User practices on music streaming platforms – RECORDS

How streaming platforms influence music listening habits? What can the massive amounts of listening data collected by these platforms reveal about contemporary music listening practices? How do musical consumption and listening habits evolve in an era of unlimited offer and personalized smart recommendation? Does the availability of gigantic musical catalogs give rise to very different individual music consumption patterns? To what extent the music people listen to on these platforms depend on the activity they perform while listening? Can the digital traces collected by those platforms support some empirical and theoretical social research that would aim to question and extend the existing social theories about cultural practices?
So far the empirical answers to these questions have been very limited. Questionnaire surveys fail to grasp the fine grain of listening practices, and to dig beyond general preferences that are declared by the respondents. Research based on interviews generally focus on the specific experience of populations that are very engaged in music listening, and those works fail to document the social diversity of listening habits and individual relations to music. On the other hand, more recent studies that have relied exclusively upon digital listening traces lack of explanatory power because these data do not include any information about the individuals that are 'behind' these data.
The ambition of the RECORDS project is to combine traditional social survey methods with big data analysis. It is based on an original partnership between social scientists, computer scientists and the research department of Deezer, one of the major music streaming platforms in France and in Europe. Through a large scale survey disseminated to hundreds of thousands of suscribing users, we will first collect self-declared data about the tastes, practices and socioeconomic characteristics of thousands of users of the platform. For the volunteer participants that will explicitely give their consent to participate to our research protocol, we will articulate this rich social information with their complete listening history data on the platform over several years. This corpus will be completed with interviews during which respondents will be able to visualize and comment their own listening history data. This will result in an unprecedented empirical material that will feed research on the social stratification of music listening, and more specifically on the diversity of content consumed on streaming platforms, at the individual and collective levels. We will measure the influence of different contexts (activities, locations, moments) and the influence of the recommendation engine on the music listened to, at different time scales. The database will also allow us to question the theory of cultural omnivorism in the light of people's actual listening data. We will measure to what extent people really listen to diverse content, and compare this effective diversity to the diversity of taste they declare in surveys and interviews. We will also compare the social explanatory power of the usual musical genre categories -- upon which are built most of the empirical results in the field of cultural sociology -- to the explanatory power of bottom-up categories, directly built from the effective plays of listeners characterized by their social properties. Finally we will focus on the role of geography in shaping listening dynamics at the national and international scales. We expect that the results will support an informed, realistic and complete understanding of contemporary listening practices on these platforms.