Algorithmic stratification in social media narrative framing: an agent-based testing of sex and educational disparities in career information access

Abstract

Abstract This study investigates how social media algorithms create information accessibility gaps through an Agent-based Testing (ABT) approach. Using 640 controlled virtual accounts on REDnote (RED) varying by sex and educational levels, 96,000 career-related search results were analyzed. Findings reveal significant algorithmic stratification patterns: lower-educated and women users receive more positively emotional and gender-stereotyped content, and the inter-sectional influence between the assigned sex and educational level of those accounts collectively shapes the narrative perspective and framing of the content, creating complex stratified recommendation patterns. This study makes several contributions: it empirically demonstrates gender discrimination in social media algorithmic recommendation systems, quantifies the differential career information delivery across sex groups, and identifies specific mechanisms through which algorithmic systems perpetuate gender inequality in career development opportunities. These findings provide empirical evidence for mitigating algorithmic gender discrimination in the occupational domain.