Diversity signaling to algorithmic versus human recommenders
(A2021-94169)

AUTHORS

Phyliss Jia Gai, Guanghua School of Management, Peking University; Anne-Kathrin Klesse, Erasmus University, Rotterdam School of Management; Eugina Leung, Tulane University

ABSTRACT

Consumers frequently receive product recommendations (e.g., which movie to watch or which song to listen) generated by algorithmic recommendation systems or human experts. To receive accurate recommendations (i.e., recommendations of products that match their preferences), consumers have to signal their preferences and provide input that the recommender can use to predict what kind of products the consumer would like. Notably, when indicating their preferences, consumers can choose whether to signal narrow (i.e., select similar items) or diverse tastes (i.e., select varied items). In five studies across various product domains (music, novels, news, videos, and paintings), we document that consumers are less likely to provide diverse input when the recommender is an algorithm rather than human. This occurs because consumers hold the lay belief that algorithms (versus humans) are less able to understand and process diverse input from consumers.