Is Collective Intelligence Less Biased?

A working paper tests the level of bias on Wikipedia.
Authors from Harvard Business School in Boston, Massachusetts, recently examined 3,918 pairs of articles about issues in U.S. politics. Each pair included one article from the Encyclopedia Britannica and one from the crowdsourced resource Wikipedia. The authors wanted to answer this question: Does crowdsourced information live up to its reputation of neutrality?

In a working paper, Shane Greenstein and Feng Zhu find that Wikipedia pays much less for the storage, production, and distribution of its information than Britannica. That means it also can present longer articles that incorporate more viewpoints and are revised far more often.

However, Greenstein and Zhu also find that Wikipedia articles generally are more biased than Britannica articles and lean toward more politically liberal views. The exception: Articles that are the most revised on Wikipedia, by the most contributors, are actually less biased than those in Britannica, showing that while Wikipedia may not live up to its ideal of neutrality, its methods eliminate bias when the crowdsourced contribution and revision process works at optimum levels.

These findings are relevant to any “closed communities,” the authors note. “Many private firms use wikis to organize internal efforts at knowledge management,” they write. “Our results imply that this strength is also a potential weakness in the absence of close managerial oversight.” To reduce bias on knowledge-based projects, managers should ensure that all sides of the issue are represented and that the project is revised many times by a wide range of contributors.

“Do Experts or Collective Intelligence Write with More Bias? Evidence from Encyclopedia Britannica and Wikipedia” is available online at www.hbs.edu/faculty/Publication%20Files/15-023_145191c4-220f-4dd8-8d41-ee2d1f693716.pdf.