Communities and Collections
Usage
  • 239 views
  • 210 downloads

Finding Bias by Characterizing Wikipedia Editing Behaviours

  • Author(s) / Creator(s)
  • Wikipedia has integrated into our information seeking needs and is a successful application of collaborative information production; however, with this combination, Wikipedia articles become susceptible to one-sided control with potentially negative impacts on our information gathering. Without a subject authority overseeing the development of an article, one or more of these volunteering editors can easily control the information through persistence in activities like rule interpretation, administrator privileges, and source selection. While anyone can refute an edit, it is likely the most relentless that will win the argument and exert control over an article. This presentation discusses an exploration of editing behaviours that might point us toward evidence of intentional bias. This presentation focuses on five of the most active contributors to Traditional Chinese Medicine (TCM) articles and gives them a behaviour profile. The analysis combines a qualitative macro- and micro-perspective of a particular editor’s history. A macro-perspective is achieved through the use of Xtools, which collects and visualizes both a user’s edit history and an article’s page history. The micro-perspective is achieved through close examination of a contributor’s edit history on a particular page. Attention is paid to acts of reverting, policy citing, large deletions and additions, and activity around sources, as these are assumed to be likely indicators of bias. Indicators of bias found in this study include localized attention on particular topics, localized attention on sections within articles while ignoring the rest of the article, frequent use of similar sources, frequent reverting, and frequent reference to policies. Odd behaviour worth investigating further for bias includes bursts of editing activity, experienced editing in early contributions, and high edit count with low final authorship of the article, which can be assessed with the data collected by Xtools.

  • Date created
    2020-02-07
  • Subjects / Keywords
  • Type of Item
    Conference/Workshop Presentation
  • DOI
    https://doi.org/10.7939/r3-mqge-qw85
  • License
    Attribution-NonCommercial-ShareAlike 4.0 International