Source-linked AI summary
Women Through the Glass Ceiling: Gender Asymmetries in Wikipedia
Claudia Wagner, Eduardo Graells-Garrido, David Garcia, Filippo Menczer
TL;DR
Wikipedia’s narrow, predominantly white and male editor community raises questions about gender inequalities in its content. This paper compares biographies of men and women across notability, topical and linguistic presentation, structural position, and metadata, finding systematic asymmetries across these dimensions.
Problem
The paper examines whether men and women depicted in Wikipedia are treated equally in notability, topical emphasis, language, metadata, and hyperlink structure.
Method
The authors compare biographies of men and women across global notability, topical and linguistic presentation, structural position, and metadata presentation.
Results
Women depicted in Wikipedia are slightly more notable than men, while biographies about women receive more family-, gender-, and relationship-related coverage and differ in language, metadata, and hyperlinks.
Takeaways & Limitations
The findings support ongoing assessment of gender bias in Wikipedia and suggest that some observed differences are attributable to Wikipedia editors rather than only offline social biases.
Takeaways & Limitations
The notability proxies are noisy and biased, and the analysis cannot determine whether the glass-ceiling effect originates with Wikipedians or reflects broader information biases.
Abstract
from arXiv · showhide
Contributing to the writing of history has never been as easy as it is today thanks to Wikipedia, a community-created encyclopedia that aims to document the world's knowledge from a neutral point of view. Though everyone can participate it is well known that the editor community has a narrow diversity, with a majority of white male editors. While this participatory \emph{gender gap} has been studied extensively in the literature, this work sets out to \emph{assess potential gender inequalities in Wikipedia articles} along different dimensions: notability, topical focus, linguistic bias, structural properties, and meta-data presentation. We find that (i) women in Wikipedia are more notable than men, which we interpret as the outcome of a subtle glass ceiling effect; (ii) family-, gender-, and relationship-related topics are more present in biographies about women; (iii) linguistic bias manifests in Wikipedia since abstract terms tend to be used to describe positive aspects in the biographies of men and negative aspects in the biographies of women; and (iv) there are structural differences in terms of meta-data and hyperlinks, which have consequences for information-seeking activities. While some differences are expected, due to historical and social contexts, other differences are attributable to Wikipedia editors. The implications of such differences are discussed having Wikipedia contribution policies in mind. We hope that the present work will contribute to increased awareness about, first, gender issues in the content of Wikipedia, and second, the different levels on which gender biases can manifest on the Web.
1 introduction
This work introduces a computational assessment of gender inequalities in Wikipedia across multiple dimensions and applies it to English Wikipedia. It finds asymmetries in notability, topical focus, language, and article structure.
- Findings: Women in Wikipedia are on average slightly more notable than men, with a larger gender gap among local heroes than superstars.The authors interpret the entry barrier as a subtle glass ceiling, especially where inclusion decisions are more subjective.
- Findings: Gender-, family-, and relationship-related topics are more dominant in stand-alone overviews of biographies about women.
- Findings: Abstract terms tend to describe positive aspects in biographies of men and negative aspects in biographies of women.The paper identifies this pattern as linguistic bias.
- Findings: Biographies of men and women differ structurally in their metadata and hyperlinks, with consequences for information-seeking activities.
- Contributions and approach: The paper presents a computational method for assessing gender bias in Wikipedia across multiple dimensions and applies it to English Wikipedia.The method is intended to support ongoing assessment, monitoring, and evaluation.
2 data & methods
The paper identifies its data sources for studying gender bias in Wikipedia.
- Data sources: The study draws on specified data sources to investigate gender bias in Wikipedia.
2. Inferred gender for Wikipedia biographies by [4].2
The study compares men’s and women’s Wikipedia biographies across notability, topical and linguistic presentation, article structure, and metadata. It uses linked Wikipedia and DBpedia data, gender metadata, lead-section text, PMI-based n-gram analysis, and hyperlink-network measures.
- DBpedia supplies article metadata, cross-language entity links, normalized hyperlinks, and person categorization across Wikipedia editions.
- Gender metadata is matched to English Wikipedia article URIs and obtained from Wikidata for other language editions.
- The analysis compares biographies about men and women across global notability, topical and linguistic bias, structural properties, and metadata presentation.
- The study focuses on biography lead sections because they are widely read and reflect editors’ selection of what matters most about a person.
- PMI ranks gender-associated words and bigrams, while linguistic analysis measures adjective ratios among positive and negative terms.
- The hyperlink analysis builds a biography network, computes PageRank, and compares women’s representation across centrality ranks.
3 results
This section presents the empirical results of the study on gender inequalities in Wikipedia.
- The section reports empirical results concerning gender inequalities in Wikipedia.
- The reported findings come from an empirical study of Wikipedia.
- The section is devoted to results about gender inequalities in Wikipedia.
3.1 Inequalities in Global Notability Thresholds
The notability analysis supports a subtle glass-ceiling interpretation: women included in Wikipedia are slightly more notable overall, while gender gaps are especially pronounced among less globally notable people. Search-based comparisons point in the same direction, but year-of-birth confounding limits interpretation.
- The local-hero gender gap is larger than the gap among global superstars, consistent with a higher Wikipedia entry barrier for women.
- 5.62 times: among people born after 1900 depicted in one language edition, men are more likely than women to appear as local heroes.
- For people born before 1900, the observed local-hero men-to-women ratio is 13.28 versus 11.73 expected by chance.Women are around 13% less likely to be depicted as local heroes than expected by chance in this population.
- 1.13 IRR: controlling for other parameters, being female increases a biography’s language-edition count, indicating slightly greater global notability.The effect is significant at p < 0.001.
- Women’s biographies show slightly greater Google search interest: the mean number of months above the global threshold is 32 for women and 30 for men.The median is one month for women and zero for men.
- Search-based notability comparisons cannot control for year of birth and profession because collecting large-scale Google Trends data was technically difficult.Subsamples matched on birth year and profession could address these confounding factors.
3.2 Topical and Linguistic Asymmetries
Wikipedia biographies differ in both topical emphasis and linguistic abstraction by gender. Women’s biographies more often foreground gender, family, and relationships, while abstract language is associated with positive descriptions of men and negative descriptions of women.
- Before 1900, women’s strongest associations include “her husband,” “women’s,” and “actress,” while men’s include “served,” “elected,” and “politician.”
- From 1900 onward, women’s strongest associations include “actress,” “women’s,” and “female,” while men’s include “played,” “league,” and “football.”
- Gender-, family-, and relationship-related n-grams are more prominent in women’s biographies, while men’s n-grams more often concern politics and sports.
- Abstract terms are used more for positive aspects of men’s biographies and negative aspects of women’s biographies.The reported effect sizes are very small.
- Women’s biographies tend to contain fewer abstract terms for positive aspects and more abstract terms for negative aspects.The pattern remains after including control variables in the regression models.
3.3 Structural Inequalities
Wikipedia biographies exhibit gendered differences in metadata, hyperlink structure, and visibility. These differences affect how biographies are represented and ranked in information-seeking systems.
- Meta-data: 33 of 340 DBpedia infobox attributes show statistically significant gender differences, although only 14 occur in at least 1% of biographies.The analysis qualitatively evaluates these differences because each attribute appears in only a small portion of biographies.
- Meta-data: Sports-related attributes are more frequent in biographies of men, consistent with men’s prominence in sports-related DBpedia classes.The attributes include active years, career station, matches, position, team, and years.
- Meta-data: Death-date and death-year attributes are more frequent for men born before 1900, while birthName is more frequent for women in recent times.The paper links these patterns respectively to historical documentation differences and surname changes after marriage in some cultures.
- Meta-data: Occupation, title, homepage, and spouse attributes are more frequent for women in recent biographies, reflecting art-related templates and relationship-focused metadata.Sport-related templates use different fields, while spouse metadata aligns with stronger relationship associations in women’s biographies.
- Network structure: All reported metadata differences have large effect sizes (Cohen’s w > 0.5).The empirical hyperlink network contains 700,706 non-singleton nodes and 4,153,978 edges after removing 192,674 singleton nodes.
- Network structure: Women are slightly less central among the highest-PageRank biographies, and their representation among top-ranked biographies remains below expected levels.For post-1900 biographies, women remain below the expected 16.8% across the entire range of top-k rankings.
- Network structure: Women’s biographies link to other women’s biographies more often than chance would predict in both observed birth-cohort networks.The comparison uses empirical networks and null models with gender-specific within- and across-group links.
4 discussion
The discussion interprets Wikipedia’s gender asymmetries as involving notability thresholds, portrayal, and visibility. It proposes policy and technical responses while acknowledging unresolved causal questions and trade-offs.
- Notability: The study estimates notability using Wikipedia editions and search engines, but leaves the causes of the observed effect for future research.The authors note that women may be less well documented and less visible on the Web.
- Notability: Women are slightly more notable than men after controlling for professions and birth year, while women’s share is smaller at low notability levels.The authors interpret this pattern as evidence of a subtle glass-ceiling effect in Wikipedia inclusion.
- Policy implications: Relaxing notability guidelines for locally notable women could increase their inclusion, but may also permit original research that Wikipedia rules prohibit.The authors propose a well-defined affirmative strategy as a possible way to grow women’s representation and ease discoverability.
- Policy implications: Editors’ wording choices are presented as a source of linguistic differences, despite Wikipedia’s reliance on secondary sources.The discussion suggests revising neutral-point-of-view guidance to address gender bias explicitly, including the Finkbeiner test.
- Visibility: Women’s visibility remains lower in link-based rankings even where editors interlink women’s articles effectively.Because men are the majority of biographies and tend to link more to men, the authors call for ranking algorithms that account for minority-group discrimination.
- Visibility: Suggested Wikipedia tools include gender-neutral-language guidance and prompts for missing reciprocal links, but these actions were not yet internal policies.The examples include linking back from a husband’s article to a woman’s article and incorporating existing language manuals.
5 related work
Prior research examined gender bias in Wikipedia’s coverage, topics, language, historical prominence, network structure, and editor community. This paper extends that work through a broader analysis of English Wikipedia content, notability, and linguistic bias.
- Coverage: Earlier studies found no gender-specific differences in coverage or article length among selected reference subjects, but Wikipedia’s missing articles were disproportionately female relative to Britannica.Other reference-list studies likewise found no significant proportional coverage difference between men and women.
- Topical bias: Prior work reported that biographies of women focus disproportionately on marriage and divorce, with similar topical biases across multiple language editions.These findings align with the present paper’s lexical and topical observations.
- Historical prominence: Across Wikipedia language editions, only 5.2 women on average appeared among the top 100 historical figures ranked by link-based algorithms.Without external reference lists, the expected number of women remains unclear in that earlier analysis.
- Network structure: Previous network studies used PageRank as an approximation of historical importance and examined link-based mechanisms contributing to the gender gap.PageRank measures node centrality based on network connectivity.
- Editor community: Research has also studied gender inequalities in Wikipedia’s editor community, including the Countering Systemic Bias WikiProject initiated in 2004.The present paper concerns gender inequalities in article content and structure rather than only editor participation.
- Present contribution: This paper advances prior work through in-depth English Wikipedia content and structure analysis, global notability signals, and linguistic-bias assessment.Its stated contribution goes beyond earlier topical and structural studies.
6 conclusions
The study finds gender differences in Wikipedia biographies that extend beyond offline social biases, while noting that its empirical analysis is limited to English Wikipedia. It presents a computational approach for assessing these differences and suggests policy revisions to address women’s visibility and language use.
- Findings: The results reveal gender differences at multiple levels that cannot be attributed solely to Wikipedia mirroring offline biases.The authors specifically attribute lexical and linguistic differences to editors’ word choices.
- Findings: Women depicted in Wikipedia tend to be more globally notable than men, indicating possible gender-specific entry barriers.The authors interpret this pattern as consistent with a subtle glass ceiling effect.
- Scope: The empirical analysis is limited to English Wikipedia, which is biased toward Western cultures.The authors leave a more detailed exploration of gender bias across all language editions for future work.
- Contributions: The paper contributes a computational method for assessing gender bias across multiple dimensions and applies it to English Wikipedia biographies.The method is intended for ongoing assessment, monitoring, and evaluation of these issues.
- Implications: The authors suggest revising Wikipedia guidelines to account for women’s low visibility and encourage less biased language use.This recommendation follows the paper’s findings about gender differences in selection and linguistic presentation.