[RCom-l] Statistics behind Wikimedians with permissions

[Milos Rancic]

I am interested in "cheap", quantitative analysis with just a couple
of notes about the structure. So, nothing should be based on
qualitative analysis of the content.

My premises are:
* Special permissions (even "rollback" permission) motivate editors.
* Loosing special permissions demotivates editors. (I am here
particularly interested in temporary admins of smaller wikis who
didn't show up the second time to ask for temp adminship again.)
* Users with special permissions influence workflow on projects.
* Thus:
** As there are more active editors with special permissions "per
capita", as more work has been done.
** During the periods of losing significant number of editors special
permissions, projects have decreased

To do so, we should answer on a number of questions, like:
* The first one is pre-question: It is possible to create
classification of the projects based on their phase. Projects with
significant number of editors are easy to classify. Smaller ones are
harder to define, but I think it is possible.
* Is retention of editors with special permissions higher than
retention of other editors?
* Is their activity higher after getting permissions?
* Is there relation between the period of time of getting two
permissions (rollbacker, then admin; admin, then bureaucrat) and their
activity? In other words, do editors need to get special permissions
from time to time to stay active? (If that's true, then granulating
permissions should be general suggestion. For example, making blocks
different permission from regular adminship. That has Portuguese
Wikipedia, by the way; but that's the result the internal problems of
their community.)
* Do Wikimedians with special permissions influence development of
wiki and how? (That's not the qualitative question, but quantitative,
as well. Do we have more edits per editor or more editors or similar
on projects with more editors with special permissions?)
* What is correlation from project to project between the project
phase and other parameters? Is it possible to make project subgroups
based on that correlation?
* What is the impact of flagged revisions and their localizations
(en.wp has different implementation than de.wp) and other parameters?
* What is the impact of introducing auto-confirmed users and other parameters?
* .. Are there some other correlations?

The most complex question, but in some cases not so complex (small
projects are under general policies) is:
* Did some policies have impact on behavior of those users; if so, which?

If some of all premisses are true, then we should:
* Educate communities, Meta community and stewards about the
conclusions. Give advices to the communities how to avoid problem
based on their phase.
* suggest particular policies for particular time of development;
* install permanent watch all over Wikimedia projects.

On Sat, Aug 20, 2011 at 21:04, Steven Walling <swalling at wikimedia.org> wrote:
> This is the kind of thing the summer researchers in the Community Dept.
> might be able to whip up pretty easily. Are you interested in all special
> permissions, or just certain ones?

I am interested in all special permissions.

> Getting different rates of retention for these permissions groups is
> relatively easy. Answering "do Wikimedians with special permissions
> influence development of wiki and how" is more qualitative and would take
> time to do comprehensively, though I think we know the answer is that admins
> in particular have an enormous impact on retention of other groups. As for
> groups like rollbacker, I would think that they have a slight but perhaps
> still measurably different impact, primarily because getting reverted with
> an impersonal explanation has a distinctly negative impact on whether people
> stick around or not. My theories aside, different permissions require their
> own individual analysis.

About the question "do Wikimedians ... and how?" -- as I said, I am
interested just in quantitative answers at this moment of time. My
premise is that more admins per capita anyway positively influence
projects. And if true, that should be quantitatively visible.

We should add other premises, including your own, as well and test them.

> I agree that working with a smaller wiki (and thus a smaller dataset) would
> be a good place to start, not least of which because it's likely that if
> there is a disproportionate impact from people with special userrights, the
> effect would be amplified on a smaller project.

We need anyway large communities to detect phases and to compare
smaller communities with them. Statistical behavior of smaller
communities is more hectic. For example, there are a lot of influences
of founder(s). But, it could be about ~10 important editors who create
specific dynamics.

> We currently have one researcher tasked with starting analysis of Portuguese
> Wikipedia, which though it's quite old/mature has only 30ish admins. I think
> I'll ask him to work on this kind of analysis soon...

Let's do the next: Goran will do that with my suggestions or
"mentoring". We will inform RCom and community from time to time; and
Goran will cooperate with the researcher of Portuguese Wikipedia about
common issues.