Diversity for what? An analysis of normative democratic assumptions in the domain of personalized news recommender systems

This paper presents a framework for linking notions of diversity to democratically sustainable recom-mender systems for news. Scholars have pointed out that news diversity is not an end in itself, but ra-ther a means for the perpetuation of democracy (Helberger, 2015; Joris et al., 2020). Helberger (2019)has argued that the conceptualization and role of diversity in news recommender systems changes inrelation to the democratic theory endorsed, both in the analysis and critique of such systems and intheir design. To better understand whether and how these different conceptualizations of diversity areoperationalized, we analyzed the current literature on diversity and news recommender systems fromthe field of computer and social sciences, using framing analysis (Entman, 1993). Firstly, we found thatthe conceptualization of diversity in news recommender systems is implicitly designed and evaluatedaccording to the participatory democracy model, which emphasizes political participation and active cit-izenship. Secondly, we found that literature from both fields have a techno-centric framing of personal-ization, which results in viewing challenges to personalization as mainly technological problems withtechnological solutions and narrow, operational, measurable definitions of diversity. We argue that thecurrent operationalizations of diversity focusing on issues bias, source diversity and content diversityoverlooks other dimensions of personalization, which are important when designing democraticallysustainable personalization systems. These missing perspectives we would argue also explain, why manypersonalization projects fail in news journalism (Møller Hartley et. al forthcoming 2023, Schjøtt Hansenand Møller Hartley 2021). Finally, the article introduces an Earth Democracy perspective (Shiva, 2006)to argue for more holistic design and evaluation criteria linked to three dimensions of democratic sus-tainability; namely 1) democratic collectiveness, 2) algorithmic reflectiveness and 3) ecological resourcedependence.