The Algorithmic Shade Room

As algorithms increasingly shape the way individuals consume digital content, many scholars have taken it upon themselves to investigate how these systems may reinforce social inequalities. One specific and often overlooked manifestation of algorithmic bias is colorism, the preferential treatment of individuals with lighter skin tones. While previous research has examined algorithmic discrimination in areas such as healthcare, policing, and education, less is known about colorist bias embedded in social media feeds and how it can affect users' perceptions. This thesis investigates the extent to which colorist algorithmic biases influence perceptions of platform trust and fairness, and whether comfort with racial diversity in friend groups moderates this relationship. A between-subject survey experiment was conducted with 216 participants. Each participant was randomly assigned to view a mock Instagram feed curated with either predominantly lighter-skinned individuals or a racially diverse set of individuals. Participants completed a pre- and post-test measuring their perceived trust and fairness toward the platform, as well as a scale measuring the racial diversity of their social circles. The findings revealed that exposure to racially diverse Instagram feeds significantly increased perceptions of platforms trust and fairness. In contrast, exposure to feeds with lighter-skinned individuals did not significantly decrease perceptions of trust and fairness. The moderation analysis revealed comfort with racial diversity in friend groups did not significantly influence the relationship between algorithmic bias and user perceptions. These results demonstrate that visible racial diversity in content matters for how platforms are perceived, regardless of users' cross-group friendships. The findings of this thesis contribute to ongoing discussions in algorithmic fairness by highlighting the psychological impact of visual diversity in algorithmically curated content. The findings also introduce colorism as a critical variable in the broader discourse of digital equity and representation, specifically in personalized recommendation systems. Even though the study was limited by factors such as ecological validity and the sensitivity of topics such as race and discrimination, the implications highlight the importance of inclusive algorithmic design and enhanced transparency in AI systems. By holding platforms accountable for their unequal values embedded in their algorithms, this research advocates for more ethical, inclusive, and socially aware technology designs.