Link Me Baby One More Time: Social Music Discovery on Spotify (2401.08818v2)

Abstract: We explore the social and contextual factors that influence the outcome of person-to-person music recommendations and discovery. Specifically, we use data from Spotify to investigate how a link sent from one user to another results in the receiver engaging with the music of the shared artist. We consider several factors that may influence this process, such as the strength of the sender-receiver relationship, the user's role in the Spotify social network, their music social cohesion, and how similar the new artist is to the receiver's taste. We find that the receiver of a link is more likely to engage with a new artist when (1) they have similar music taste to the sender and the shared track is a good fit for their taste, (2) they have a stronger and more intimate tie with the sender, and (3) the shared artist is popular amongst the receiver's connections. Finally, we use these findings to build a Random Forest classifier to predict whether a shared music track will result in the receiver's engagement with the shared artist. This model elucidates which type of social and contextual features are most predictive, although peak performance is achieved when a diverse set of features are included. These findings provide new insights into the multifaceted mechanisms underpinning the interplay between music discovery and social processes.

• The paper demonstrates that music sharing is most effective when recipients and senders share similar tastes, measured by cosine similarity and enhanced by sender engagement.
• The paper finds that strong social ties through reciprocal interactions and direct messaging significantly increase the likelihood of recipient engagement.
• The paper shows that social cohesion, where a recipient’s network already favors an artist, plays a key role in promoting lesser-known music discovery.

An Analysis of Social Music Discovery on Spotify

The paper presented in this paper offers a comprehensive examination of social music discovery on Spotify through person-to-person music content sharing. Utilizing Spotify's large dataset, the authors investigate how individual user interactions impact music exploration and engagement, focusing primarily on explicit link sharing events. The research is grounded in the idea that music sharing among users can be influenced by social ties, user engagement, and artist popularity. The authors implement a Random Forest classification model to understand the intricate relationships influencing these social recommendations.

Key Findings

The paper's principal findings are rooted in three hypotheses regarding music engagement:

1. Music Taste Similarity: The research establishes that users are more inclined to engage with a new artist when the receiver and sender have similar music tastes, and the shared content aligns with the receiver's listening history. The authors affirm that taste similarity—quantified as the cosine similarity between user and track vectors—plays a crucial role in music engagement. Sender-artist engagement also enhances the probability of successful sharing, suggesting that senders' enthusiasm or familiarity with the artist may influence the receiver’s response.
2. Social Tie Strength: The likelihood of receiver engagement increases with the strength of social ties, as measured by reciprocity, direct communication methods, and quantity of past interactions. Direct communication through one-to-one messaging platforms positively affects engagement, surpassing the efficacy of broadcast-style sharing. Reciprocal relationships and pre-existing interactions further heighten the probability of engagement, reflecting the impact of trust and intimacy in social exchanges.
3. Social Cohesion and Network Effects: The paper presents evidence that social cohesion among the receiver’s connections influences engagement. Receivers are more likely to engage with new music if their existing social circle is already engaged with the artist. This effect is more pronounced for lesser-known artists, emphasizing the role of social influence in the discovery of new music.

Methodology and Model Evaluation

The researchers utilized a rich dataset from Spotify, capturing user interactions, sharing events, and music preferences, which informed the training of a Random Forest classifier to predict engagement. Feature sets included music taste similarity, sender-artist engagement, tie strength, and user platform activity metrics, with the model achieving a ROC-AUC of 0.73.

The feature isolation test, which evaluated individual feature sets, highlighted that while taste similarity and sender-artist engagement were essential in predicting engagement, a diverse feature pool provided the highest predictive performance. Notably, features involving social cohesion and network positions contributed to high precision, indicating their informative value when present.

Implications and Future Directions

While this paper offers critical insights into the dynamics of music discovery on platforms like Spotify, it also opens avenues for further exploration. Understanding the persistence of engagement beyond initial discovery could reveal how transient interest transitions into long-term fandom. Additionally, the underlying causal mechanisms—whether social influence or homophily—are still complex and warrant further investigation.

Future work could also expand to examine broader group interactions within social networks, moving beyond pairwise user data. This could inform strategies for enhancing social features on streaming platforms, potentially facilitating more effective music discovery and sharing experiences.

In summary, this research underscores the multifaceted nature of music discovery in digital environments, driven by interpersonal connections and network structures. Such insights could significantly inform the design of social features in music streaming services, enhancing user experiences and promoting diverse music exposure.