User:Collect/personas

Early stage of user essay.

Typical number of "total editors in a month" on WP is about 150,000.

Actual number of active personas (accounts) editing more than a single article is about 30,000.

Of which 2,000 are admins, and 6,000 are casual users. about 12,000 are "experienced" editors, but not admins. 10,000 are "multiple persona users." Of those, I suspect the majority are held by admins who do not want to have their admin persona and editing personas get confused, and who earnestly try to keep the types of edits by each account fairly separate.

Some, however, are not used for such a proper purpose -- rather they are used for two or more distinct purposes contrary to Wikipedia policies.

First -- the average "community size" in WP (note that WP as a whole is not a community, but those who are active in editing an article or group of related articles can form a type of community) is under 20 editors. By using additional personas on such a micro scale, two or three editors can effectively create "consensus." With only a few editors on any given article, loading the dice with an extra three or four voices will almost invariably tip the scales. It is virtually impossible to prove any impropriety, other than by circumstantial evidence. Usually the slip up occurs in vocabulary usage, or in times logged online. Really strange styles (like using no caps, or using obvious horrid spelling) are tip-offs that the persona is a multiple persona. One method of "gaming the system" is to canvass editors who are thought to have supported one side in the past. When a "non-canvassed" editor appears, moreover, the probability of it being a surrogate approaches one.

The second type is the "jump in and anger the other guy" persona. The aim here is to get the people one disagrees with to be blocked or just to disappear. If an editor says his primary desire is to drive you off Wikipedia, that is a pretty good clue. Or if he says he has been a gumshoe on your case.

Signs of a "multiple persona":

Setting up a user page and blanking it so it will not be "redlinked." While a great number of multiple personas just do not care, this is a sure sign that the account user has more than one account.
Having a very restricted edit history - usually focussed on perhaps ten to fifteen articles, generally in a fairly narrow field. Very few casual users have such an edit pattern at all. And none of the genuine casual users will suddenly pop away from their small editing area. Usually their edits bear a relationship of some sort either to each other or to an acknowledged background of the editor.
Where amy page is suddenly placed up for deletion, the appearance of a new entrant placing a single "bullet vote" on the issue is highly suspicious.
Where any person with a limited number of edits suddenly appears on an unrelated ANI page, the account should be considered suspicious.

WP is neither the first nor last online venue for multiple personas, but I was shocked at the apparent percentage of such.

Several editors have, or have had, extensive directions on how to set up alternate accounts on their userpages. Without encouraging in any way such activities here, the directions include statements about avoiding interaction with a primary account except when needed, making several hundred edits on miscellaneous pages without being noticed, keeping the account for more than three months before interacting with yourself, deliberately using the account during hours where you have not generally been active, and using an "IP anonymizer" to prevent a normal "checkuser" finding the connection.

Examination of substantial editors (roughly 60K edits each) shows a few interesting traits - out of over two hundred full sets of intersection counts, I found a total of two where the number of intersections on user talk pages outnumbered the number of intersections on articles. For the 90th percentile, the ratio was well under 50%. In zero cases were there any overlap of greater than two on userpages. In zero cases did the intersection on article talk pages hit the 20% mark, while the 90th percentile was with under a 2% overlap. Any intersection characteristics of editors with fewer edits should not show greater results for intersections if the editors are fully unrelated.

Additional source material via en.wikichecker.com