Algorithmically curated music listening and academic stress recovery in university students: examining algorithmic...

Academic stress is among the most commonly reported psychological difficulties among university students in China, shaped by intense competition, high family expectations, and a rapidly changing graduate employment landscape. Music listening is one of the most widely used informal coping strategies in this population, and Streaming platforms increasingly shape everyday music listening through algorithmically personalized recommendation systems. However, the psychological mechanisms linking algorithmically curated music listening to stress outcomes remain poorly understood. However, the psychological mechanisms linking algorithmically curated music listening to stress outcomes remain poorly understood.

A sequential explanatory mixed-methods design was employed. A two-wave survey (T1 and T2, 4 weeks apart) was administered to 312 undergraduate students from three Chinese universities. Participants were classified into two listening-type groups based on their self-reported habitual music engagement: predominantly AI-curated (n = 178), predominantly self-selected (n = 108), or mixed-mode (n = 26; excluded from group-comparison analyses). Validated measures of perceived stress (PSS-10), emotional state (PANAS), algorithmic trust, emotional autonomy, and privacy concerns were collected. Multiple regression analyses examined between-group differences and individual-level predictors. A purposive subsample of 24 participants took part in follow-up semi-structured interviews analysed using reflexive thematic analysis.

Students who habitually listened to algorithmically curated music reported moderately lower perceived stress and negative affect at T2 compared to self-selecting listeners, though associations were small in magnitude. Algorithmic trust was the strongest predictor of lower stress (β = -0.31, p = 0.001). Emotional autonomy showed an unexpected positive association with stress (β = 0.19, p = 0.027). Privacy concerns were marginally associated with elevated negative affect. Qualitative analysis yielded four themes: low-effort relief, trust as accumulated experience, the double edge of autonomy, and privacy as background noise.

Associations between algorithmically curated music listening and stress recovery depend substantially on the quality of the user-algorithm relationship. Algorithmic trust is a key correlate of favourable outcomes. The design of digital wellbeing tools in higher education should prioritise transparency and user control.