Algorithmic Selves: How AI Recommendation Systems Shape Identity, Belonging and Well-Being among University Students

Abstract

Addressing the concept of algorithmic selves, this study examines how engagement with AI-curated Instagram and TikTok environments relates to identity-relevant self-evaluation, belonging, and psychological well-being among university-age users. It investigates whether platform engagement and the importance assigned to likes and followers are associated with problematic social-media use, low self-esteem, loneliness, and depressive symptoms. A secondary cross-sectional analysis was conducted using data from 252 respondents. Reliability assessment, hierarchical regression with HC3 robust standard errors, parallel and sequential mediation, exploratory moderation, sensitivity analyses, and external longitudinal triangulation were undertaken. Ordinal confirmatory factor analysis was not estimated because a suitable SEM estimator was unavailable. Platform engagement predicted problematic use, B=0.923, p=.001, as did likes salience, B=1.475, p=.004. Problematic use was associated with low self-esteem, B=0.208, p=.004, and depressive symptoms, B=0.561, p<.001, but not loneliness. The indirect effect through low self-esteem was significant, IE=0.233, 95% CI [0.070, 0.410]. The loneliness pathway was nonsignificant. TikTok-preferring respondents reported higher engagement and depressive symptoms, although platform preference did not moderate the primary relationship. Problematic use and self-evaluation may explain how algorithmically curated engagement relates to well-being, but causal claims require direct longitudinal measures of recommendation exposure.