AI-Driven Recommender Systems, Filter Bubbles, and Romanian Facebook Users: A Qualitative Analysis
TL;DR
This paper investigates how Romanian Facebook users experience AI‑driven recommender systems, focusing on perceived filter bubbles, echo chambers, and their emotional and relational consequences, and contributes to the understanding of how AI‑driven recommender systems impact Romanian users’ information environment, emotions, and social connections.
Abstract
Filter bubbles describe the phenomenon in which algorithm‑driven personalization of online content narrows users’ exposure, insulating them from broader or opposing viewpoints. Recommender systems, used widely across most digital platforms, are the foundational technology underlying the creation of filter bubbles, as they are designed to predict preferences and suggest relevant items, being especially prone to creating such insulated environments. This paper investigates how Romanian Facebook users experience AI‑driven recommender systems, focusing on perceived filter bubbles, echo chambers, and their emotional and relational consequences. Based on 12 in depth interviews with current and former users, the study explores three questions: how users describe personalization and bubble effects in their feeds, how they experience the emotional and psychological impact of AI‑curated content, and what changes they envision to improve their social media experience. Respondents report highly personalized, repetitive feeds that bring forward political, news, and group content while partially ghosting everyday posts from friends. They specifically link bubble formation to both algorithmic curation and their own practices of unfollowing, muting, and engaging with content. Respondents describe complex emotional states, ranging from amusement and connection to anxiety, anger, envy, guilt, and feelings of addiction. Former users who reduced or quit Facebook report improved well‑being, greater focus, and more deliberate information consumption from alternative sources. In all interviews, users express a strong desire for greater control and transparency, including adjustable parameters for their news feed, better filtering of unwanted content, and stronger moderation of negative speech and disinformation. The study contributes to the understanding of how AI‑driven recommender systems impact Romanian users’ information environment, emotions, and social connections, highlighting both active coping strategies but also the limits of individual solutions.