Two essays on consumers' relative preference for algorithm-based versus human recommendations

Abstract

Recommendation systems, the information filtering tools that provide options of products or services to customers, are powered by programmed algorithms or human experts. Both algorithm-based and human recommendation systems are widely adopted by companies to help consumers make optimal purchasing decisions amid their access to product information. However, the current state of the art lacks a systematic understanding of when companies should provide product recommendations by human experts or by algorithms, which impairs customer enjoyment as well as consumer loyalty. To address this question, it is necessary to understand the factors that influence consumers’ relative preferences for algorithm-based and human recommendations. In this thesis, I present two essays and examine two factors that influence consumers’ relative preferences for algorithm-based and human recommendations. The first essay investigates consumer-centric characteristics and shows that consumers who focus on the breadth (vs. depth) of product knowledge prefer recommendations by an algorithm over a human expert. I further explore several other marketing-related manifestations of the breadth (vs. depth) of product-knowledge focus. Eight studies demonstrate that consumers focus more on the breadth than the depth of product knowledge, and thus prefer algorithm over human recommendations, when they are less knowledgeable about the product category, expect to receive a large number of recommendations, or are in a relatively early decision-making stage. The second essay explores a ubiquitous situational factor, time of day, that can influence consumers’ relative preference for algorithm-based versus human recommendations. Across five studies, I show that consumers exhibit a preference for algorithm-based recommendations over those offered by human experts in the evening, as opposed to the morning. This enhanced preference is driven by the belief that human cognitive functioning is worse in the evening than in the morning. I further reveal that the effect is attenuated if individuals are led to believe that human cognitive functioning is not necessarily better during the daytime, or if cognitive efforts are not helpful in improving the recommendations. By discussing a consumer characteristic and a situational factor that influence consumers’ relative preference for algorithm-based versus human recommendations, this thesis deepens our understanding of algorithm-human comparison, enriches the literature on product knowledge, as well as extends the research on time-of-day effects on consumer behavior. In addition, guidelines are developed for marketers on how to utilize different recommendation agents in various situations.