Wikipedia was founded on the notion the Internet is a self-correcting machine: by harnessing collective intelligence through an open-source platform, the facts will ultimately come to light. But a new study shows that collective intelligence generally produces biased information, except in a narrow range of circumstances. Northwestern’s Shane Greenstein and the University of Southern California’s Feng Zhu analyzed a decade’s worth of Wikipedia articles on U.S. politics and found that only a handful of them were politically neutral.
Large numbers of contributors did, in fact, help make articles more unbiased, they note, confirming what’s known in the tech community as Linus’ Law: “Given enough eyeballs, all bugs are shallow.” However, Greenstein and Zhu found that “the majority of articles receive little attention, and most articles change only mildly from their initial slant.” Overall, Wikipedia is less biased and partisan than in its early years–when the site’s political entries leaned Democrat–but most of its content hasn’t benefitted from the true wisdom of crowds, they conclude.
It’s worth, noting, however, that the researchers used a very technical index to measure political slant, one developed by two economists to measure media bias. The index measures how often 1,000 phrases are used, drawn from utterances by Democrats and Republicans in the 2005 Congressional Record. According to the index, Democrats are more likely to use phrases like “civil rights” and “trade deficit” while Republicans are more likely use phrases like “economic growth” and “illegal immigration.” But the vocabulary of partisans has doubtlessly shifted somewhat since 2005.