Conceptual work:
From the theoretical side, we developed a conceptional framework for understanding the impact of social media on liberal democracy by examining how social media transform citizenship, identity, and the public sphere in liberal democracies. Its main result was to move beyond binary claims that social media are either democratic or anti-democratic and instead to develop a historically contingent yet robust framework that defines democracy more clearly and shows that digital platforms can simultaneously widen participation, intensify polarization, strain trust, and alter the conditions of public reasoning and political inclusion.
Providing evidence:
We successfully connected the overarching theoretical framework — grounded in polarization, trust, and participation as the principal dimensions of digital citizenship — with empirical case studies examining important emerging forms of political participation: online calls to participate in petition-signing campaigns and fundraising drives, but also new practices such as open source investigations (OSINV). This objective was achieved by operationalising polarization and trust as measurable dimensions, which we inferred for large populations of social media users across nine EU countries using ideology scaling calibrated with political survey data.
Concerning the methodological advances, the project was successful in developing an inductive method for the extraction of narratives from large textual corpora. The method rests on the idea that narratives are often not present in their entirety in digital traces, but rather alluded to through narrative traces. Such traces include specific actors, their actions, their motives, and the other actors they interact with. The traces are re-assembled as networks that allow insights into the dominant narratives present in the analysed corpora and their role for issue alignment and polarization.
Modelling work:
The overarching objective of modelling activities in SoMe4Dem has been to advance agent-based models (ABM) as viable tools for digital policy support. In order to achieve this goal, we integrated different empirical insights into computational models to allow a collective-level analysis of how platform affordances interact with adaptive user choices.
Within the project, significant progress towards more realistic models of online platforms has been made. We developed a framework which focuses on how users in community-based platforms like Reddit select between different available actions, such as choosing to which subreddit to contribute and tested the collective effect of different ranking and recommendation algorithms in synthetic message streams generated by ABMs.
Improving algorithms and interfaces:
The objectives were to understand and improve how platforms and users can jointly foster more competent, democratic use of social media, and to test scalable interventions that enhance citizens’ digital literacy. The objectives were pursued by two complementary routes: (1) modifying recommender algorithms and interface elements to surface epistemic-quality cues and produce more trustworthy recommendation rankings; and (2) boosting citizens’ competences via behavioural interventions, for example through lateral reading.
The first route resulted in developing and validating a family of context‑based indicators (“exogenous cues”) that let ranking algorithms favour higher‑quality news without ever inspecting content itself, and by showing that such rankings systematically improve the information environment. For the second route, two interventions - online search and lateral reading were tested. Behavioural measures suggest increases in information search in both intervention groups. On average, lateral reading reduced trust in untrustworthy sources, online search increased trust in trustworthy sources.