Breaking Echo Chambers with Personalized News

Abstract

When a news aggregator platform such as Google News or Apple News selects personalized news for its user, will it select news that conforms to its user’s bias, thus creating an “echo chamber”? To answer this question, this paper studies a game between a click-maximizing platform and a rational Bayesian user who reads news for its instrumental value. We derived the conditions for the user to exhibit a preference for headlines that contradict her prior bias, which induces the platform to recommend prior-contradicting headlines. Common statistical distributions, including the Normal distribution, satisfy these conditions. Our theory helps explain why, empirically, some media pander to their consumers’ biases while others do the opposite, and why some consumers seek prior-conforming news while others seek news to challenge their views.

Keywords: media bias, personalized news, echo chambers, news aggregator, platform

JEL Classification: C72, D83, L82