History Flow Visualizations

Although it’s about three years old at this point, a group of researchers at MIT and IBM have an interesting paper (pdf, Google HTML version) showing interesting visualizations of Wikipedia edits. The visualization they use is derived from the same data as an annotate command in CVS or Subversion would give: it displays which user contributed which line. They color each user’s contribution’s a different color, and shows how these lines appear in the order they are in the article over time.

One interesting part is the analysis of vandalism, which is certainly a central issue to any wiki, especially one that aims to be a general encyclopedia. In some of the visualizations, one can see “holes” in the graph; these are places where a vandal deleted all or most of the article. However, changing the horizontal axis from equally-spacing edits to an actual time scale makes these disappear. They did not survive long enough to even show up when the graph was viewed in this way.

Wikis are an inherently democratic system; things that survive are those that everybody looking at the page can agree on. In this way, we can see visually how the actions of editors affect other editors. One example of this, in a sort of information cascade, is that the text put down by the person who created the article tends to survive much better than text added later. They explain this by saying that the first person sets the tone of the article, so everyone editing the article after them does things in the same way. There is also a very interesting picture of a zig-zag pattern demonstrating an edit war.

On a related note, here is a time-lapse video of the first 12 hours of the article on the Virginia Tech shootings.

Posted in Topics: General, Technology

Responses are currently closed, but you can trackback from your own site.

Comments are closed.



* You can follow any responses to this entry through the RSS 2.0 feed.