Impact of AI-Based Content Filtering on Self-Esteem: Moderating Role of Digital Literacy

Abstract

The use of artificial intelligence (AI)-driven content filtering systems has become an intrinsic part of the design of modern-day digital platforms, which then influence the information landscapes that users are exposed to at a previously unheard-of level of accuracy and magnitude. Although the use of these systems has been widely implemented, their psychological implications, especially concerning the self-esteem of users, have not been properly studied in empirical research, especially in non-Western learning institutions. This paper was aimed at evaluating how AI-based content filtering affects the self-esteem of university students in Islamabad, Pakistan, and digital literacy as a modulating factor. The design used was a quantitative cross-sectional survey design and a total of 384 students were sampled and included in six universities (three government and three privately owned) in Islamabad through convenience sampling. The Rosenberg Self-Esteem Scale (RSES) was used to measure self-esteem, 10 modified items based on validated digital experience frameworks were used to measure AI-based content filtering perception and the Digital Literacy Scale (DLS) created by Hiller Spires and others was used to measure digital literacy. There was a good internal consistency among the instruments with Cronbach alpha values ranging between.79 and.93. The analyses of the data were performed with the help of the IBM SPSS Statistics Version 26 which helped to conduct preliminary analyses and with the help of SmartPLS Version 4.0 and AMOS Version 26 which helped to conduct the structural equation analysis and test the moderation. The findings showed a significant and negative correlation between AI-based content filtering perception and self-esteem ( beta = -.31, p =.001). Digital literacy moderated this relationship in a significant manner, whereby the students who reported to be more digitally literate showed less negative effects of AI filtering perception on self-esteem (interaction beta =.26, p <.01). Model fit indices were satisfactory (CFI = .958, RMSEA = .047, SRMR = .061). The implications of these findings are on the education of digital literacy, counseling in universities, and responsible design of AI filtering systems in educational digital ecosystems.