Wow, thanks for making that.  I had thought that the combination of shifting vandalfighting to the edit filters and the change of intrawikis from peer to peer to hub and spoke had reduced the number of bot edits. But then I find there are bots following me around - when I semi protect an article a bot follows me and adds the little silver padlock icon on the page. Which gets me to the more serious point about measuring things by edit count. All edits are far from equal, someone who manually writes several paragraphs of encyclopaedic content is contributing something far more valuable than for example my recategorising 30 images  from one category to another. I'd go so far as to say that in that example the one edit that writes paragraphs of text involves more work and is of perhaps thirty times the "value" of my thirty edits. But measured on edit count we would value them the other way round.

I think that this is such a distortion that it risks skewing our whole understanding of the project, and I'm wondering if anyone has found another way to try and measure someone's contributions? In other spheres of human endeavour you might try and estimate how many hours of work had been contributed by particular people or groups of people. You could I suspect get part of the way there by only counting the number of unique calender hours in which someone had made edits. Some people routinely do hundreds of manual edits in an hour, and such people would on this measure have contributed far fewer hours than those whose edits are rarely less than ten minutes apart. Of course estimating the hours of effort by counting the number of unique hours in which one has edited would over estimate the effort of the hypothetical editor who occasionally contributes the odd couple of minutes, and under estimate the efforts of someone who takes more than an hour before they actually hit save. But it would probably be much closer to people's relative donation of time to the project than simply looking at raw edit count.

On a related note, I'm looking for ways to test the hypothesis that some of the drop in edit count between our 2007 peak and the introduction of the edit filters in 2009 was simply due to bots reverting vandals faster, and getting individual vandals through the warning levels and blocked quicker and for fewer total edits than would have happened in manual vandalfighting. Can anyone suggest ways to test that hypothesis or measure its effect?

Regards

Jonathan

 

On 14 October 2013 17:40, Diederik van Liere <dvanliere@wikimedia.org> wrote:

Very cool. If you include wikidata then more than 50% of the edits on the Wikimedia projects are made by bots. One of the dead horses I like to beat is that bot editors should be treated as first class citizens of Wikipedia and this data nicely illustrates that.  I think this is a bigger watershed moment (we might have reached this threshold a while back) then mobile vs non-mobile and we should have a way more rigorous discussion about the future of bots on Wikipedia. Particularly as all our big features are aimed at human editors :)

D

On Mon, Oct 14, 2013 at 12:30 PM, Dario Taraborelli <dtaraborelli@wikimedia.org> wrote:

A new app by Thomas Steiner (@tomayac) counting bot vs human edits in real time from the RecentChanges feed:

http://wikipedia-edits.herokuapp.com/

(read more [2]). The application comes with a public API exposing Wikipedia and Wikidata edits as Server-Sent Events. [1]

Dario

[1] http://blog.tomayac.com/index.php?date=2013-10-14&time=16:49:46&perma=Bots+vs.+Wikipedians.html
[2] https://en.wikipedia.org/wiki/Server-sent_events

_______________________________________________
Analytics mailing list
Analytics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/analytics

_______________________________________________
Wiki-research-l mailing list
Wiki-research-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wiki-research-l

_______________________________________________
Wiki-research-l mailing list
Wiki-research-l@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wiki-research-l