Comparing User Interfaces for Customizing Multi-Objective Recommender Systems

The goal of Multi-Objective Recommender Systems (MORSs) is to adapt to the needs and preferences of the users from different beyond-accuracy perspectives. When a MORS operates at the local level, it tailors its results to the needs of each individual user. Recent studies have highlighted that, however, the self-declared propensity of the users towards the different objectives does not always match with the characteristics of the accepted recommendations. Therefore, in this study, we delve into different ways for users to express their preference toward multi-objective goals and observe whether they have some impact on declared propensities and overall user satisfaction. In particular, we explore four different user interface (UI) designs and perform a user study focused on the interactions with both the UI and the recommendations. Results show that multiple UIs lead to similar results w.r.t. usage statistics, but users’ perceptions of these UIs often differ. These results highlight the importance of examining MORSs from multiple perspectives to accommodate the users’ actual needs when producing recommendations. Study data and detailed results are available from https://osf.io/pbd54/.