Recommender system information trustworthiness: The role of perceived ability to learn, self-extension, and intelligence cues

Abstract

Making consumers trust in product descriptions provided by recommender systems is important for marketers, managers, and IT developers dealing with AI-based recommenders. The previous related research has focused on the relationship between the personalization of a recommender system’s offering and consumer trust in the system. This paper aims to extend this literature by investigating how the perceived recommender system ability to learn influences the perceived trustworthiness of the recommended product descriptions. Additionally, it is studied what role the consumer self-extension into system recommendations and intelligence cues play in this relationship. Two studies (Study 1 and Study 2) were conducted among young adults participating in an online research panel. Both studies used smartphones as a product category. All study participants were exposed to the same set of product descriptions ostensibly provided by recommender systems. Study 1 surveyed actual Facebook users (N = 204) and assumed Facebook as a recommender system. Perceived ability to learn, self-extension, and perceived trustworthiness were rated by the participants, and the measurements were analyzed with a mediation model. Perceived ability to learn had a positive effect on perceived description trustworthiness, and selfextension mediated this relationship. Study 2 (N = 515) was an experiment. The participants were exposed to a fictitious recommender system. Ability-to-learn cues (like asking about preferences; present vs. absent) and anthropomorphic cues (like first-person and conversation-like communication; present vs. absent) were manipulated between-subject. Perceived ability to learn, perceived anthropomorphism, and trustworthiness were rated by the participants, and the measurements were analyzed with ANOVA and moderated mediation models. Ability-to-learn cues × anthropomorphism cues interaction effect occurred on perceived ability to learn and trustworthiness. The positive effect of ability-to-learn cues on perceived ability to learn (perceived trustworthiness) was more positive (occurred only) in the presence of anthropomorphic cues. The paper extends the existing knowledge on consumer response to recommender systems by linking perceived system ability to learn and perceived recommended product description trustworthiness (with the evaluated offerings unchanged). Moreover, the paper provides novel insights into the role of self-extension and system intelligence cues in building consumer trust in recommender systems. The results may guide marketers, managers, IT developers, and policymakers dealing with AI-based systems through the possible consequences of developing AI technology in recommender systems.