Understanding user preferences in recommender systems: A systematic review with a user-centric perspective in the audiovisual domain

Abstract

User preference is central to the design and effectiveness of recommender systems (RS), as its satisfaction influences both user experience and platform success. However, the concept of user preference is inherently complex, and existing knowledge remains diverse and fragmented, lacking a synthesis that considers users’ multiple and evolving needs. Adopting a user-centric perspective, this study systematically reviews 135 RS papers focused on how user preferences are defined and operationalized in the audiovisual (AV) domain. Grounded in theories of multi-attribute utility and bounded rationality, this review focuses on key dimensions related to the conceptualization and application of user preferences in RS studies. We develop a typology to categorize the main issues and methodological approaches and identify prevailing trends and research gaps across these dimensions. Our findings reveal that: (1) user preference definitions are frequently absent, with few studies offering explicit conceptualizations. (2) user participation in preference elicitation is limited, with most studies relying on offline datasets and prioritizing content relevance metrics over user experience; (3) User attributes, such as demographics and personality traits are underrepresented compared to item attributes like content features, leading to overspecified models and reduced interpretability; and (4) interest in temporal and contextual preference dynamics is growing, although stability remains crucial for consistent recommendations. These findings indicate critical shortcomings in understanding and operationalizing user preferences from a user-centric perspective. We emphasize the need for greater integration of insights from psychology and behavioral studies to ensure that RS design aligns with users’ intrinsic, evolving, and long-term needs.