Online Exploration When Search Topic and Popularity Ranking Are Decoupled: Insights on Echo Chambers

Abstract

Personalized search algorithms produce results that are both topically relevant and ranked by their general popularity and individual fit to users’ previous searches and choices. New choices from such tailored lists feed back into the algorithms, over time creating content echo chambers, where content exposure is increasingly biased toward users’ and their friends’ interests and views. We create an online search environment for TED Talks, where topic and popularity are separately controlled, and study the relationship between users’ characteristics and their reliance on own interests vs. crowd-based popularity sorting in content exploration. In topic-based searches, we randomly block/show popularity information to examine its impact on the tendency to explore. We find that high levels of sociability, previous experience with similar content, and a younger age are associated with an increased susceptibility to echo chambers, represented by little to no exploration and popularity sorting prior to content choice. Opinion leaders may alleviate echo chambers in their social circles as they conduct more topic-based exploration and exhibit lower popularity reliance. Showing popularity information increases opinion leaders’ popularity sorting, but does not impact non-leaders’ exploration. Our findings identify users’ echo chamber risk factors, and suggest that reducing the salience of popularity information may contribute to more balanced content exposure facilitated by opinion leaders.