Automated gender identification vs. manual review: Gender assignment of editorial board members in Information and Library Science Journals

In recent years, computer programmes that automatically assign gender based on a person’s name and country of affiliation have been increasingly used in gender studies of authors of scientific publications. This study aims to compare the results generated by the automated genderize.io programme with those obtained through manual gender identification. To assess the accuracy of genderize.io, the gender of Editorial Board Members (EBMs) from 84 journals in the field of Information and Library Science was analysed. The comparison revealed discrepancies: genderize.io incorrectly classified 80 out of 1,419 men as women, and 124 out of 2,580 women were misidentified as men. Additionally, genderize.io classified the gender of 123 EBMs as unknown. While the manual method achieved a 99.15 percent accuracy rate, genderize.io had a slightly lower accuracy of 91.51 percent. There was, however, strong agreement between the two methodologies (Cohen's Kappa = 0.829, p < 0.001). Genderize.io exhibited a 7.71 percent inaccuracy rate, particularly underestimating the number of women. The study concludes that while automated software like genderize.io is effective for large-scale analyses and useful for library and information professionals, manual review is recommended for smaller studies to ensure higher accuracy.