I wonder whether I've been a moderator here for long enough
(four months now)
that I can
talk about our decision-making process in aggregate, without
revealing anything untoward about any particular suspension.
How big is the Physics Stack Exchange community?
Well, there are several answers to that question.
I'm not a database person and you're probably not either, so here are
some answers based on public information.
We're one of the most active
Stack Exchange sites:
my count is about
4100 new questions
since the start of 2017, a little under a hundred per day.
If you assume we're maintaining our long-term average of about
1.5 answers per question, that's about ten thousand posts to the site
over the first six weeks of the year.
It's a lot.
There is not any one user that has read every contribution,
and determining overall quality is a group effort.
The quality of a post is estimated by the
voting and reputation systems.
My experience is that most votes occur on posts that are relatively
fresh, so we can use the
list of users with reputation
changes
to estimate the population of the site over a time interval.
I count about 4700 users whose reputations went up
and about 650 whose reputations went down over the past six weeks,
so I estimate a population of five or six thousand logged-in users
visiting in a typical quarter.
Most of them must make one or two posts and not come back,
to get to our ten thousand questions and answers over that period.
We can also look at
voters
(current rough count: 450)
and
editors
(rough count: 65)
since the start of the year.
And what we see is the unsurprising result that, as a participation
task gets more complicated, the number of folks who participate in it
is smaller.
About 10% of our visiting population is actually voting on questions
or answers.
About 1% of our visiting population is involved in making edits, and
a slightly different 1% has
enough reputation
to access the
"10k user" moderation tools.
Close queue reviewers?
This month,
there are about thirty.
So as far as monitoring the site, making good things appear and bad
things disappear, we have somewhere at or around a hundred people.
If you're reading this answer, you're probably one of them,
and you'll probably recognize most of the names of the others.
That's a small group, which matters for what I want to discuss next.
That was the size of the "community", for a few definitions of the term.
How big is the role played by the diamond moderators?
The moderator activity
log
lets me summarize activity
over a quarter ---
which we saw above corresponds to about 8,000 questions and 12,000 answers.
During that interval, the six diamond mods
processed
1200 flags.
Another 1200 flags were apparently handled by the community in the
low-quality queue.
removed 3000 comments (mostly to chat, in batches)
closed 1300 posts, and deleted 500 posts
edited 2500 posts (of which probably 2000 tag-only
edits
--- Qmechanic is an amazing retagging machine)
cast 4000 votes on questions and answers
created 250 posts and 2200 comments
sent 28 private moderator messages, including 10 account
suspensions, of which 6 were for a week or less.
This is a lot of activity to see in aggregate!
(It works out to
typically handling two or three flags per mod per day, which doesn't
feel like much.)
Hopefully two things leap out at you.
First, a solid majority of the twenty thousand question and answer
contributions to Physics Stack Exchange from its user community over
the past three months didn't get any diamond moderator intervention at all.
That's a good thing.
And second: reaching out to a user privately, and especially
suspending a user, is quite rare for us --- about three times a
month.
About half of these suspensions have been for voting irregularities or
misusing sockpuppet accounts; those problems we typically identify on our own.
The other five-ish suspensions
(one or two a month, if you're counting)
were shared among other suspension reasons:
continued low-quality contributions, vandalism or deletion of upvoted
contributions, self-promotion, making other users get fighty with each
other, and so on.
Those are things that happen in public, that are
usually pointed out to us in flags or in chat by our 3k and 10k users.
And the 10k users can continue to see material even after it's been
deleted, if they feel like going hunting for it.
Now, what's the point of issuing a suspension?
Suspensions are about behavior, not about
people.
In particular,
At the end of this timed suspension period, your reputation will be
recalculated, and your account will resume as normal. We don't hold
grudges. The point of all this is to address the behavior. If the
behavior improves, you are welcome back.
An important part of making people feel welcomed back is allowing the
community to forget about their punishment.
But this is a website: literally the only we have to interact with
each other here is by making entries in a searchable database.
Once a thing is said, it's hard to un-say.
We do not issue suspensions without substantial agreement among
the diamond moderators, and the suspension messages are clear about
the (mostly public) activities that led to the suspensions.
But keeping the details private --- or at least, keeping our concise
statement of these details private --- is an important part of making
the user feel welcome to return after the suspension.
What we'd like is to encourage our suspended users to return,
make quality contributions,
and eventually join the relatively small cadre of users who help to
curate the site.
And part of the clean slate that's needed to become active in that
community is for the community to forget about old problems ---
something that's harder if a concise description of those old problems
is just a mouse click away.
So the answer to my first question is no: I can't say much more about
suspensions in aggregate than I could individually.
There have been so few suspensions that essentially any remark I made
would permit you, an alert reader, to figure out who I was talking about.
A commenter asks whether prior suspensions are "expunged" somehow
for purposes of escalating duration (usually a week, then a month,
then a year).
The records of private communications and suspensions are not deleted
from the servers or hidden from the moderators;
we exercise our judgement.
For a hypothetical example, suppose a user had a temper tantrum,
collected a bunch of rude/abusive flags, and needed a time-out.
If their history showed a previous suspension for the same reason in the last
year, we'd almost certainly escalate.
But if their history showed a brief previous suspension for the same reason from
five years ago, followed by a positive contribution record, I don't
think we would hold that ancient history against them.
Where exactly is the boundary?
That I can't say. It's a judgement call, which we'd try to make using
all the information that's available to us.
If a user is active in the community, makes good contributions, and is
encouraging to and respected by other users, we would probably tend to
be lenient.
If a user spends a lot of time testing the boundaries of what is and
isn't acceptable, we would tend to be more harsh.
But I won't pigeonhole myself in the future by inventing some
algorithm here.
SELECT p.Id as [User Link], * FROM (SELECT Users.Id, SUM(Posts.Score) AS TotalVC FROM Users LEFT JOIN Posts ON Users.Id = Posts.OwnerUserId WHERE Posts.CommunityOwnedDate IS NULL GROUP BY Users.Id) p LEFT JOIN Users ON p.Id = Users.Id WHERE TotalVC > 0 AND Users.Reputation = 1;
It is not a perfect query, it detects some false positives, too, but all practically important caged users are there. The list is not much smaller as the same on the SO... but the SO is 15 times bigger site. $\endgroup$ – peterh Feb 23 '17 at 13:51