Understanding Echo Chambers in E-commerce Recommender Systems
The echo chamber phenomenon in Alibaba Taobao --- one of the largest e-commerce platforms in the world --- is analyzed and evidence suggests the tendency of echo chamber in user click behaviors, while it is relatively mitigated in user purchase behaviors.
Abstract
Personalized recommendation benefits users in accessing contents of interests\neffectively. Current research on recommender systems mostly focuses on matching\nusers with proper items based on user interests. However, significant efforts\nare missing to understand how the recommendations influence user preferences\nand behaviors, e.g., if and how recommendations result in \\textit{echo\nchambers}. Extensive efforts have been made in examining the phenomenon in\nonline media and social network systems. Meanwhile, there are growing concerns\nthat recommender systems might lead to the self-reinforcing of user's interests\ndue to narrowed exposure of items, which may be the potential cause of echo\nchamber. In this paper, we aim to analyze the echo chamber phenomenon in\nAlibaba Taobao -- one of the largest e-commerce platforms in the world. Echo\nchamber means the effect of user interests being reinforced through repeated\nexposure to similar contents. Based on the definition, we examine the presence\nof echo chamber in two steps. First, we explore whether user interests have\nbeen reinforced. Second, we check whether the reinforcement results from the\nexposure of similar contents. Our evaluations are enhanced with robust metrics,\nincluding cluster validity and statistical significance. Experiments are\nperformed on extensive collections of real-world data consisting of user\nclicks, purchases, and browse logs from Alibaba Taobao. Evidence suggests the\ntendency of echo chamber in user click behaviors, while it is relatively\nmitigated in user purchase behaviors. Insights from the results guide the\nrefinement of recommendation algorithms in real-world e-commerce systems.\n