PRISM: A Personalized News Recommendation System for Mitigating Political Echo Chambers
Abstract
Personalized news recommendation systems often reinforce users' pre-existing beliefs by limiting their exposure, contributing to political echo chambers. Existing mitigation approaches—such as rewriting biased text or diversifying recommendations—risk distorting article semantics or deepening polarization, while typically offering no rationale for why content is judged biased. We present PRISM, an interactive web-based news recommendation system that mitigates political echo chambers across three levels of granularity: feed level, article level, and event level. Building on KHAN (political stance prediction) and CROWN (personalized news recommendation) from our prior work, PRISM lets users monitor the political distribution of their reading history and current recommendations, inspect each article's predicted stance with bias-indicative keywords and natural-language rationales, and voluntarily explore cross-stance coverage of the same news event, preserving user autonomy in line with self-determination theory. PRISM is available at https://github.com/Hoyeol-Yang/PRISM, with a demonstration video at https://youtu.be/NbcvHME__cI.