Can Algorithm Knowledge Stop Women from Being Targeted by Algorithm Bias? The New Digital Divide on Weibo

• Published in: Journal of broadcasting & electronic media Vol. 67; no. 3; pp. 397 - 422
• Main Authors: Zhang, Yang, Chen, Huashan
• Format: Journal Article
• Language: English
• Published: Philadelphia Routledge 27.05.2023; Routledge, Taylor & Francis Group
• Subjects: Algorithm Bias; Algorithm Knowledge; Algorithms; Bias; Digital Divide; Gender aspects; Gender differences; Gender divide; Internet access; Personalization; Questions; Recommender systems; Women
• ISSN: 0883-8151; 1550-6878
• DOI: 10.1080/08838151.2023.2218955

Algorithm knowledge of users plays a crucial role in avoiding them from algorithm bias in recommendation systems. Gender of users has been found to correlate with algorithm bias, but also leaving behind a question of whether this relationship can be described by algorithm knowledge. By using Weibo as an example system, we clarify the aforementioned question from a digital divide theory perspective. We combine a traditional method (questionnaire) with a deep learning computational method to explain algorithm bias in two sequential studies. Our findings suggest that algorithm knowledge solely works for men while fails to protect women. Who users follow helps determine what information they are exposed to on Weibo, and this renders female users' algorithm knowledge useless. This work provides a valuable perspective on algorithm bias: we view algorithm bias as a new digital divide and contribute to the understanding of gender differences by applying the digital divide perspective. Methodologically, we contribute by integrating traditional and computational methods to explain algorithm bias from a folk theory perspective.