Recommender systems and market diversity
The last ten years have seen a large increase in the number of products available. Many believe this increased variety will allow consumers to obtain more ideal products for themselves. One difficulty that arises, however, is how consumers will find such niche products among so many choices. Recommender systems are one solution to this problem. These systems use data on purchases, ratings, and product content to identify which items are best suited to each user. Although a large body of work exists on designing recommender systems, we know much less about how they affect the market and society. This thesis begins a line of research in that direction, asking what effects recommenders have on the products sold through them and the consumers who use them. Part one asks how recommenders affect products: do recommenders increase the diversity of products sold? Two anecdotal views exist. A common view is that recommenders help consumers discover new products and thus increase sales diversity. Others believe that recommenders only reinforce the popularity of already popular products. Modeling the consumer-recommender interaction as a stochastic process, we find that some recommender designs can reduce sales diversity. In turn, consumers may be underserved if there exist better product matches outside of the hits. We also discuss design modifications that limit these popularity effects and promote exploration. Part two asks how recommenders affect consumers: do they create fragmentation among users? Recommenders give consumers a powerful means to focus on their interests and filter out all other content. As a result, critics argue that recommenders will reduce commonality and create fragmentation. Others, however, contend the opposite: recommenders may homogenize users because they share information among those who would otherwise not communicate. These are opposing views for which there is not yet empirical evidence. In an empirical study of a large service provider in the music industry, we find that recommenders are associated with an increase in commonality among users, and so concerns of fragmentation may be misplaced. The thesis thus identifies a debate about recommender systems in each part, products and consumers, and in each case, the reconciliation appears to challenge a popular view.
Fleder, Daniel M, "Recommender systems and market diversity" (2009). Dissertations available from ProQuest. AAI3363295.