Rapid technological advances in artificial intelligence (AI), data analytics, big data, the semantic web, the Internet of Things (IoT), and cloud and mobile computing have given rise to a new generation of AI-driven recommendation agents (RAs). These agents continue to evolve and offer potential for use in a variety of application domains. However, extant information systems (IS) research has predominantly focused on user perceptions and evaluations of traditional non-intelligent product-brokering recommendation agents (PRAs), supported by empirical studies on custom-built experimental RAs that heavily rely on explicit user preference elicitations. To address the lack of research in the new generation of intelligent RAs (RAs 2.0), this dissertation aims to study consumer responses to AI-driven RAs using an affordance perspective. Notably, this research is the first in the IS discourse to link RA design artifacts, RA affordances, RA outcomes, and user continuance. It examines how actualized RA affordances influence user engagements with and evaluations of these highly personalized systems, which increasingly focus on user experiences and long-term relationships. This three-essay dissertation, consisting of one theory-building paper and two empirical studies, conceptually defines "RAs 2.0," proposes a comprehensive theoretical framework with testable propositions, and conducts two empirical studies guided by smaller carved-out models to test the validity of the comprehensive framework. The research is expected to enrich the IS literature on RAs and identify potential areas for future research. Moreover, it offers key implications for industry professionals regarding the effective system development of the new generation of intelligent RAs. / Doctor of Philosophy / Rapid technological advances in artificial intelligence (AI), data analytics, big data, the semantic web, the Internet of Things (IoT), and cloud and mobile computing have given rise to a new generation of AI-driven recommendation agents (RAs). These agents continue to evolve and offer potential for use in a variety of application domains. This three-essay dissertation, consisting of one theory-building paper and two empirical studies, conceptually defines "RAs 2.0," proposes a comprehensive theoretical framework with testable propositions, and conducts two empirical studies guided by smaller carved-out models to test the validity of the comprehensive framework. The research is expected to enrich the IS literature on RAs and identify potential areas for future research. Moreover, it offers key implications for industry professionals regarding the effective system development of the new generation of intelligent RAs.

The quantity and variety of product information available online today has increased significantly in recent years. This situation has exacerbated user information overload perceptions and made it difficult for online shoppers to choose between various online products and services. This is especially true for older adults, who typically have limitations in cognitive abilities due to the natural aging process and, as such, may perceive additional difficulties processing large amounts of information online. In response, Recommendation Agents (RAs) have become popular as decision support tools for online consumers in general, and older adults in particular. However, in the information systems literature, there is a lack of understanding regarding the design of RAs to suit the needs of different segments of the population, including older adults. Grounded in the theory of planned behaviour, and the “aging and IS adoption” literatures, this study investigates the impact of cognitive age and RA comprehensiveness on user perceptions towards the complexity of the input and output stages of an RA, and their subsequent impact on the antecedents of a user’s intention to utilize the RA for online shopping. This experimental study finds that: (i) an individual’s cognitive age significantly increases perceived RA input and output complexity perceptions; (ii) higher levels of RA comprehensiveness increases a user’s RA input and output complexity perceptions significantly; (iii) RA output complexity plays a more critical role than RA input complexity in shaping user perceptions of the overall complexity of an RA; and, (iv) increased levels of RA comprehensiveness increases individual perceptions of RA usefulness. Additionally, and as expected, cognitive age moderates the relationship between RA comprehensiveness and input/output complexity such that the effect is stronger for older adults. Surprisingly, however, cognitive age also moderates the relationship between RA comprehensiveness and perceived RA usefulness such that it is stronger for older adults. Theoretically, this study helps us to better understand how different levels of RA comprehensiveness, in terms of both the input and output stages of the RA operation, impact the intention of users of different cognitive ages to use online RAs. For practitioners, the results highlight the importance of customizing the design of RAs, in both their input and output stages, for consumers with different cognitive ages. / Thesis / Doctor of Philosophy (PhD)