Making digital publics: algorithmic media and the fabrication of publ...

Recommender systems reorganize the dissemination of information and thereby give way to new kinds of publics. The increasing use of recommender systems can thereby be understood as a new algorithmic regime of information distribution that has the potential to splinter the public sphere into many different small-scaled publics by creating filter bubbles (Pariser, 2012) or echo chambers (Sunstein, 2009). This poses a challenge for democratic discourse .However, while there is some evidence for the formation of filter bubbles (O’Callaghan et al., 2013), other findings speak against their existence (Borgesius et al., 2016; Haim et al., 2018). This indicates that the emergence of a filter bubble is contingent and context-sensitive. In our contribution we argue that different algorithmic techniques (Rieder, 2017) for recommendations impact the construction of publics. While not determining how publics are constituted, the algorithmic techniques mediate between actors according to their own specific logic, impacting the way these publics are constructed (Ang, 1991). Drawing on empirical fieldwork on the development of a public broadcasting recommender system powered by machine learning, we contrast two prominent ideal-typical algorithmic techniques for recommender systems. Collaborative filtering and content-based filtering both aim to provide personalized information. However, the techniques connect different actors in a wider network of interactions and make use of tracking data in a different way to computationally produce publics. In our contribution we investigate these algorithmic techniques to understand how they relate manifold interactions between news providers, data production, and users to producing different publics. Describing algorithmic techniques is essential to understanding new algorithmic forms of information distribution.Critically analysing how algorithmic techniques are embedded in and mediate between databases, interfaces,and practices sensitizes us to the ways that the formation of digital publics. This also opens up perspectives for rethinking and redesigning new algorithmic regimes of information distribution.