I have heard talk in the past about various biases that the Wikipedia editors’ community seems to have. For example, that community has a reputation for being gender-biased. But is that reputation accurate—is it actually biased, or is that a subjective perception? And how would we be able to tell, empirically and objectively?

The results “uncover significant gender differences at various levels that cannot only be attributed to the fact that Wikipedia is mirroring the off-line world and its biases.”

That’s the set of questions that researchers at the University of Koblenz-Landau, GESIS, Telefonica, ETH-Zurich, and Indiana University set out to answer. In an article just published in EPJ Data Science titled, “Women through the glass ceiling: gender asymmetries in Wikipedia,” Claudia Wagner, Eduard Graells-Garrido, David Garcia, and Filippo Menczer describe how they developed an approach for measuring gender bias on Wikipedia. And their results “uncover significant gender differences at various levels that cannot only be attributed to the fact that Wikipedia is mirroring the off-line world and its biases.”

To get to this conclusion, Wagner et al. compared the Wikipedia articles (the English-language version) on men and women against external sources, and they also compared the articles’ linguistic and meta-data structures. For example, as an external measure of “notability,” they used Google search volume between 2004 and 2015, and compared the interest expressed by Google searches for notable men and women, and compared that against Wikipedia content, and found that women overall had to be “more notable” to have a Wikipedia article than men.

In addition, they also analyzed topical and linguistic biases in the articles that are posted on English Wikipedia.

They conclude by translating their findings “into some potential actions for the Wikipedia editor community to reduce gender biases in the future.” You can read the entire article here.