« first day (79 days earlier)      last day (2862 days later) » 
00:00 - 20:0020:00 - 00:00

12:00 AM
Alright, so I got some feedback on the flag decisions.
 
that was quick, considering I asked and got a "I'm busy atm"
 
Well, in honesty they were already talking about it so the majority of my questions were answered when I set foot in there.
 
TL?
 
probably the arqade room
 
Anyway. So, it seems that the biggest beef is that not only was it a giant wall of flags, but it looked very clearly like it was entirely just a wall of "Every comment containing the word 'edit'".
 
12:02 AM
that's not actually completely true, I flagged every comment on each post that had a comment containing the word "edit" or "thanks"
if it just so happened that was the only comment that needed a flag, then so be it
 
Which isn't always something to action on, and having them sift through the mess is just a bother. As well it disregarded context, so for example an exchange that goes like "Hey, did you account for X?" that is responded to with "I have edited to account for X", deleting the latter only leaves an unanswered comment, while deleting both may remove potentially useful data (sometimes, the inquiry for clarification is signal of its own).
 
@GraceNote I also didn't flag all the things that were on the list
 
As I expected, in some cases near the end, actions were taken near the end even though the flag was rejected because even though the approach to flagging wasn't agreed with, there was still things that should be done, so action was taken but the decline was indeed to send the message of not to take this sort of "shotgun" approach to flagging.
 
very well, today I will flag 100 comments and 100 NAAs
 
@GraceNote Maybe they should have done a custom decline reason...
 
12:07 AM
@hichris123 There's no custom decline for comments.
 
Oh, then that makes sense.
 
It seems to me like comment flags are neglected by everybody - Stack Exchange and the moderation team
and that perhaps some work needs to be done around them?
 
Comments are low priority compared to most things. The highest priority dealings with comments, which are spam and offensive, can be handled swiftly. It's by design as it were.
I'll reiterate that doing 100 flags, especially after having the selection of declines and that I've delivered the message behind said declines is in fact related to the bulk and the methodology behind the flagging, is not all that helpful to actually getting things actioned on.
 
It was a joke, I'll submit like 25 at a time
so that nobody has an "oh crap! FLAGS!" moment.
but everybody on these teams should know how I am by now, I have no issues doing 200 edits in a day, they shouldn't be surprised when I submit 100 flags.
 
12:24 AM
I don't think the two are quite comparable since one is an action type that has its results performed, while the other is a request for another's action. What makes for a comfortable and manageable volume differs.
 
Just throwing this out there, comment flags are kinda neglected and need some serious work. I.e. 10k handling or something.
4
 
49
Q: Make community comment moderation a more public process

PëkkaComment flagging could be a community tool against the often rough tone on SO. However, the current comment moderation system isn't suitable for the purpose. It's not useful for anything but the removal of really outrageously offensive stuff. That's because it's almost completely opaque: There ...

 
@Undo probably 20k but yeah I agree
 
There's the feature request we were speaking about earlier
Well, actually, a separate feature request than what I expected
 
the problem with feature requests is that it's time, and a lot of time there is no official response to the requests
posting a feature request and getting no response is essentially just being ignored by the people who can do anything about it
 
12:31 AM
@GraceNote Yeah, I thought there was a different one...
 
It's fun when you're one of the people that usually reviews these things and you still get no response.
But better to have something out there. We check on these things. Things which are long old, not much spoken of, or even declined, we do get to implementing.
 
@ManishEarth Swagmail has arrived.
 
@Undo swag mail ?
oh, swagbox
I knew it would (TL magic)
Jan 31 at 18:40, by ManishEarth
You can do this one in every 30 times and still have 97% positive feedback.
What'd you get?
 
@ManishEarth He's posting it in the Tavern.
 
@GraceNote can you give me a screenshot of what the interface that these comment flags appears in looks like?
It may influence how I flag
 
12:45 AM
A link to the post, and the singular comment that is flagged is provided alongside what comment flag was used.
 
do mods have to click the link to the post to do anything about it?
or are there buttons (delete, etc) inline/underneath
like in the 10k tools
 
Technically, no, which is why it's great for things like spam/offensive that are pretty easy to sight-check.
But for anything that isn't straightforward, they'd have to click. Multiple times, too.
Oh, dinner time
 
enjoy
 
Even battleships need maintenance.
 
@GraceNote /keeps forgetting that the username refers to a battleship and not the musical note-thing
 
12:53 AM
@GraceNote's seemingly random terminologies have always confused me
 
1:14 AM
@ManishEarth Well, also, my current avatar is a battleship
 
SLEEPY TIME
night
You can perform this action again in 5 seconds - retry / cancel
I have totally got to get rid of some of these rooms I'm in
 
 
3 hours later…
 
3 hours later…
7:15 AM
Arqade mods ^
obviously they didn't like me fixing all of the things they said were wrong
cc @GraceNote
 
 
2 hours later…
9:38 AM
6
Q: Obsolete comment flag handling

kalinaWe've got well over 5000 obsolete comments on Arqade. Since these cannot be handled using any of the user level tools, I have been flagging these so that they can be actioned by a moderator. Unfortunately, our moderators don't appear to appreciate this effort in keeping the site clean? All of th...

 
@ManishEarth damn, that one disappeared quick
 
 
2 hours later…
@kalina Those weren't the Arqade mods.
 
SE invalidated them?
I think you should get involved in that meta post
 
Yep. Probably after I get some breakfast, if we can be patient.
 
apparently I'm on my third warning and potentially facing a suspension, after getting an aggressive and insulting meta post from a mod that claims I'm doing this for the flag count and to cause trouble
even though it's provable that cleaning stuff up is basically what I do
yeah take your time, I'm a little upset and think this is being handled wrong but I'm handling myself
 
12:41 PM
Now the mod team on Arqade are firing personal attacks at me for trying to get a concise answer
 
1:18 PM
@GraceNote I will bring this up here instead of there until I can compose myself to write it up without all of the emotion, but as you can see from the screenshot in my meta post, I only used the script as a basis of where to start after you said "don't flag everything with edit in it" (paraphrase)
So this comment:
"Do not flag based purely on naive keyword searches. If I thought a simple keyword search was sufficient to remove these comments, I would remove them en masse and cut out the middlemen."
It's not completely valid, when you look at how many flags I raised per question:
I checked every single thing I flagged lastnight after our conversation to ensure I didn't leave any instances where I was flagging "edited" of "please add this into your post", like we discussed
 
@kalina Be sure to keep calm while discussing this with Grace
 
and I went through every comment on every part of the post for each post I looked at to make sure I got everything in one go
@JanDvorak I'm nearly always calm when I speak with Grace
I'm pretty sure deep down Grace knows I'm not here to cause trouble
and even if that's not the case, does a good job of hiding the bias
Only a small number of the posts I flagged in the early hours of today were generated by the scripts
the rest were all manually chosen
so... with that in mind, is this purely a case of numbers, and can you give me a number that I can work with?
or is that number 0
 
I'd hold off a day or three, let them get the new mods in and oriented.
 
I'm not even sure I'm talking about Arqade now
personal attacks and threats from moderators without any consequences to them is seriously putting me off
 
s/ors /or /
 
1:29 PM
@kalina Tim is dealing with that end.
 
1:41 PM
@kalina The best thing to do, right now, is just enjoy the sites. You mentioned poking around the network to see if we had stuff in other areas of interest - that could be fun. Answer some questions on Arqade, that's always fun - so is helping new users get to learn the ropes. The obsolete comments really should be part of a more organized cleanup, unless you just happen on them opportunistically.
 
@3ventic The review process being lost is a fault here, and part of that is because of the limitations of the flag queue system. If you pick out of a list of 500 comments that all have specific keywords, the moderators still get a list of 50+ comments that all have specific keywords, which doesn't give any illustration that the situation has changed. — Grace Note 33 secs ago
This is basically what I was meaning to say before I got pulled by FEichinger to discuss comments in general. It isn't visible that you did any change to your flagging pattern, and even when we reviewed the flags, it still looked pretty much the same. That's why we acted as we did - the changes to your procedure are entirely invisible to us and we still see the same thing we saw before.
 
Ah, I understand now
I have too many chats open, I keep losing track of where I was talking last..
 
@kalina What I would recommend is that you take those scripts, take the numbers you get out of them, raw and post-analysis-filter, and make a list of the comments and post it on Meta or somewhere else. Bonus points if you group comments of similar ilk together with reasonings behind their suggestion for removal.
 
@GraceNote I would have expected it would be quite obvious, since before it was one flag per post and after it was 3+ flags per post
 
@kalina Comment moderation tools (even on the mod end) are hardly optimal. In fact, they really stink. This is our fault and we need to fix it.
 
1:49 PM
@kalina It wasn't obvious.
 
@GraceNote does it look obvious from my screenshot?
 
Nope, because when I look at the posts in the screenshot, I see a number of comments containing "edit" or "thanks" equal to the number of flags you raised.
 
@GraceNote you don't see that they're all grouped together by the post they were on?
whereas before they were one flag per post?
 
I'm not sure how that changes anything about it still being a big blast of keywords flagged.
 
because it shifted from 80 flags = 80 posts to 80 flags = 20 posts
which is one of the things you mentioned as an issue
I'm starting to feel like I'm just destined to be perpetually misunderstood
 
1:56 PM
I'm feeling quite misunderstood, too, because "Number of flags per post" wasn't remotely the issue.
The issue was "Flagging a bunch of things based on keywords, enmasse". 80 flags on 20 posts is still 80 flags. They're bunched up in the UI but it's still 80 flags.
 
it was one of the thing mentioned, you mentioned orphaning comments, the second pass made sure that didn't happen
with a greater portion of the things flagged not even related to the keywords
 
It was one thing I mentioned in a series of lines explaining why this kind of flagging isn't helpful.
Again, I'll assert that when I took a look at the list just now, I saw an equal number of flags to the number of comments that contained keywords. If there were some in other parts that didn't contain the keywords, that got lost in the mess caused by the first, which brings us entirely to why this approach is bad in the first place.
 
 
2 hours later…
3:43 PM
@ManishEarth Your little app makes me wonder if we should block all caps titles in the quality filters. I'm just sayin' - I didn't realize there were so many of those coming in
 
@TimPost I'm for
they aren't that many. Fewer than 1/hr
 
@TimPost all caps should be banned in the quality filter
 
WHY DO YOU SAY THAT?
 
@Grace looking at one of your latest answers, are you saying obsolete flags can stack and auto delete posts?
and if so does that follow the same 3+(score/3) rule?
ie: 3 for an unvoted comment, 4 for a +1 comment
 
Correct.
 
3:51 PM
there will be a new clean up post up by the end of the day, then
I'm right in thinking that during said process, until the Nth flag is cast, these will sit in the mod queue, yes?
but that, once the Nth flag is cast, it will then vanish again
 
Correct.
 
wait, are you going to synchronize flagging from N "users" using your script?
 
Good that you're doing a post to organize this - it's kinda hazardous to work with comment flags in a sort of organized front without a physical plan ready that can be reviewed and previewed.
 
@TimPost haha good point
Maybe I ought to add some machine learning magic to it -- monitor the questions that get closed on some sites and build up from there
But that would be hard
Also, making a markov chain impersonator-bot would be so much more fun :p
 
@ManishEarth If you contribute the memory and persistence, I'll be glad to contribute the algorithm
I'm thinking in terms of: for each n-gram, how many spam-posts and non-spam-posts contained that n-gram?
 
3:56 PM
@GraceNote this is all I ever wanted to happen, all of this drama from today has annoyed me and some of the comments made have seriously upset me when ultimately I just wanted to clean shit up.
@ManishEarth would you be ok with us using charcoal for this cleanup effort?
 
then convert that to some sort of significance weights (think Bayesian filters) and sum up.
 
@JanDvorak Persistence I can do, at least for a few months. How much memory?
jibber jabber jargon jargon probability blah blah
:p
 
Hm... how much memory does the list of all 6-grams take?
 
what's a 6-gram?
 
I just read that as "how much money for 6 grams" and had to pick my jaw up
 
3:58 PM
sequence of six consecutive characters from a post
 
@JanDvorak I was thinking in terms of closed posts as well, but yeah, best focus on spam
 
@ManishEarth depends on how you train it
you could even have two separate sets of weights for spam and closure
 
@Undo In light of the recent posts, why not speed up work on my initial idea of converting Charcoal to a "better place" for comment moderation, where normal users valid/invalidify things and mods do a second check and can script-delete
> We're actually thinking of blocking this script from functioning any further on our API.
@GraceNote If you want I can disable the app for now
 
@ManishEarth Grace meant the comment search engine. The spam search engine is awesome.
 
@JanDvorak I'm talking about the comment search
 
4:01 PM
yeah... stop it
 
I don't have access to the server, but I can force a pull to a locked down version
 
do you have a line of vision to the server at least? If so, finding a gun shouldn't be hard.
 
Yeah. That would be helpful.
 
Ok /doing
 
@ManishEarth how much memory can I get? Say, 256MB?
 
4:06 PM
@JanDvorak do you have a Charcoal login?
Just need to test if it still works
 
> Sorry, Charcoal is locked down at the moment
 
@3ventic thanks
 
@ManishEarth nope.
G2G now, bye. I'll start working on the trainable filter within the smoke detector when I get home.
 
Alrighty :)
@JanDvorak hmm, not sure
@hichris123 @Undo Charcoal is now locked down (in case you don't get github notifications)
 
.oO( I get the feeling Manish is unhappy with me )
@GraceNote now that a solution has been found can we lock that Obsolete comment flag question?
 
@ManishEarth Still around?
 
yep
wait no
 
oh ;(
 
[ SmokeDetector ] Master is not around right now
:p
Yeah, I'm around. What's up?
 
@kalina That doesn't even remotely relate to normal processes on these.
 
4:21 PM
@GraceNote it's to stop it from becoming a thing that ends up with further commenting for days to come
I don't mind either way
 
(hope you don't mind the lockdown, found it the best thing to do at this stage -- we can sort out what to do with it later)
 
@ManishEarth No, it's right.
 
yay
 
It makes me sad, though.
 
yeah
But I think we can now improve it to a more useful and less mod-annoying stage
 
4:23 PM
@ManishEarth So is this a primarily-Arqade problem, or... network?
has hard time parsing transcript
 
Network
 
ok
 
primarily arqade imo
 
sigh
And I thought we had something great here. Drat.
 
@kalina The warnings and concerns about this were voiced before Arqade even had an issue.
 
4:25 PM
@GraceNote Really?
I knew about the 'don't do it to annoy Diamonds', but...
hadn't heard anything else.
 
You were the one who acknowledged that Shog9 poked you about this.
 
Yessss... but I guess I didn't make to connection to a concern
 
Which dials back to before Arqade, and is on a network level. Ascribing it to be an Arqade issue is short-sighted and unproductive.
 
FWIW Undo has been stopping when asked and after that was pinging before flagspamming
 
@GraceNote true
 
4:26 PM
But that doesn't apply to the other users of the app I guess.
 
I suppose.
Umm @ManishEarth I can still log in ;(
 
@Undo Not locked down for you, me, or Chris. We'll need to be able to improve-test it.
 
Ah, ok.
I do think, though, that we've uncovered a huge flaw in comment flags at least.
 
Yeah.
 
It seems a little hacky to have passing lists to someone as the recommended method.
 
4:29 PM
Well, that is for mass-cleanup
Passing lists has always been the mass-cleanup solution
Mass edits -- on Phy meta, we made meta posts and as a community edited them, taking care not to flood
 
For any mass job, really. Lists help organize things, allow people to review contents, and track how progress is doing in one neat place.
 
So we need to pretty much quit mass-flagging comments.
 
Lists
 - help organize things
 - allow people to review contents
 - track how progress is doing in one neat place
 
hah
 
@Undo not a total loss, I can still use it to mass-delete comments on Phy
I used to sometimes do that via data.SE, but it's hard
The tool is an awesome one for bored mods
 
4:31 PM
Yeah, but that means we need to track down bored mods.
 
Note here that the trick is that he's using the system to take direct action.
 
And it also means I can't use my own (not really, but kinda) tool.
And the SO mods never get bored...
 
@Undo pretty much your own. I made the javascript flagmagic, that's all.
Though ... that's the thing that caused the problem :p
 
Meh, it's open source. No one "owns" open source :P
 
hah
 
4:32 PM
@Undo We'll see about that!
 
I think FOSS code is still owned. Just accessible
 
@GraceNote tries to figure out what this means
At this point, it seems almost like the best thing to do is to integrate it into the mod tools.
 
It will probably have to evolve into something requiring a mod step, yes
 
Which makes me sad ;(
Although having worked on something that's in the mod tools would be really neat.
 
Technically what you've made is already in the flag tools, just without the comment search feature
 
4:35 PM
has no idea what Manish is talking about
Oh, as in the flag queue?
Yeah.
 
Yeah
The only thing missing is that the flag queue can't SEDE
 
But they have access to the DB, which is a whole lot better.
 
@Undo "So it has come to this" isn't the only phrase that can be tossed around to suddenly shift the tone of a discussion.
 
hahahaha
 
Ah.
heh
I'm still a little shocked that I woke up in the morning to an email from GitHub with the title: Charcoal-SE/Charcoal: Lockdown #3
:P
I'm still torn - I want it to be a mod tool, but I still want to use it.
 
4:51 PM
Solution: Become a mod!
 
Not going to happen anytime soon.
 
With some retooling, it doesn't necessarily have to be a mod-only tool.
I don't know what such retooling entails yet, but there's potential.
 
I just can't imagine how we can retool it without making it very very targeted to privileged users.
 
Well, part of the discussions at current have been pushing forward the idea of somehow opening things to 20k users.
It's also a thing to take into account that it might not just be retooling on your end, but we can also retool things on our end.
 
I fully support that.
 
4:55 PM
What you have is a good idea going. In its current state it's problems, but it, like most endeavors in this room, were done with positive intent.
 
5:13 PM
@ManishEarth So mSO post about Charcoal?
 
:o Charcoal's made in php?
 
@3ventic Yeah.
 
That's something I'm good at
 
@Undo dunno. your call
 
@ManishEarth Well, shog said so. We obey the great Shog, so...
I vote we wait until @hichris shows up then make a Google Doc draft thingy.
 
5:16 PM
I won't be around as much
@Undo just about the Charcoal app, or everything under Charcoal-SE?
 
Aww.
 
I guess Cinder isn't too harmful
 
@Manish It's /Charcoal that's causing problems, right?
 
yes
 
Cinder doesn't even work yet :P
 
5:17 PM
SmokeDetector is a bit annoying, but most people like it, and the spam is disappearing
 
It's not really in the same league as Charcoal either.
 
yeah. It's headless, meant to be run by one person
 
Anyway, I have to go get something done. I sure hope we can pull out of this low ;)
 
@Undo not really a low in my eyes. I had sorta forgotten about Charcoal, much more interested in Cinder and the rest
 
6:05 PM
@Undo next up: an app that aids in tracking robo reviewers
No, I'm not writing it yet, but have always thought of one
Because when I find a robo reviewer I usually find more by looking for folks who vote similarly
 
@ManishEarth Sounds like a good idea, but what/how would we track?
People with weird approve/decline ratios?
 
6:37 PM
@Undo basically, you find one "possible robo-review", and then the script looks for more by the users who approved it. It finds more bad reviews that way and propagates
 
@ManishEarth So we would have to screen-scrape, right?
 
something like that
 
Wait.
@ManishEarth SEDE has a suggested edit table!
 
yeah
We can start with stuff like backtick edits and generalize
 
hmmm
me likes it
 
6:40 PM
approves
 
You take a seed bad review, and look for controversial reviews (multiple decline+accept on on review) by the same user. Ask the operator to validate the review (were the approvers wrong?), and if so, feed that into the machine too
Don't know exactly what the logic would be like, but I have spent a lot of time doing automate-able review bot digging
Usually it branches out too quickly
 
I bet we could get minitech to use it.
 
Maybe it could be a set of tools instead of an autobot
 
Yesss... me likey...
plays with SEDE
Are suggested edits in the data dump?
 
The current tools don't have a "approves on a rejected review" statistic
We can add that
Dunno
 
6:43 PM
They're in SEDE
Me wants them locally.
 
I don't think the review system is there, though
 
No.
 
Suggested edits != s.e. reviews
 
The SuggestedEditVotes table is what we're after.
 
I was about to say that
 
6:48 PM
hm
k
What about the other queues?
Though suggested edits seem like where we want to be
 
Yes, suggested edits are where the robo-reviewers that I've seen tend to hang out.
@ManishEarth how do I do a LIMIT 10 in SEDE?
 
select top 10 *?
 
Ah
 
not sure
one is mysql, one is ms sql
 
that works.
thanks
Ok, this is interesting.
The suggested edit approvals are stored as upvotes and rejects as downvotes.
 
6:51 PM
ah
 
I'm currently downloading the CSV of all Physics.SE suggested edit votes.
 
@Undo we don't have robo reviewers :p
well, some, but the tools are sufficient
 
@ManishEarth Are you sure?
:P
@ManishEarth So where should we get them from?
 
very few reviewers
 
ah
 
6:53 PM
@Undo SO :p
all of them
db dump
 
@ManishEarth I don't think they're in the data dump.
looks
 
we need the revision text though
@Undo everything on SEDE is in the dump
 
@ManishEarth Do we?
Just a minute.
connects big drive to closet-server
tries to remember pmount syntax
there.
@ManishEarth I can't find suggested edits in the data dump.
 
00:00 - 20:0020:00 - 00:00

« first day (79 days earlier)      last day (2862 days later) »