Under the Hood of Carousels: Investigating User Engagement and Navigation Effort in Multi-list Recommender Systems

DOI

A ranked list has been the standard way for several generations of recommender systems to present recommendations to their users. Even today, the majority of research focuses on achieving the best possible ranking and assesses recommender approaches by the quality of rankings they can generate. Yet, many user-facing recommender systems, especially in streaming and e-commerce domains, have switched to more interactive and flexible carousel-based interfaces. Is it just an attempt to make a better-looking interface, or is the popularity of carousel-based interfaces well supported by the better service they offer to users? This paper compares carousel-based recommendation interfaces with traditional ranked lists through three key experiments. First, a profile recovery simulation using the MovieLens dataset assesses how effectively each interface adapts to evolving user preferences. The results show that the carousel interfaces support faster adaptation and encourage broader content exploration compared to ranked lists. Second, a simulation examining navigation effort demonstrates that carousel interfaces reduce the effort that users exert to locate relevant content. Finally, we reexamine these findings through a large-scale user study in a real-world setting, demonstrating that carousel interfaces can enhance user experience by supporting efficient content discovery and promoting broader engagement with diverse content.