Modeling news recommender systems' conditional effects on selective exposure: evidence from two online experiments.

This article introduces the Recommender Influenced Selective Exposure (RISE) framework to model the conditional effects of news recommender systems (NRSs) on selective exposure—the tendency of users to prefer news aligning with their existing attitudes. Through two preregistered online experiments involving Norwegian internet users browsing a simulated news site, the study empirically demonstrates that NRSs can either amplify or reduce selective exposure depending on their design goals, specifically by promoting attitude-consistent or attitude-inconsistent content via relevance and salience nudges. The findings suggest that selective exposure occurs under baseline conditions shaped by human editors or random ordering, but system-driven NRSs can causally influence this behavior by altering the prominence and perceived relevance of news articles. This research highlights that the democratic implications of NRSs depend on design decisions, shifting responsibility from the technology itself to its implementation, and encourages further exploration of NRS design to balance exposure diversity in online news environments.