Automating Publics: A Multi-Level Framework for Analysing Algorithmic Public Formation

Beskrivelse

This paper advances a conceptual framework for analysing how publics are shaped through and across personalised, datafied, and hybrid media infrastructures. The increasing implementation of AI-driven and personalised content recommender systems across the media industry raises fundamental questions about how algorithmic curation reconfigures citizens’ engagement with issues of public concern (Birkbak & Carlsen, 2016a). While existing scholarship has typically examined personalisa-tion and in particular News Recommender Systems (NRS’s) at the level of single platforms or single news outlets - often focusing on measurable effects such as click behaviour (Einarsson et al., 2025; Haim et al., 2018; Möller et al., 2018) or the values inherent in the design of NRS (Møller, 2023; Schjøtt Hansen & Hartley, 2023)- there is a lack of theoretical frameworks capable of explaining how personalised media environments reshape public formation and linking these developments across individual, organisational, and societal levels. The framework mobilizes previous literature on three levels and combines concepts from audience studies with public theoretical and media sociolog-ical approaches to media production in order to develop a coherent framework for analysing the au-tomation of publics.
At the micro level, the model conceptualises personalised media use as a communicative practice of decoding within what we term personalised mediascapes: cross-platform flows of content characteristic of contemporary, datafied, hybrid media environments (Chadwick, 2013; Lai & Flensburg, 2020; Møller Hartley et al., 2023). Building on encoding/decoding theory (Hall, 1991) and research on media experiences (Ytre-Arne & Moe, 2021), this level highlights how individuals interpret personal-ised content normatively and emotionally. At the meso level, the model includes personalization algorithms as algorithmic assemblages (Ananny, 2016) and the cultural dimensions of algorithms (Seaver, 2017), conceptualizing NRSs as socio-technical systems that encode normative values and organisa-tional imaginaries into processes that match users with public issues. This builds on previous work showing that personalisation systems embed assumptions about users and their concerns (Bucher, 2018; Lomborg & Kapsch, 2020). Through “ethnographies of algorithms,” this level captures how personalisation collapses traditional boundaries between encoding and decoding as aggregated user data become integral to producing personalised content flows (Birkbak & Carlsen 2016b). Finally, at the macro level, the model mobilises the concept of issue publics (Birkbak, 2013; Dewey, 1927; Marres, 2005) to theorise how algorithmically distributed content shapes collective attention, contro-versies and the emergence of public issues. In contrast to platform-specific notions such as networked (boyd, 2011) or calculated publics (Gillespie, 2014), issue publics emphasise how shared stakes in problems generate publics across media sites and institutional arenas. This perspective enables analy-sis of how personalisation modulates the visibility, articulation, and coherence of issues across the broader media ecology.
Having presented the framework the article discusses how these levels are linked, and what these means for how researchers approach the analysis of personalised media production and use empiri-cally and methodologically.

Periode	9 apr. 2026
Begivenhedstitel	Controversies of AI Society
Begivenhedstype	Konference
Placering	Copenhagen, DanmarkVis på kort
Grad af anerkendelse	International