A More Human Future For Music Discovery

Abstract

Mainstream music recommendation systems rely heavily on prediction algorithms that translate listeners’ unformed desires into explicit, nameable categories, leaving limited space for open and exploratory listening. This study addresses this limitation by exploring how to support Stimmung, the felt and preverbal attunement through which listeners sense and orient themselves toward a musical desire before it can be consciously articulated or named. To support this form of emergent musical desire, the study introduces nourishment as an interaction design condition that sustains the listener’s active interpretive work rather than replacing it with passive, automated delivery.
Using a Research through Design methodology, the project identified dimensions of experiential loss associated with the shift from physical record stores to frictionless streaming. These insights informed the iterative, autobiographical development of Curio, a domestic, object-based AI system. Curio invites users to place personally meaningful household objects on a surface. A multimodal AI then interprets the objects’ material qualities without identifying them by name, using this interpretation to retrieve unfamiliar music. This exploration resulted in an annotated portfolio of provotypes examining how agency is distributed among the user, the object, and the machine.
The research demonstrates that making AI reasoning fully transparent does not necessarily empower users. Instead, explicit justifications can function as persuasive devices that encourage passive acceptance of algorithmic recommendations. By contrast, withholding definitive explanations and introducing “productive ambiguity” can invite listeners to engage in meaningful co-interpretation. The study also shows that interaction design must correspond to the level of commitment associated with different media formats. Albums, as high-commitment formats, warrant slow and deliberate gestures, whereas songs, as lower-commitment formats, require more fluid and continuous interaction landscapes.
Ultimately, this study transforms AI-mediated music discovery from a transactional exchange into an embodied and reflective ritual. It contributes transferable design knowledge through a framework comprising the stages Locate, Name, Decide, Translate, and Provoke. This framework enables designers working in other automated domains to reintroduce nourishment into digital contexts characterised by abundance.