Workshop on Online Social Networks
Participation by invitation only

Microsoft Research Cambridge, UK.

December 7, 2007.

Reputation, Coordination, and Article Quality in Wikipedia

Gueorgi Kossinets (Cornell University)

Abstract:

Wikipedia is a free online encyclopaedia written and edited by hundreds of thousands of volunteers. Because all edits and discussions are automatically recorded, Wikipedia is a great data set with which to study coordination and self-organization in the context of distributed collaboration systems. For example, how can reliable knowledge production be sustained in an open community that relies on individual efforts in the absence of a formal incentive structure? And to what extent does the quality of output derive from the attributes and social network positions of the contributors? I will describe the findings from an analysis of the complete editing history of the English language Wikipedia from 2001 through 2006, which contains about 2 million articles edited by ca. 650,000 registered users.

Specifically, I will focus on the relationship between the social network of editors and the quality of entries they produce. In Wikipedia, high quality articles are selected through peer review, productive contributors receive token awards from other members of the community, and furthermore, members communicate with each other by collaborating on articles and discussion pages. One could expect that these mechanisms would lead to what what I call "reputation clustering" in the network of wikipedians, whereby contributors proximal to well-regarded editors will tend to produce better work themselves. I will discuss both theoretical and practical implications of this phenomenon and describe a statistical model to predict article quality by incorporating contributors' reputation and network positions