As recommender algorithms increasingly shape daily choices, they quietly pick up people’s emotional behaviours across devices and platforms, using them as content filters and decision aids. However, little research has examined the perspectives of users, whose data seed the emotional inferences by algorithms to hyper-personalise recommendations. Filling this gap, the current study draws on Self-Determination Theory to unpack how and why users grapple with emotion-aware algorithms. Through in-depth interviews with 18 daily users of Spotify, we uncovered authentic narratives about how algorithms recognise and respond to user emotions in music streaming, where feelings, identities, and personal data are closely intertwined. The findings reveal core tensions between users’ needs for autonomy and relatedness, and how competence-supportive designs can buffer such tensions in human-algorithm interaction. This work extends research on emotion recognition in human–computer interaction from embodied agents to embedded recommender systems, foregrounding user voices and dynamic roles of emotions within the ecosystem of algorithmic personalisation. Further, we offer practical insights to foster user trust and well-being by calibrating user expectations for transparency with responsible interface designs.
Shuer Zhuo (Tue,) studied this question.