Systemic Gendered Citation Imbalance in Computer Science: Evidence from Conferences and Journals

Abstract

Gender imbalance persists across science, technology, engineering, and mathematics (STEM) fields, including computer science, where it appears in researcher demographics, productivity, recognition, hiring, and career progression. Given computer science's rapid expansion and global influence, addressing this imbalance is essential for broadening participation and fueling innovation. Although journal-oriented disciplines exhibit consistent gender imbalances in citation practices, it remains unclear whether similar patterns arise in the conference-centric culture of computer science. Here, we systematically investigate gender imbalance in citations of conference and journal papers in computer science. We find that papers for which a woman is listed as either first or last author receive fewer citations than expected, partly because of homophilic citation tendencies (i.e., authors tend to cite papers that share specific attributes). This imbalance is especially pronounced for conference papers--particularly those published at top-tier venues--relative to journals. Moreover, we find that the prominence of the first or last author and the structure of their local co-authorship networks are potential drivers of these imbalances. By exploring how conference-centric publishing practices can amplify systemic imbalances in computer science, our study offers insights that may inform efforts to foster more equitable representation in academia.

Figures (10)

• Figure 1: Time-varying demographics of published papers by gender category in computer science. (a) Proportions of papers by gender category over time. (b) Proportions of papers in gender categories MW, WM, or WW across 11 subfields of computer science over time.
• Figure 2: Gender imbalance in citations made by papers in computer science. Panels (a), (d), and (g) show results for all papers; (b), (e), and (h) for MM papers; and (c), (f), and (i) for WW papers. We used the random-draws model in (a)--(c), the homophilic-draws model in (d)--(f), and the preferential-draws model in (g)--(i). See Supplementary Tables S6--S8 for the statistical significance of the over/under-citation under these three reference models.
• Figure 3: Gendered citation imbalance across subfields and topics in computer science. (a) Gender imbalance in citations made by all papers to MM and WW papers in each subfield of computer science. See Supplementary Table S9 for the statistical significance of the over/under-citation. (b) Proportion of WW papers in a subfield versus the over/under-citation made by all papers to these papers. (c) Proportion of WW papers in a topic versus the over/under-citation made by all papers to these papers.
• Figure 4: Temporal trends of gender imbalance in citations made by papers in computer science. (a): All papers. (b): MM papers. (c): WW papers. See Supplementary Figure S3 for the statistical significance of the over/under-citation over time
• Figure 5: Gendered citation imbalance in different publication venues of computer science. (a) Over/under-citation received by MM and WW papers published in conferences and journals. See Supplementary Table S10 for the statistical significance of the over/under-citation. (b) Over/under-citation received by MM and WW papers in each venue rank. See Supplementary Table S11 for the statistical significance of the over/under-citation. (c) Proportion of WW papers in a conference versus over/under-citation received by these papers. (d) Proportion of WW papers in a journal versus over/under-citation received by these papers.