Abstract
Wikipedia is a community-created encyclopedia that contains information about notable people from different countries, epochs and disciplines and aims to document the world's knowledge from a neutral point of view. However, the narrow diversity of the Wikipedia editor community has the potential to introduce systemic biases such as gender biases into the content of Wikipedia. In this paper we aim to tackle a sub problem of this larger challenge by presenting and applying a computational method for assessing gender bias on Wikipedia along multiple dimensions. We find that while women on Wikipedia are covered and featured well in many Wikipedia language editions, the way women are portrayed starkly differs from the way men are portrayed. We hope our work contributes to increasing awareness about gender biases online, and in particular to raising attention to the different levels in which gender biases can manifest themselves on the web.
Introduction
As the world's most influential source of "neutral" information, Wikipedia’s content is shaped by its community of volunteer editors. However, this community is predominantly male, leading to long-standing concerns regarding a systemic "gender gap." While previous discussions have often focused on the raw number of biographies, this study argues that gender inequality is multi-dimensional. We examine four distinct types of bias: coverage, structure, language (lexical), and visibility across six major language editions: English, German, Spanish, French, Italian, and Russian.
Methodology and Datasets
To assess coverage bias, we required external "ground truth" lists of notable individuals to compare against Wikipedia’s actual content. We utilized three reference datasets to mitigate individual source bias:
- Freebase: A broad collection of approximately 120,000 notable people.
- Pantheon: A curated list of 11,000 biographies with a presence in at least 25 languages.
- Human Accomplishment (HA): A list of 4,000 eminent individuals in arts and sciences prior to 1950.
Our analysis employed network science (k-core decomposition and in-degree centrality) to measure structural importance, and Natural Language Processing (NLP) using a Naive Bayes classifier to identify gender-discriminative language.
Results: Coverage and Visibility
Contrary to common assumptions, our analysis of coverage yielded surprising results. When comparing Wikipedia to our reference datasets, we found that notable women are often proportionally better represented than notable men. In almost all language editions and across all three datasets, the percentage of notable women included in Wikipedia was higher than the percentage of notable men included.
This trend is further illustrated by the male-to-female coverage ratio. A ratio below 1.0 indicates that women from the reference dataset are more likely to have a Wikipedia article than their male counterparts. As shown in the figure below, this is the case for the majority of language-dataset pairs, particularly in the Human Accomplishment (HA) and Pantheon datasets.
Furthermore, we found no significant evidence of male bias in the selection of "Featured Articles" on the English Wikipedia homepage, suggesting that visibility—at least in terms of editorial promotion—is relatively balanced.
Results: Structural and Lexical Bias
While coverage numbers are encouraging, deeper structural and linguistic biases remain prevalent.
Structural Centrality: In the English, Russian, and German editions, men occupy significantly more central positions in the internal link network. Women are often relegated to the "periphery." While women’s articles frequently link to men (often husbands or male colleagues), the reverse is less common, creating a structural asymmetry where men appear more "essential" to the knowledge graph.
Lexical Disparity: The language used to describe women differs fundamentally from that used for men. Articles about women are significantly more likely to discuss romantic relationships, marriage, and family. Between 23% and 32% of the most discriminative words for women fall into categories like "Family" or "Relationship," whereas for men, these categories account for only 0% to 4%. Men’s articles are more likely to be defined by their professional fields (e.g., "infantry," "baseball"). This suggests a failure of the "Finkbeiner Test," as women are frequently defined by their domestic roles rather than their professional achievements.
Discussion and Limitations
Our findings suggest that the "gender gap" on Wikipedia is not merely a matter of inclusion, but of representation and connection. Even when women are included, they are often portrayed through a different lexical lens and positioned less centrally in the network.
A critical limitation is that our reference datasets may themselves reflect historical biases. If history has ignored notable women, Wikipedia’s "equal" coverage may simply be mirroring a biased past. Future work should investigate whether these biases are reinforced by Wikipedia’s software design or the specific behaviors of its editor community, and how search algorithms might inadvertently amplify the peripheral status of women's biographies.
Summary updated