Gender Bias in YouTube Exposure: Allocative and Structural Inequalities in Political Information Environments
Gender Bias in YouTube Exposure: Allocative and Structural Inequalities in Political Information Environments
Authors: Jipeng Tan, Weifeng Zhang, Ye Wu, Jialin Guo, Yŏng Min Date: 2026-04-30 Paper ID: openalex:2604.27479
Summary
This paper investigates gender-based algorithmic bias in YouTube’s political recommendation environment using a controlled social-bot field experiment. The authors identify both allocative biases—differences in content, ideology, and entities—and structural biases characterized by community-level clustering. By modeling these dynamics, the study illustrates how recommendation algorithms reinforce societal inequalities through personalized information exposure pathways.
Key Contributions
- Identifies statistically significant gender-based allocative bias in YouTube recommendations regarding issue distribution, ideology, and political entities via controlled bot experiments.
- Demonstrates structural bias on YouTube through distinct clustering patterns that differentially segregate male-coded and female-coded profiles.
- Provides a collaborative-filtering model that successfully reproduces observed gender-based inequalities in political information exposure.
Open Questions & Future Work
Archivist Review
The paper provides a valuable empirical analysis of algorithmic bias but does not propose new technical architectures, mechanisms, or formal methods that qualify as reusable ‘concepts’ for the vault. I have approved the open question regarding the political consequences of algorithmic segregation as it highlights a significant and unresolved research direction in algorithmic fairness and digital democracy.
Approved Open Questions
- Political consequences of algorithmic segregation: Understanding the downstream effects of algorithmic bias on democratic citizenship is critical to evaluating the normative implications of platform-mediated political communication.
Rejected Candidates
- [concept] Allocative bias (
allocative-bias) - not_novel: This is a standard term in algorithmic fairness and not a novel methodology specific to this paper. - [concept] Structural bias (
structural-bias) - not_novel: Similar to allocative bias, this is a general social-technical concept rather than a reusable research architecture or mechanism from the paper.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2604.27479
- paper_id
- 2604.27479
- paper_source
- openalex
- domain
- nlp
- tags
-
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-05-03T06:03:40Z