Questions about ideological bias on Wikipedia are reflected in academic analysis and public criticism of Wikipedia, and especially its English-language site, in relation to whether or not its content is biased due to the political, religious, or other ideology of its volunteer Wikipedia editors and any effect it may have on the reliability of the online encyclopedia. Wikipedia has an internal policy which states that articles must be written from a neutral point of view, which means representing fairly, proportionately, and, as far as possible, without editorial bias, all of the significant points of view that have been verifiably published by reliable sources on a topic.[a]
Collectively, findings show that Wikipedia articles edited by large numbers of editors with opposing ideological views are at least as neutral as other similar sources, but articles with fewer edits by a smaller number of ideologically homogeneous contributors were more likely to reflect editorial bias.
Shane Greenstein and Feng Zhu, professor and associate professor respectively at the Harvard Business School, have authored several studies examining Wikipedia articles related to U.S. politics and the editors that work on them to identify aspects of ideological bias within its collective intelligence.
In Is Wikipedia Biased? (2012), the authors examined a sample of 28,382 articles related to U.S. politics (as of January 2011) measuring their degree of bias on a "slant index" based on a method developed by Matthew Gentzkow and Jesse Shapiro in 2010, to measure bias in newspaper media. This slant index purports to measure an ideological lean toward either Democratic or Republican based on key phrases within the text such as “war in Iraq”, “civil rights”, “trade deficit”, “economic growth”, “illegal immigration,” and “border security”. Each phrase is assigned a slant index based on how often it is used by Democrat vs Republican members of U.S. Congress and this lean rating is assigned to a Wikipedia contribution that includes the same key phrase. The authors concluded that older articles from the early years of Wikipedia leaned Democratic, whereas those created more recently held more balance. They suggest that articles did not change their bias significantly due to revision, but rather that over time newer articles containing opposite points of view were responsible for centering the average overall.:4-5
In a more extensive American follow-up study, Do Experts or Collective Intelligence Write with More Bias? Evidence from Encyclopedia Britannica and Wikipedia (2018), Greenstein and Zhu directly compare about 4,000 articles related to U.S. politics between Wikipedia (written by an online community) and the matching articles from Encyclopædia Britannica (written by experts) using similar methods as their 2010 study to measure "slant" (Democratic vs. Republican) and to quantify the degree of "bias". The authors found that "Wikipedia articles are more slanted towards Democratic views than are Britannica articles, as well as more biased", particularly those focusing on civil rights, corporations, and government. Entries about immigration trended toward Republican. They further found that "(t)he difference in bias between a pair of articles decreases with more revisions" and, when articles were substantially revised, the difference in bias compared to Britannica was statistically negligible. The implication, per the authors, is that "many contributions are needed to reduce considerable bias and slant to something close to neutral".
The study Ideological Segregation among Online Collaborators: Evidence from Wikipedians (2016) by Greenstein, Zhu, and Yuan Gu focused on the behaviors of contributing editors themselves. Working again within a subset of articles related to U.S. politics and using terminology introduced in Is Wikipedia Biased?, the authors offer several significant findings.
They found that editors are slightly more likely to contribute to articles which exhibit an opposite slant to their own—a tendency that the authors called opposites attract. They further found that debates on Wikipedia tend to exhibit a "prevalence of unsegregated conversations over time", meaning that the debates on Wikipedia tend to involve editors of differing view – which the authors called unsegregated – as opposed to debates involving only editors with homogenous views (segregated). Unsegregated conversation is supposed to favor the convergence towards a neutral point of view.
They also found that the degree of an editor bias decreases over time and experience, and decreases faster for editors involved in editing very slanted material: "[t]he largest declines are found among contributors who edit or add content to articles that have more biases". They also estimated that, on average, it takes about one year longer for Republican material to reach a neutral viewpoint than for Democratic material. These results must however be taken with care, as they were not peer-reviewed.
A subsequent peer-reviewed study found that a model of this productive friction, which is defined as the collective resolution of socio-cognitive conflicts, can explain and predict the dynamics of knowledge production on Wikipedia, further supporting the hypothesis that collaborative work from multiple editors with opposing views help reach neutrality. Furthermore, another study found on the French Wikipedia that a majority of editors had a propensity to share equally in a dictator game, and that this propensity was correlated with their involvement on Wikipedia (as measured by the time spent and attachment).
American Christian conservative activist Andrew Schlafly founded the online encyclopedia Conservapedia in 2006 based on his view of "liberal bias" on Wikipedia, which he also described as "increasingly anti-Christian and anti-American". He said that he "found that the biased editors who dominate it censor or change facts to suit their views" and that "facts against the theory of evolution are almost immediately censored", that some articles use British English, and that Christianity is not given credit for the Renaissance. Conservapedia has itself received negative reactions from figures, journalists, and scientists for its bias and factual inaccuracies.
In 2013, the Croatian-language version of Wikipedia drew media attention after the daily newspaper Jutarnji list reported on critics' concerns that administrators and editors on the website were projecting a right-wing bias into topics such as the Ustashe regime, anti-fascism, Serbs, the LGBT community, and gay marriage. Many of the critics were former editors of the website who said they had been exiled for expressing concern. The small size of the Croatian Wikipedia — as of September 2013, it had 466 active editors of which 27 were administrators — was cited as a major factor. Two days after the story broke, Croatian Minister Željko Jovanović advised students not to use the website. In 2018, historians with the University of Zagreb told the Balkan Investigative Reporting Network (BIRN) that the Croatian Wikipedia has "many shortcomings, factual mistakes and ideologically loaded language" and that students are often referred to the English Wikipedia instead of their native Croatian, especially for topics on Croatian history.