You are here

Understanding Music Recommender Systems as Cultural Intermediaries

The next Web Science 2016 presenter is Jack Webster, who focusses on music recommender systems. Such recommender systems could generate filter bubbles, but that threat is nothing new; the cultural intermediaries described by Pierre Bourdieu fulfilled very similar roles and could have engendered very similar patterns.

But music recommender systems do not draw on a lifetime of personal experience, but rather utilise large datasets on users' engagement with music streaming and download services, applying algorithms designed by human coders. How do these different human and technological actors come together in creating and operating recommender systems, then?

Cultural intermediaries are the contextual actors operating in a given field, and their actions and positioning in the field and regulated by habitus. If we consider recommender systems as a new type of cultural intermediaries, they can be seen as a power operating from within the music system – but they need to be further conceptualised.

Cultural capital is the basis of cultural intermediaries' authority; they are positioned in a structured field of relations. But as technological systems the recommender systems also have an agency of their own, enabled and constrained by both human and technological actors. Actor-network theory can be used to better understand this complex set of interactions.

Capital is the accumulation of labour, and delegation is the transformation of a major effort into a minor one. Recommender systems are a product of such delegated labour – so what are music recommender systems' valued cultural assets? Such systems rely on users and their ratings: human actors contribute insights into their tastes that can be evaluated by the system using a range of mathematical models. This can be seen as one aspect of the systems' cultural capital.

Non-human actors take on the selective attitudes of their engineers, including ethics, values, and duties – habitus. The recommendation of new items that are similar to a user's known tastes is regulated by habitus in the form of users' past music uses and ratings which are operationalised by the system. There is therefore a constant interaction between the competencies of the human and technological actors.

Such systems then provide a lens through which to view changing cultural consumption preferences and practices, especially building on large datasets of consumption.