Beyond Wikipedia: Coordination and conflict in online production groups
In this paper, Kittur & Kraut study "online production groups" by trying to understand whether "some common processes are involved in the growth of all wiki communities". The generalizations on wikis are generated by comparing 6000+ Wikia wikis to (English) Wikipedia. Wikipedia has specific qualities that make it hard to generalize:
- it is an established community
- has established policies and norms
- many active volunteers
- social network/reputation
To provide a basis for generalizing, much of the study focuses on Wikia, which uses similar structures and norms as Wikipedia (including MediaWiki software).
Background theory and questions
Goals, norms, policies, and contributors may affect how work is organized. Coordination, in particular, changes as groups grow in size. They are particularly interested in conflict (under what conditions does it arise? How effective are "coordination mechanisms in managing conflict at different scales"?)
Wikipedia is more than the accretion of numerous edits (see Power of the few vs. wisdom of the crowd: Wikipedia and the rise of the bourgeoisie). Coordination is more complex in wikis because the group is more diffuse, without a control-command structure. Thus (citing Discontinuities and continuities: A new way to understand virtual work (ref 40)), collaboration requires "developing a common view of their task, their work processes and the work product." Three main coordination mechanisms (from the "classic" understanding of group process and organizational behavior) are expected:
- direct peer-to-peer contact and communication
- group structure (role differentiation, formal and informal management).
- shared mental models - either from spending time in a common environment or the use of standards, guidelines, and policies
These form the core of the paper, where coordination, group structure, and use of standards, guidelines and policies are later studied.
They used the full history from 6811 public wikis hosted on Wikia - 5 million pages created and maintained by 1 million editors. These show a skewed distribution of contributions and "very few communities survive to become highly successful with continued activity". (This notion of success focuses on the community, rather than the wiki knowledge base.)
The amount of communication increases over time, to become 10-15% of edits in Wikipedia and Wikia wikis. They hypothesize that "user talk scales with the number of users while article talk remains relatively constant"; they suggest looking for a common law governing coordination.
In Wikipedia article talk edits are 6-8% of all edits "for most of the lifespan of Wikipedia". User talk edits grew substantially from 2003-2005 (increased by a factor of 78); as of 2008, user talk edits were 6-7% of all edits.
Standards, Guidelines and Policies
Wikipedia is different from the other wikis studied in this respect: it uses policies and procedures much more heavily. "While policy and procedure edits grew up to 10% of all edits made in Wikipedia, their influence remains relatively small in other wikis, hovering around 3%." They provide some evidence that the Wikipedia namespace is a reasonable choice for review of standards, guidelines, and policies; however, as they note, deletion, WikiProjects, and other material is also contained in this space.
They use the Gini coefficient to show the differences between heavily edited and non-highly edited wikis; these are evident even "at inception". A core group of contributors seem to make most of the edits -- which they say points to the relevance of a workgroup structure.
They look to revert patterns to understand conflict, based on several past studies (se ref 3, 22, 35 below). They find the number of contributors to to be the most important factor, and find that the contributor density (number of contributors per page) is also important. This may relate to "the degree to which those editors are forced to interact".
Influence of Coordination on Conflict
They try to predict the number of reverts based on certain figures from the previous month: number of editors and coordination methods.
The proportion of article talk is associated with conflict though "It is difficult to distinguish whether article talk is a cause of conflict or a marker of conflict, since much conflict happens during the discussion of an article, but given that we are predicting conflict for a given month using the article talk from the previous month, this suggests that the more communication there is, the more likely it is for conflict to occur. ".
Having work concentrated in a core group is also associated with increase conflict (but also with improved article quality): "This suggests that there may be a risk/reward situation with having a small group of editors doing most of the work: if they work well together, avoid territoriality, and help structure the work of less involved contributors they can be extraordinarily effective in improving quality; but if they do not, having a few highly committed editors who are highly committed only to their own personal viewpoints can be a source of high conflict. An intriguing possibility is that in some situations both may be true, if conflict between editors is constructive and helps to clarify arguments and improve the page."
Policy and procedural work may help manage conflict, but changes in policy and procedure "are associated with greater conflict in subsequent time periods".
Article talk, user talk, and concentration of workgroup structure seem to increase in effectiveness for managing conflict as the wiki grows, counter to previous research. The authors hypothesize a "social benefit" to communication, perhaps by promoting shared mental models.
Further work is also needed to explain differences in these results, compared to related research.
- ref 3: Brandes, U. and Lerner, J. Visual analysis of controversy in user-generated encyclopedias. Information Visualization 7 (2008), 34–48.
- ref 22: Kittur, A., Suh, B., Chi, E., & Pendleton, B. A. He says, she says: Conflict and coordination in Wikipedia. In Proceedings of CHI (2007).
- ref 35: Suh, B., Chi, E., Pendleton, B. A., & Kittur, A. Us vs. them: Understanding Social dynamics in Wikipedia with revert graph visualizations. In Proceedings of VAST (2007).
- ref 40: Watson-Manheim, M., K. Chudoba, and K. Crowston, Discontinuities and continuities: A new way to understand virtual work. Information Technology & People, 15 (2002), 191- 209.
Theoretical and practical relevance:
- Importance of workgroups and core groups.
- Early results about counterintuitive results about conflict
The authors suggest further research on "how task and relationship factors interact". That might help predict "more optimal coordination patterns"; they suggest running experiments and interventions combining macro and micro models -- large scale network dynamics and micro-level coordination dynamics.
This was published in an open access journal.