Understanding the Effects of Recommender Systems on Consumer Views and Behaviors

Interacting with online personalization and recommendation systems can have unintended effects on consumer preferences and economic behavior.  In particular, consumers’ self-reported judgments can be significantly distorted by the system-predicted ratings that have been previously observed by consumers.  This talk provides an overview of our recent stream of studies on this phenomenon.  In particular, we find that a recommendation provided by an online system affects consumers’ ratings for products, even when the ratings are submitted immediately following consumption.  We also find that recommendations displayed to participants significantly sway their willingness to pay for items in the direction of the recommendation.  As recommender systems continue to become increasingly popular in today’s online environments, removing or mitigating such system-induced biases constitutes an important research problem.  We further discuss user-interface-based approaches to proactively preventing biases before they occur, i.e., at rating collection time.  Through laboratory experimentation, we identify and report relative benefits and limitations of graphic vs. numeric user interface designs of recommender systems.