Supporting Personalized Music Exploration through a Genre Exploration Recommender

Recommender systems have been largely focused on the task of predicting users' current preferences and finding the most relevant items that users currently like. However, this approach is not sufficient as users may want to explore and develop new preferences, for example about a new genre. Allowing users to explore new preferences has many advantages, such as helping users to stay away from the so-called ``filter bubbles'', supporting new preference exploration and development, and promoting under-explored niche tastes, in addition to the mainstream preferences. Therefore, in this dissertation, we explore how recommender systems can be leveraged to support users' new preference exploration in the context of music genre exploration. The research takes a multidisciplinary approach in which we explore music recommendation algorithms and interactive exploration interface design for supporting music genre exploration, paired with insights from individual's music preference evolution and theories on decision making (such as digital nudges). For this purpose, we propose a music genre exploration tool and refine the tool over subsequent studies. We evaluate the music genre exploration tool with multiple single-session user-centric studies and one longitudinal user study on the long-term effectiveness of the tool to drive new preference exploration with various types of users’ objective behavior and their subjective user experience. From the studies, we find that users perceived the music genre exploration tool to be a new and helpful way to explore and develop new music tastes. By allowing users to make trade-offs between their current preferences and the new music genre they want to explore, the music genre exploration helps users make an easy personalized first step out of their comfort zone and towards the new preferences. The newly designed interactive exploration interface of the music exploration tool improves the usability and helpfulness of genre exploration by improving transparency, controllability and understandability. We further investigate individual differences during musical preference evolution by checking individuals' musical preference consistency and identify a relevant personal factor associated with this consistency (i.e., musical expertise). Our findings suggest that users with different musical expertise tend to show different musical exploration behavior. We further enhance the exploration tool with digital nudges to see if digital nudges can promote more exploration from users, and based on insights on individual differences, how this differs among individuals with different expertise levels. Based on our findings, we discuss opportunities and implications for future recommender systems to support new preference exploration and development.