I've just come across a really important bit of research on EN wiki.
User:Mr.Z-man analyzed the new accounts created in February to see how many were still editing in September.
Key findings:
Over two thirds of new accounts had still not become editors after 6 months.
Editors who start by creating articles are only about a quarter of new editors, three quarters edit existing articles. I find this credible if a little on the low side. But over ten thousand newbies created an article in February 2010, had their article deleted and ceased editing.
Editors whose articles are not deleted are over seven times more likely to remain than editors whose articles were deleted.
As one would expect, the retention rate of Article creators was much lower than of other new editors.
http://en.wikipedia.org/wiki/User:Mr.Z-man/newusers
WereSpielChequers
Thanks for pointing this out. This is very cool stuff. Philippe & James Alexander are working on some related questions. I just asked them to reach out to you and Mr.Z-man.
Zack
On Mon, Mar 21, 2011 at 12:27 PM, WereSpielChequers < werespielchequers@gmail.com> wrote:
I've just come across a really important bit of research on EN wiki.
User:Mr.Z-man analyzed the new accounts created in February to see how many were still editing in September.
Key findings:
Over two thirds of new accounts had still not become editors after 6 months.
Editors who start by creating articles are only about a quarter of new editors, three quarters edit existing articles. I find this credible if a little on the low side. But over ten thousand newbies created an article in February 2010, had their article deleted and ceased editing.
Editors whose articles are not deleted are over seven times more likely to remain than editors whose articles were deleted.
As one would expect, the retention rate of Article creators was much lower than of other new editors.
http://en.wikipedia.org/wiki/User:Mr.Z-man/newusers
WereSpielChequers
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l
Yes, thanks for sending this along.
I created a fallout chart which helped me understand the relative weights of different paths: http://bit.ly/ggyypQ
I'm also trying to put this in the context of the Editor Trends Research [1] and the Product Whitepaper [2]. These numbers seem to suggest that registered users actually have a relatively small percentage chance of getting their first edit reverted (3.2% = 1,223/38,404). This seems to be pretty consistent with what we're seeing in other revert research (e.g., Zachte's revert trends [3] has registered editors being reverted 4.2% of the time).
While reversion rates for registered editors appears to have grown quite a bit from 2005-2007, the numbers are still relatively small, especially when compared to the reversion rates of anonymous editors. This seems to suggest that important areas for research are:
1) Understanding reversion of edits from anonymous users and their downstream impact on our other metrics (e.g., New Wikipedians) 2) Obtaining a more nuanced understanding of reversions to registered editors (e.g., in addition to understanding the % of edits that get reverted over time, understanding the nature of the reversions and whether they have gotten more contentious over time).
Howie
[1] http://strategy.wikimedia.org/wiki/Editor_Trends_Study/Results [2] http://strategy.wikimedia.org/wiki/Product_Whitepaper [3] http://strategy.wikimedia.org/wiki/Product_Whitepaper#Reversion_and_Newbie_T...
On 3/21/11 1:24 PM, Zack Exley wrote:
Thanks for pointing this out. This is very cool stuff. Philippe & James Alexander are working on some related questions. I just asked them to reach out to you and Mr.Z-man.
Zack
On Mon, Mar 21, 2011 at 12:27 PM, WereSpielChequers <werespielchequers@gmail.com mailto:werespielchequers@gmail.com> wrote:
I've just come across a really important bit of research on EN wiki. User:Mr.Z-man analyzed the new accounts created in February to see how many were still editing in September. Key findings: Over two thirds of new accounts had still not become editors after 6 months. Editors who start by creating articles are only about a quarter of new editors, three quarters edit existing articles. I find this credible if a little on the low side. But over ten thousand newbies created an article in February 2010, had their article deleted and ceased editing. Editors whose articles are not deleted are over seven times more likely to remain than editors whose articles were deleted. As one would expect, the retention rate of Article creators was much lower than of other new editors. http://en.wikipedia.org/wiki/User:Mr.Z-man/newusers WereSpielChequers _______________________________________________ RCom-l mailing list RCom-l@lists.wikimedia.org <mailto:RCom-l@lists.wikimedia.org> https://lists.wikimedia.org/mailman/listinfo/rcom-l
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l
I think that we need to consider three other factors here. 1 The proportion of reverts that are legit 2 The community's dissonance over unsourced edits. 3 Transparency
In my experience a substantial minority of IP edits are vandalism, but the community is fairly effective at reverting that vandalism. Much of it comes from school IPs with long block logs and a prominent place on the screens of hugglers.
There are inconsistencies and complexities about the reversion of unsourced edits. Complexities in that newbies maybe surprised to discover that an unsourced change to a biography of a living person will get reverted by many editors who would not be as cautious about an edit to an article on a battleship, volcano or Bollywood film. Inconsistencies in that some editors would respond to the addition of an unsourced fact with a citation needed tag, a revert or a subvocalised meh.
But the big difference between the deletion of new articles and the reversion of new edits to existing articles is transparency. People do make mistakes with rollback, huggle and the other vandal reversion tools. But the system is very effective at picking them up, not least because anyone can look at the edit history of live edits. Vandalfighters who make frequent mistakes get told to slowdown and ultimately lose rollback rights. We are far less effective at dealing with errors at speedy deletion.
WereSpielChequers
On 21 March 2011 21:36, Howie Fung hfung@wikimedia.org wrote:
Yes, thanks for sending this along.
I created a fallout chart which helped me understand the relative weights of different paths: http://bit.ly/ggyypQ
I'm also trying to put this in the context of the Editor Trends Research [1] and the Product Whitepaper [2]. These numbers seem to suggest that registered users actually have a relatively small percentage chance of getting their first edit reverted (3.2% = 1,223/38,404). This seems to be pretty consistent with what we're seeing in other revert research (e.g., Zachte's revert trends [3] has registered editors being reverted 4.2% of the time).
While reversion rates for registered editors appears to have grown quite a bit from 2005-2007, the numbers are still relatively small, especially when compared to the reversion rates of anonymous editors. This seems to suggest that important areas for research are:
- Understanding reversion of edits from anonymous users and their
downstream impact on our other metrics (e.g., New Wikipedians) 2) Obtaining a more nuanced understanding of reversions to registered editors (e.g., in addition to understanding the % of edits that get reverted over time, understanding the nature of the reversions and whether they have gotten more contentious over time).
Howie
[1] http://strategy.wikimedia.org/wiki/Editor_Trends_Study/Results [2] http://strategy.wikimedia.org/wiki/Product_Whitepaper [3] http://strategy.wikimedia.org/wiki/Product_Whitepaper#Reversion_and_Newbie_T...
On 3/21/11 1:24 PM, Zack Exley wrote:
Thanks for pointing this out. This is very cool stuff. Philippe & James Alexander are working on some related questions. I just asked them to reach out to you and Mr.Z-man. Zack
On Mon, Mar 21, 2011 at 12:27 PM, WereSpielChequers werespielchequers@gmail.com wrote:
I've just come across a really important bit of research on EN wiki.
User:Mr.Z-man analyzed the new accounts created in February to see how many were still editing in September.
Key findings:
Over two thirds of new accounts had still not become editors after 6 months.
Editors who start by creating articles are only about a quarter of new editors, three quarters edit existing articles. I find this credible if a little on the low side. But over ten thousand newbies created an article in February 2010, had their article deleted and ceased editing.
Editors whose articles are not deleted are over seven times more likely to remain than editors whose articles were deleted.
As one would expect, the retention rate of Article creators was much lower than of other new editors.
http://en.wikipedia.org/wiki/User:Mr.Z-man/newusers
WereSpielChequers
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l
Howie,
There are two things that I'm pretty sure the editor trends survey has missed.
Firstly it looks at reverted edits but not deleted ones. That's why Mr Z man's survey is so important - it looks at the article creators whose deleted articles don't count if you only look at live edits.
Secondly, and I'm not sure whether we want to make a big public thing about this, much of wikimedia was contributed by people who were at work. As companies got more slick about monitoring internet usage and disciplining staff for misuse of the Internet then I'm fairly sure we lost a lot of those editors, or only got their edits outside working hours. I started editing in early 2007, just after the peak period, and I knew not to edit at work because I'd been involved from a Privacy angle in the implementation of monitoring software. I don't know what the market penetration was during the middle of that decade for such software, but I'm pretty sure it is now ubiquitous on corporate setups.
We don't have either the IP address data to directly verify this nor do I think e could or should ask our 2004/2006 editors if they were editing on their employers time. Though we could indirectly test this by taking editors who have declared their age and timezone and look at how the profile of edits in the working day has changed over time.
I know we that in the UK mornings we usually have a build up of backlogs as the Australians go to bed, the Americans haven't woken up yet and most of the Brits are at work. I suspect that would have been different five years ago.
If my theory on that is correct then one element of our dropoff in editing may be a one off rather than an ongoing trend.
Though I think we ought to be discreet about this, we could have some editors still in the same jobs as they had then who wouldn't want their employers to realise that we still have records of the times they edited Wikipedia several years ago.
WereSpielChequers
On 21 March 2011 21:36, Howie Fung hfung@wikimedia.org wrote:
Yes, thanks for sending this along.
I created a fallout chart which helped me understand the relative weights of different paths: http://bit.ly/ggyypQ
I'm also trying to put this in the context of the Editor Trends Research [1] and the Product Whitepaper [2]. These numbers seem to suggest that registered users actually have a relatively small percentage chance of getting their first edit reverted (3.2% = 1,223/38,404). This seems to be pretty consistent with what we're seeing in other revert research (e.g., Zachte's revert trends [3] has registered editors being reverted 4.2% of the time).
While reversion rates for registered editors appears to have grown quite a bit from 2005-2007, the numbers are still relatively small, especially when compared to the reversion rates of anonymous editors. This seems to suggest that important areas for research are:
- Understanding reversion of edits from anonymous users and their
downstream impact on our other metrics (e.g., New Wikipedians) 2) Obtaining a more nuanced understanding of reversions to registered editors (e.g., in addition to understanding the % of edits that get reverted over time, understanding the nature of the reversions and whether they have gotten more contentious over time).
Howie
[1] http://strategy.wikimedia.org/wiki/Editor_Trends_Study/Results [2] http://strategy.wikimedia.org/wiki/Product_Whitepaper [3] http://strategy.wikimedia.org/wiki/Product_Whitepaper#Reversion_and_Newbie_T...
On 3/21/11 1:24 PM, Zack Exley wrote:
Thanks for pointing this out. This is very cool stuff. Philippe & James Alexander are working on some related questions. I just asked them to reach out to you and Mr.Z-man. Zack
On Mon, Mar 21, 2011 at 12:27 PM, WereSpielChequers werespielchequers@gmail.com wrote:
I've just come across a really important bit of research on EN wiki.
User:Mr.Z-man analyzed the new accounts created in February to see how many were still editing in September.
Key findings:
Over two thirds of new accounts had still not become editors after 6 months.
Editors who start by creating articles are only about a quarter of new editors, three quarters edit existing articles. I find this credible if a little on the low side. But over ten thousand newbies created an article in February 2010, had their article deleted and ceased editing.
Editors whose articles are not deleted are over seven times more likely to remain than editors whose articles were deleted.
As one would expect, the retention rate of Article creators was much lower than of other new editors.
http://en.wikipedia.org/wiki/User:Mr.Z-man/newusers
WereSpielChequers
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l
RCom-l mailing list RCom-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/rcom-l