UC San Diego

Music has not been exempt from the so-called "curatorial turn" that is visible in so many parts of our culture -- the turn, that is, toward machine learning to help consumers sort through the glut of media with which we are all confronted today. Automated music recommendations, in fact, are probably the single biggest driver of the music industry's recovery from the crisis it faced at the beginning of the 20th century. Nobody knows what shape the music industry will take if and when this recovery is complete, but it seems certain that automated curation will be at its center. This fact represents a confrontation of the human faculty of aesthetic judgment and machine learning at a scale the world has never seen. The issue of automated music recommendation raises many of the philosophical problems familiar from the critique of technology in culture. It does so, moreover, in a way that offers new insights into this familiar problem. In spite of this, little critical attention has been paid to music as a subject in the field of critical algorithms studies. In this dissertation, I provide an introduction to thinking critically about this crucial topic in algorithmic culture, taking Spotify as a case study that exemplifies many broader trends. I situate Spotify in the history of American copyright law, I perform a close reading of the Spotify platform, and I conduct quantitative experiments to analyze the large-scale behaviors of Spotify's recommendation engine. Although Spotify has often been seen as a singularly innovative company that somehow managed to "save" the music industry as a whole, this study shows that in many ways Spotify is better understood as an expression of attitudes toward musical meaning and commerce that are quite traditional in the music business.