Saikore has teamed up with Optimice to do an preliminary study of applying social network analysis tools to a Wiki. Read on to find a report with our early findings.
Using social networking tools we can make use of wiki transaction logs to visualize the relationship between wiki pages and editors. Who is editing what; who are the most active editors, both in page edits and breadth of pages edited; which pages are being edited the most and by whom etc.
For this study, the Wikipatterns wiki site has been chosen as a subject, mainly because of its subject, community size, technical accessibility and public nature. Wikipatterns is a wiki site about wiki usage. It is built with Atlassian's Confluence wiki and contains a toolbox of wiki usage patterns and anti-patterns. The site aims to be a guide to major stages of wiki adoption and it explores patterns that apply at each stage. Wikipatterns is a relatively new site, it started in January 2007, and can probably claim to have a very wiki literate membership. As of June 2007, there are about 400 users, of which 200 have edited a page and of these 200, about 40 users have made edits to 5 or more page.
Note that the founding member of the site, Wiki Evangelist Stewart Mader (blog), was removed from the analysis because he swamps the network with his connections to nearly every page and every other user. Sorry Stewart!
The map below provides a visualization of wiki pages, editors and their interactions through common page editing (click for a larger version).
The size of the nodes reflect the number of connections between editor and page. For example, the larger editor nodes reflect relative editing activity in terms of different pages edited. The larger page nodes reflect the number of different editors having written to the page. The thickness of the links reflects the relative number of edits an editor has conducted on a given page. In this analysis, the creation of a page was not considered as an 'edit', and so pages that have been created and then never modified will not appear in the diagram because they will appear to have no edits and hence no links. Isolated pages and editors have also been removed. Only page edits have been considered, comments - which are an equally valuable source of connections - have been ignored in this analysis.
Some interesting observations
The home page is one of the most edited pages, but the edits appear to be mostly done by infrequent editors, many for whom the home page is their only edit.
The broad based editors appear to be seeding many new pages.
There is a pattern of many lightly edited pages and many one page only editors compared to multi-edited pages or multi-page editors. This is in fact common in the world of social networks and is called the "small world" effect.
One other interesting thing we can do with the data is to look at which editors are affiliated through editing common pages. That is, we could infer a relationship between editors based on their co-editing of a page. The map below illustrates these affiliations:
The nodes are all editors. The link thickness represents the relative strength of affiliation through common page edits. The size of the nodes reflects the relative number of affiliations with other editors. The colours show an attribute gleaned from the e-mail addresses i.e. whether they were free e-mail accounts like hotmail or gmail; dot gov or edu; Atlassian (company who created the wiki space) or a corporate e-mail account.
What we can see is that the Atlassian editors are pretty much central to the network of editors and all are active. There appears to be two clusters emerging, one of which is around five of the Atlassian editors. Cmiler and fschop are playing "broker" roles between the two clusters. Trevor Pike is also brokering to relatively new editors on the periphery. Of course the above interpretation may be over-reading the relationship aspects as we don't really know to what degree joint editing creates relationships between the collaborative editors. For that we need to do a little more research.
For more information, please contact the authors, James Matheson (Saikore )and Laurence Lock Lee (Optimice)
