Recommender Systems and their Effects on Consumers: The Fragmentation Debate

Recommender systems are becoming integral to how consumers discovermedia. The value that recommenders offer is personalization: inenvironments with many product choices, recommenders personalize thebrowsing and consumption experience to each user's taste. Popularapplications include product recommendations at e-commerce sites andonline newspapers' automated selection of articles to display based onthe current reader's interests. This ability to focus more closely onone's taste and filter all else out has spawned criticism thatrecommenders will fragment consumers. Critics say recommenders causeconsumers to have less in common with one another and that the mediashould do more to increase exposure to a variety of content. Others,however, contend that recommenders do the opposite: they may homogenizeusers because they share information among those who would otherwise notcommunicate. These are opposing views, discussed in the literature forover ten years for which there is not yet empirical evidence. We presentan empirical study of recommender systems in the music industry. Incontrast to concerns that users are becoming more fragmented, we findthat in our setting users become more similar to one another in theirpurchases. This increase in similarity occurs for two reasons, which weterm volume and taste effects. The volume effect is that consumerssimply purchase more after recommendations, increasing the chance ofhaving more purchases in common. The taste effect is that, conditionalon volume, consumers buy a more similar mix of products afterrecommendations. When we view consumers as a similarity network beforeversus after recommendations, we find that the network becomes denserand smaller, or characterized by shorter inter-user distances. Thesefindings suggest that for this setting, recommender systems areassociated with an increase in commonality among users and that concernsof fragmentation may be misplaced.