Algorithms and the Diffusion of Stereotyped Information
Abstract
This research examines the role of TikTok's recommendation algorithm in the dissemination of stereotypes, based on an analysis of 840 videos drawn from three contrasting thematic test accounts (entrepreneurship, beauty, politics). As algorithmic systems increasingly mediate access to information for younger audiences, understanding how they shape exposure to stereotyped content has become a pressing concern for information science and the broader academic study of online social networks. This digital ethnographic study unfolds across three successive phases conducted over four consecutive weeks : passive observation, differentiated active interaction, and deliberate diversification, designed to test the algorithm's responsiveness to user behaviour. The results indicate that 48.9% of suggested content conveys stereotypical representations, with significant proportions (40 to 51.4%) observed from the moment the accounts were activated. The analysis establishes a hierarchy of engagement signals: viewing time predominates over comments, which in turn exert greater influence than "likes." The study also confirms the existence of filter bubbles, whose intensity varies according to thematic domains, thereby questioning the relevance of algorithmic literacy as a remedy for information enclosure. These findings call for reinforced critical education and structural regulatory responses targeting platform design itself, rather than placing the burden of vigilance on individual users alone. Keywords: recommendation algorithms, TikTok, stereotypes, filter bubbles, algorithmic literacy, online social networks
