[Foundation-l] Friendliness: a radical proposal -- some proposed details and a diagram

[Neil Harris]

Here are some more details to flesh out my proposal for new admin creation.

Proposed rate of automatic new admin creation: 5% a month, until back to 
early-Wikipedia proportions of admin number relative to edit rate.

Although this sounds a lot, it's only about 3 new admins a day.

---------------------

State transitions:

IP user
|
|  Creates an account, passes captcha test
V
User
|
|  Time passes
V
Autoconfirmed user
|
|   Time passes. User gets chosen at random from pool of all editors, 
followed by machine checking for good participation. The daily rate of 
random selection is tuned to generate the correct rate of new admins 
over the long term.
V
Proposed new admin
|
|   Gets message. Sends a request message to a list. Any "old admin" 
checks for human-like edits, then performs one-click action to issue 
admin bit. If they don't respond within (say) two weeks, the invitation 
is withdrawn, and they have to wait to be be drawn again at random.
V
New admin, with limited powers
|
|   One year passes without being de-adminned
V
Old admin, with full powers

----------------------

Some possible machine-detectable criteria for "good participation", 
based on edits:

* Account age: Has been a Wikipedia contributor for at least two years.
* Recent activity: Has made at least one edit in at least X days in the 
last three months.
* Recent blocks: has not been blocked at all in the last year
* Responsiveness: Has edited a user page of an editor who has edited 
their user page, at least Y times in the last three months.
* Edit comments: Has added a non-trivial edit comment to at least Z% of 
their edits
* Namespaces: Has edited some balanced mix of articles, talk pages, user 
talk pages, and project talk pages, within the last three months

Note that this is a satisficing activity -- the aim is not to find the 
best editors, or to be fair, but just to select active Wikipedia 
participants who know their way around, and are not misbehaving, and 
then select some of them by lot.

The final test, for humanness, necessarily needs to be performed by a 
human being, to avoid the threat of bots gaming the system, but, if as 
suggested above, there are only about three or four candidates proposed 
each day.

Note also that almost this process can be implemented in a bot, 
independently of the actual wikipedia software itself.

-- Neil