The Daily Me versus The Daily Others: Can Social Recommender Systems Diversify User Interests?

Abstract

Recommender systems have been accused of polarizing user attention and consumption on online platforms. We examine this phenomenon using a field intervention on the largest online knowledge sharing platform (or Q&A community) in China. This platform originally adopted a content-based filtering algorithm that recommended content based on the topics to which each user subscribed. After operating for over a year, the algorithm was changed to social filtering which recommended content that each user’s social connections engaged with. We find that this intervention increased the creation of social ties by 21%, but this effect came at the cost of a 22% decrease in question subscriptions and a 19% decrease in answer contributions. We show that users’ increased social interests mainly involved following established and socially engaged users, leading to a greater concentration of social interests on the platform. However, we find that users’ own topical interests became less concentrated, as only popular (versus unpopular) topics received significantly fewer subscribers. We explain these findings by exploring their underlying mechanism. We show that compared with content-based filtering, social filtering algorithms are more likely to expose general users to content consumed by their followees who are more interested in niche topics than general users.