lost 'em a user; site got disabled; policy was suggested; policy was a good idea; site got re-enabled; policy is a thing

very succinct!

@ArtOfCode How did the comment read?

[ SmokeDetector | MS ] URL-only title, bad NS for domain in body, bad NS for domain in title, bad keyword in body, bad keyword in title, +6 more: healthflyup.com/test-troxin/ by jgxkuwgem on askubuntu.com

@gparyani shrug

tpu- by K.Dᴀᴠɪs

just the bog standard spam auto comment

please read how not to be a spammer blah blah exterminate blah

part of the problem is the auto comments but eh

which have now been changed to make them less confrontational

yep yep

!!/test earn cash

> Would not be caught as a post, title or username.

!!/test making money

> Would not be caught as a post, title or username.

Think those should be added?

you're going to catch like all of PFF

And a decent fraction of Workplace, probably

New metasmoke user 'Stephen Rauch' created

Spammer went though the trouble of disclosing their mailing address: meta.stackexchange.com/users/383953/tamahagane

@gparyani hmm, did they edit it out already?

@tripleee No, it's in the location field

ah yes

a shame that they've resorted to spamming since it sounds like a decent idea

why did they go to the dark side

although I find it to be somewhat hypocritical if they're going to then charge money

that's not "democratizing" shit

@K.Dᴀᴠɪs No

!!/rev

@quartata teward/Solar Flare 58105da (metasmoke: Merge pull request #1674 from Charcoal-SE/auto-blacklist-1520227812.2381415 --autopull)

huh...

Did anyone get the user count before the meta post? I'm interested in seeing how many new users pop in

still confused by the double CI

@Andy 374

@Andy Not sure where to get that number, but it looks like only two people have joined since: chat.stackexchange.com/…

So, two less than whatever it is now

374 was on Thursday

Er sorry. 371.

there are a few deleted users but just keeping track of the ID is probably easier

@ArtOfCode What do you think of meta.stackexchange.com/questions/138284/…?

Only deleted users are 2 4 and 6 I think