[Wikimedia-l] Thanking anonymous users

Mark delirium at hackish.org
Thu Jan 16 22:02:30 UTC 2014

On 1/14/14, 5:56 AM, Tim Starling wrote:
> On 14/01/14 15:38, Marc A. Pelletier wrote:
>> On 01/13/2014 11:20 PM, Tim Starling wrote:
>>> The English
>>> Wikipedia edit rate has been declining since about January 2007, and
>>> is now only 67% of the rate at that time. A linear regression on the
>>> edit rate from that time predicts death of the project at around 2030.
>> That's...  come /on/ Tim!  You know better than to say silly things like
>> that.
>> The abuse filter alone could very well account for this (the prevented
>> edits and the revert that would have taken place).  :-)  I used to do a
>> lot of patrol back in those years and - for nostalgia's sake - I tried
>> doing a bit over a year ago.  The amount of "surface" vandalism has gone
>> down a *lot* since.
> Reversing the decline in editor population has been a major strategic
> priority of WMF for many years. You are saying you have never heard of
> it before? Well, here is some reading material for you:

I have heard much about the strategic priority, but much less about the 
rigorous data analysis. In particular, I have yet to see a demonstration 
that there is actually a decline in what we might call the "productive 
editor" population, people adding things to articles or otherwise 
improving them. Instead what's usually quoted are raw counts, things 
like "number of accounts that have made >5 edits in a month". But of 
course this kind of "blind quantitative" analysis is not a legitimate 
social-science methodology, at least not if some extremely strong 
ceteris-paribus assumptions are first validated.

To just pick one hypothesized confound among many that have been 
discussed on and off, there may have been a decline in the joint 
population of "vandals + vandal-fighters". These are counted as editors 
by the ">5 edits" criterion, but between them produce no net editing, so 
a decline in their joint population is not a real editor decline, and an 
increase in their joint population is not a real editor increase.

Another hypothesized confound is that there has been a wholesale 
replacement of "recent changes patrollers" with bots. A loss of net-95 
editors because 100 people who solely did recent-changes patrol were 
replaced by 5 bots that do the same job would be a decline of 95 
raw-data editors, but not really a net loss in productive editors.

These confounds might, in the end, not account for much after all. But I 
have been looking and haven't found even an attempt to *really* 
substantiate claims that the number of actual encyclopedia editors has 
declined, versus just superficial quantitative analysis of the 
accounts-making-edits raw data.


More information about the Wikimedia-l mailing list