To your point that users continually providing low quality content are likely to end up banned even if some of their questions get edited/upvoted, the main issue there is, from everything that I can tell, upvotes are weighted much more heavily than downvotes

even a single question that attracts a reasonable number of upvotes will allow that user to post quite a few questions that get rather heavily downvoted before the two have evened out.

oh, Tavern ain't good enough for ya?

;-P

so if even (and I know that these won't be exact numbers here) something like 1/4 of the user's questions get fixed up by others, they'd keep their head above water.

@Servy not really

So... I probably shouldn't link directly to specific users

agreed

But, I'm looking at someone who has, to date, asked four questions

First one, deleted with -5

second, answered with +2

third, answered with +4

He just asked another question a few minutes ago. Reasonably well-written, but a duplicate.

He would currently need to wait at least 1 day before asking another one.

He had to wait several days to ask this one.

what's the average time between his existing questions?

is he usually waiting a few days anyways?

well, he's had to lately

Usually only the very worst contributors are asking more than a few questions in any several day span, even if they have the choice

First was may 10th, second Oct 21, third Oct 24...

the rate limiting certainly helps with the very worst of offenders

It does, but that's not even the primary purpose.

Banning helped with the very worst of offenders

The basic idea is, if you make folks wait, some of them will start to treat it like what it is: a limited resource.

Folks can still get banned of course, but even there we're looking at changing it to... a really really long rate-limit.

I almost think that most of the rate limiting that we put on people now, when the've done just a bit wrong, is pretty darn close to what the rate limiting should be for everyone at all times

Like, 1/year

people asking great qeustions don't ask 5 questions a day

well, it varies

And that's one reason why we added the newer question badges - folks don't tend to notice consistently good askers otherwise.

You remember the guy who asks six terrible questions in a day

You forget the guy who's been asking one good question a day for 4 years

true, although people don't look at anyone else's badges very often

they usually just see what's in the user card

yeah, we need baseball cards

lotsa stats

YOU'RE ASKING .400! WOOO!

lol

I know just out of my personal experiences the very best questions almost always come from 1 rep users

with rare exceptions

I've yet to see someone with 400 questions that had questions better than mediocre

well... An awful lot of questions come from 1-rep users. Which is a big reason why content analysis / triage is so important

yeah, which is why it's not particularly harmful for me as an answerer to actively avoid answering questions from people with more than say 1k rep

a 1 rep user is rolling the dice between great and terrible

the people asking a question several times a week tend to be consistently marginal. Good enough to not usually warrant downvotes/closures, rarely actually interesting

might depend on the topic there too

sure

As a programmer though I just find it actually quite hard to imagine coming across several technical questions I couldn't solve using basic Google research each week, week in week out

and I say that having been working in a technological stack that has abysmal documentation and very marginal existing resources online (particularly several years ago)

most of those users tend to be people with great question asking skills, but to which the actual underlying question isn't terribly interesting

There's definitely a difference in people, in how folks learn.

I generally learn more answering than I do asking - but then, I've learned an awful lot because someone else asked at some point in the past.

Which is where you get a big difference between topics. Someone asking questions about a newer language / library might often hit a lot of areas that haven't really been documented before.

The questions themselves aren't necessarily all that different from what folks are asking elsewhere, but the answers are novel

I suppose I just consider cases like that exceptional enough that I'd have the rate limiting systems inverted a bit

default to reasonably strict rate limiting and just stop counting "hits" as it were

rather than starting people off at rate limits that are more or less just there for overt spam and moving down as there are problems

Well, then you get into a bunch of unpleasant edge cases where you're implicitly encouraging bad behavior.

by that I assume you mean creating new accounts?

Or editing existing questions to repurpose them

true

Think of the case where someone asks a question and immediately realizes they didn't need to.

Happens all the time: they see the answer in the related sidebar, or just figure it out by "talking to the duck"

But, now they have a different question

The proper thing to do is to delete the one they just asked (or self-answer it if there's a useful answer)

and post another one

so cases like that is why I'd want to have rate limits not be time since last question, but rather have it roll, i.e. you can ask X questions over Y interval of time

adjusting X and Y based on past behavior

which is how it works, to a degree

There has to be a hard limit of some sort, just to prevent abuse

sure

I think my main point is just that this type of rate limiting, while it obviously needs detailed tuning, is something that should be there by default, and even without a demonstrated history of bad behavior

and that once you have started to build up a history of bad behavior it makes sense to start moving beyond just rate limiting to rates more in line with what is expected

I know one useful tool that you guys have said you're looking into is trying to affect the visibility of questions

i.e. through the homepage

seems hard to get right, but the type of thing that could be more useful

if done right

mind if I ask what's happened along that front since the idea was mentioned a few months back?

@Servy well, there's where that content analysis thing comes into play

if we can get at least a rough idea of what a question is like immediately, we can throw a subset into a triage system

good stuff goes straight to the home page

awful stuff gets closed/deleted

meh stuff sits in limbo until/unless someone edits it

Note the emphasis on editing again

a non-trivial number of questions aren't... awful

but they don't need to be promoted unless someone is willing to take the time to polish them up

is, and if so to what degree, a user's past history used to determine the "rough idea of what a question is like immediately"?

I know a lot can be done by looking at things like word choice, use of code, length, etc., but one would think a user having a lot of closed/deleted questions could at least be another factor

@Servy it helps considerably

moving an otherwise "meh" question into the "probably needs closure" perhaps

but

it doesn't exist in a great many cases

of course

By the same token, rate-limiting from the start sounds good, but... Until you have at least one question, there's nothing to limit

apart from the per-IP limits that already exist

but one of the nice things with that system is that a user with enough bad contributions to be of concern, but not enough to outright ban, isn't encouraged to make a new account, make abusive edits to other questions, etc.

correct

while still avoiding giving them all of the attention of a normal new question

That's a big hope for this - that if we get it right, an awful lot of questions will just sorta float on by, many eventually being deleted automatically, without clogging up the front page or the close queue

Which brings me back around to the concern over editing - the notion that someone could be pretty bad at asking questions, but do well on the site because they always get edited

That does happen

it also allows us to control which questions are more likely to get the attention of reviewers willing to spend the (usually considerable) time to actually save a question

Even today

rather than just leaving it up to luck

But... It's rare

It's like the guys who make it through life by winning contests

You hear about 'em and think "man, what a charmed life, they don't deserve that shit"

but of course, there are many thousands who try to do that crap and never pull it off

worrying about a lucky few isn't really worth the effort

And, as you say, we could do a whole lot better about not leaving it to chance

I suppose the one risk is that the few questions that do get saved result in lots of upvotes, giving the user a much better history, making the algorithm think they're better at asking, making them more likely to have their next question get reviewer attention

creating a positive feedback loop for negative content

it's just a trap the algorithm needs to avoid

which seems like a hard trap to not step in

Basically the algorithm needs to figure out how to ignore the effects that it itself creates

Yeah, so... The other part of this is taking VLQ flags out of the Low Quality queue (where they're borderline useless) and feeding them into triage.

I think you might just end up with the same problem again

the VLQ queue was once just a place to find crappy content and go fix it, but people didn't actually want to spend all of the time it took to actually do that to up their review count by one

That's the point: triage can't fix your question. All it can do is re-cateogrize it.

so the queue ended up feeding in posts that could be "resolved' quickly an easily

okay, that does sound helpful

now we just need to deal with the problem that the average quality of reviewers doesn't seem to be where (I imagine) either of us would hope

If a post escapes crap/meh categorization because we miscategorized the user, or our content analysis sucks, or both... Then that breaks us out of that loop and forces reevaluation.

@Servy asking reviewers to make hard decisions usually doesn't end well

agreed

Asking multiple reviewers to make relatively straightforward decisions (and the acting on consensus) works out ok

so a triage queue of asking if a post is trash, meh, or great, seems actually kinda hard

This is where the LQ queue works so much better for answers than it does for questions

I'd expect to see a lot of meh, meh, meh, meh, ...

@Servy me too - in fact, I'm banking on it

if the majority of posts going into the queue are not very good, and the majority coming out are put into limbo...

the idea being that anytime you see a "trash" you can really count on it meaning something/

?

...Then we're doing what we set out to do: filtering out a whole bunch of mediocre questions and giving them less attention without wasting a ton of time on the home page or close review.

The big opportunity for failure is if folks default to "Ok" or "Delete!" - then we're back to crudding up the homepage or filling up close review

I suppose you can push people in that direction through the UI

by making "meh" be one click, but requiring additional clicks (through a reason selector) for either "trash" or "awesome"

let's hope

the "delete" would probably look fairly similar to the flagging reason dialog, but then you'd probably need to come up with some sort of barrier for people marking a post as awesome

maybe

maybe give people like 2-3 "awesome" votes per day, out of a review queue of 20

folks are reasonably resistant to saying something looks good

basically you get to give extra attention to just the few of the post you think were the best?

If you require 3 people to agree on it, then the chances of something looking good are going down significantly

But the truth is, we won't know until we've seen it in action for a couple weeks

also true

as with LQ, the queue will tend to self-select for folks looking for crap

I like where this is going though, is this whole thing currently your pipe dream, something SE is seriously planning to implement, something currently being worked on, or what?

various bits of all three

Hoping to have something up on meta soon

looking forward to it