I want to share some new research in to Wikipedia article creation trends
on the 10 largest Wikipedias: English, German, French, Italian, Portuguese,
Spanish, Polish, Chinese, Japanese and Russian.
This work was led by Aaron Halfaker and done in part as background work in
to potential future design work by the Growth team, aimed at helping
newcomers be more successful at creating their first articles.
- Slides are at:
- and Aaron's talk was recorded as part of our first public research
showcase at the Wikimedia Foundation: http://youtu.be/arO9YzcTWGE
This is really important insight in to the nature of who creates articles
and how on large Wikipedias. Aaron compares the overall success rates of
editors based on their experience level, as well as the workflow used to
create a page (direct creation, userspace draft, or a more elaborate review
process like English Wikipedia's Articles for Creation system).
In particular, some important or unusual conclusions/questions we have
1. Retention of articles by newly-registered users is actually getting
worse over time. How can we use new software and better social policies to
turn this around? In addition to Aaron's thoughts in the presentation, we
have notes on mediawiki.org.
2. Except in Polish and English (where anonymous article creation is
turned off) anonymous editors are actually much more prolific and
successful article creators than users who create a page in their first 24
hours after registering an account. How can we support these anonymous
editors more? Our hypothesis about why they are more successful is
currently that they include some experienced editors, including some small
number of logged-out Wikipedians.
3. Why is survival of new articles so high on Japanese Wikipedia?
4. How can we seriously reform review processes like "Articles for
Creation", Flagged Revisions, and so on? These backlogs of review hamper
throughput of new articles created by newbies. In English Wikipedia's case,
it is seriously choking off new article creation. The quality of articles
that make it past is high, but is not enough to make up for the 50% (!)
drop in volume of new articles in my view.
More thoughts are welcome,
and its /Usability