Are Recommendation Systems Annoying? An Empirical Study of Assessing the Impacts of AI Characteristics on Technology Well-Being

Abstract

Recommendation systems—that is, a class of machine learning algorithm tools that filter vendors' offerings based on customer data and automatically recommend or generate personalized predictions—are empowered by artificial intelligence (AI) technology and embedded with AI characteristics; but the potential consequences for customer well-being are greatly overlooked. Hence, this research investigates the impact of AI characteristics on technology well-being (self-efficacy, technology satisfaction, emotional dissonance, and autonomy) through two mechanisms: intuitiveness versus intrusiveness. A literature review which conceptualizes AI characteristics and technology well-being in the recommendation system context is followed by a US-based survey approach which shows that higher levels of information optimization, predictability, human likeness, and customizability lead to higher levels of intuitiveness, whereas only information optimization and human likeness leads to increased intrusiveness. However, both intuitiveness and intrusiveness are found to promote technology well-being in the context of a recommendation system, especially for those more vulnerable individuals who respond positively to intrusiveness. Hence, the conclusion is “the recommendations are not always annoying,” whereby the relationships between AI characteristics and technology well-being are significantly influenced by perceived intrusiveness. These findings help business practitioners to identify how consumers perceive and engage different AI characteristics, and therefore could better take care of technology well-being while boosting AI development.