Towards a Theoretical Approach for Analysing Music Recommender Systems as Sociotechnical Cultural Intermediaries

As the rate and scale of Web-related digital data accumulation continue to outstrip all expectations so too we come to depend increasingly on a variety of technical tools to interrogate these data and to render them as an intelligible source of information. In response, on the one hand, a great deal of attention has been paid to the design of efficient and reliable mechanisms for big data analytics whilst, on the other hand, concerns are expressed about the rise of 'algorithmic society' whereby important decisions are made by intermediary computational agents of which the majority of the population has little knowledge, understanding or control. This paper aims to bridge these two debates working through the case of music recommender systems. Whilst not conventionally regarded as 'big data,' the enormous volume, variety and velocity of digital music available on the Web has seen the growth of recommender systems, which are increasingly embedded in our everyday music consumption through their attempts to help us identify the music we might want to consume. Combining Bourdieu's concept of cultural intermediaries with Actor-Network Theory's insistence on the relational ontology of human and non-human actors, we draw on empirical evidence from the computational and social science literature on recommender systems to argue that music recommender systems should be approached as a new form of sociotechnical cultural intermediary. In doing so, we aim to define a broader agenda for better understanding the underexplored social role of the computational tools designed to manage big data.