Do citizens engage in content moderation of toxic speech against historically marginalized groups?
Jesper Sommer Rasmussen (Lasse Lindekilde, Tanja Marie Hansen, Simon Tobias Karg)
Do partisans engage in crowd moderation of toxic speech targeting historically marginalized groups? We provide a comprehensive test of whether and when individuals engage in crowd moderation of toxic speech targeting gender, ethnic, and religious groups. The results show that, on average, individuals are more likely to engage in crowd moderation when toxic speech targets historically marginalized groups and minorities. These findings are robust across cross-national differences in cultural norms, free speech traditions, and political contexts in the United States, Germany, Denmark, and Italy. However, we find that this effect is conditional on partisanship and that variation in the target of speech generates greater differences on the political left than on the political right. The findings have important implications for when and whether citizens engage in crowd moderation.
Protecting financial stability in the age of social media – are online platforms our only hope?
Katarzyna Parchimowicz
Social media is reshaping the financial system as we know it. In recent years, we have witnessed several instances of unexpected volatility in global markets, undermining financial stability. In 2021, users of the Reddit group WallStreetBets coordinated to boost GameStop’s stock price via the Robinhood trading app, aiming to hurt investment funds that had bet on its decline. As a result, GameStop’s value surged by 600% in four days, markets panicked, and Robinhood suspended trading in these stocks. This instability was driven by social media (Hayes et al., 2022). Similarly, the run on Silicon Valley Bank (SVB), was amplified by social media – news of SVB’s alleged financial weakness spread rapidly through its highly interconnected depositor network on Twitter/X (Cookson et al., 2023). Its collapse sent shockwaves through the global financial system, marking the first Twitter-fueled bank run and the largest, fastest bank run in history. Unfounded chatter about Credit Suisse’s insolvency also contributed to its demise in 2023 (Egkolfopoulou, 2023). Comparable dynamics appear in crypto markets, where celebrity endorsements can move prices. Elon Musk's comments on Dogecoin and Bitcoin are well‑known examples (Cary, 2021). Importantly, these events fall outside financial regulation. They rarely meet the criteria of market manipulation (Gale, 2022; Möslein and Jetzen, 2024) and no macruprudential framework treats them as systemic risk. Occasionally, actions by “finfluencers“ are subject to advertising laws, but those only address the issue after the fact. Overall, regulatory and supervisory tools are too slow to counter social media-triggered financial instability – especially as false information spreads faster than the truth (Vosoughi et al, 2018). This research project aims to analyze content moderation regimes (USA, EU, UK) in order to see whether they could be adjusted to extend their protection also to financial stability. Main legal acts explored include Digital Services Act (DSA - EU), Communications Decency Act (USA) and Online Safety Act (UK). As the DSA offers the strongest potential in this area, the analysis focuses on its core mechanisms - notice‑and‑action, the definition of illegal content, and the list of systemic risks - to assess how they might be adapted to safeguard financial stability without giving platforms disproportionate influence over markets or financial expression.
Theories of risk in the Digital Services Act
LK Seiling
Understanding social media influence increasingly depends on access to platform data — and under the Digital Services Act (DSA), the scope of that access is tied directly to the DSA's conception of risk. The DSA establishes special due diligence obligations for very large online platforms and search engines (VLOPSEs) to manage systemic risks (Husovec, 2024), covering platforms like Facebook, Instagram, TikTok, and X.com that reach more than 45 million recipients in the EU. Critically, researcher data access under Article 40 is purpose-limited to these systemic risks. This means that what researchers can study is, in part, determined by how systemic risk is defined — yet the DSA offers no clear definition, providing only non-exclusive lists of risk categories and mitigation measures (Kaesling, 2023a, 2023b). For researchers seeking to understand social media influence through DSA-enabled data access, this indeterminacy creates uncertainty about what phenomena are legible and researchable within the regulation’s framework. This presentation addresses this indeterminacy by developing a theoretically grounded account of systemic risk in the DSA without falling into the trap of “premature standardisation“ (Loi et al., 2025). By drawing on the DSA and established risk literature, it provides a meta-theory of both systemic risk governance as well as a risk formation on digital platforms. It argues for broadening the focus beyond the framing of much existing academic work, which describes the DSA as a mere “risk management“ exercise (Griffin, 2024; Griffin & Fornasari, 2025), and proposes the concept of “systemic risk governance“ (IRGC, 2018) as an alternative, establishing a system-based approach to risk. This is then used as the baseline for the integration of risk factors mentioned in the DSA with classical risk literature to theorise the role platforms play in the formation of “systemic risks“.
From Participation to Presence: Informal and Ephemeral Citizenship in the Age of Social Media
Jakob Linaa Jensen
The Internet, and particularly social media, has profoundly transformed contemporary citizenship practices, not by revitalizing traditional forms of political participation, but by reshaping how citizenship is enacted, experienced, and sustained in everyday life. This article contributes to debates on distributed and mediated citizenship by presenting a longitudinal analysis of changing citizenship practices in Danish national elections over a fifteen-year period (2007–2022). Conceptually, the study builds on Couldry and Livingstone’s public connection project, which understands citizenship less as a set of discrete participatory acts and more as an ongoing orientation toward shared public concerns maintained through media use (Couldry, Livingstone & Markham 2007). In addition, the article draws on Hallvard Moe’s work on digital citizenship, which highlights how political engagement is increasingly fragmented, individualized, and embedded in routine media practices, often detached from sustained or institutionalized participation (Moe 2020). Together, these perspectives provide a framework for understanding citizenship as both distributed across media environments and increasingly ephemeral in form. Empirically, the article draws on five repeated survey studies conducted across Danish national elections between 2007 and 2022. The longitudinal design allows for a fully comparable time-series analysis of changes in social media use, attitudes toward digital political engagement, and perceived democratic effects. The findings show that participation during elections has increasingly taken the form of informal, low-intensity practices such as liking, following political actors, sharing content, and using voting advice applications. These practices rarely translate into sustained or formal political engagement. Instead, they point to a mode of citizenship that is situational, affect-driven, and temporally compressed, adjusting to the rapid circulation and short attention cycles of social media. Overall, the article demonstrates how citizenship in digital democracies is not so much expanded as reconfigured: from stable, institutionally anchored participation toward informal, mediated, and ephemeral practices. Rather than reversing the long-term decline of formal political participation, digital media have contributed to a gradual informalization of citizenship, characterized by episodic, low-threshold, and often fleeting forms of engagement.
Concepts for the Democratic Ramifications of Social Media Platforms – A Scoping Review
Ulrikke Dybdal Sørensen (David Budtz Pedersen, Joachim Wiewiura)
Concerns about how new information and communication structures affect democracy are not new, but today’s unprecedent scale and speed of digitisation, platformization, and technological advancements have intensified debates about how platforms are altering democratic infrastructures. In a prominent systematic review Lorenz-Spreen et al. (2022) provide grounds for concerns as the existing empirical evidence depicts an online platform environment very loosely linked to democracy. Yet, while the review is often referenced as an empirical basis for discussion, it raises the need for theoretical and conceptual considerations and grounding. Especially as the digital transformation has resulted in a diversification of key concepts such as democratic debate online, digital public debate, online sphere(s), digital public(s), etc. As a result, key concepts and core terms are variously applied, interchangeably conceptualised, ascribed differing meanings, and/or characterised by fragmented attributes in social media research. Still, despite the centrality of these concepts and the potential conceptual challenge, no systematic overview of the basic yet often used concepts in the research literature currently exist. We address this gap by exploring how the same concepts can be assigned different meanings, how different concepts are assigned the same meaning or used to denote similar phenomena, and how notions of public debate, public(s), and public sphere(s) risk being conflated in research on social media. We do this by taking a systematic look on key concepts used as signifiers for the democratic ramifications of social media platforms by exploring available social media research. We do this with the concept of democratic debate online as a starting point since it forms a central signifier for the challenges facing democracy in a digital age. Especially within a Danish context where democratic debate online is a recurring focal point in policy discussions of social media, democracy, and governance (Regeringen, 2021). Hence, we provide an overview of various conceptual components currently used for the intersection of social media and democracy. Subsequently, we contribute to a better understanding of the systemic risk category civic discourse in the Digital Services Act as it closely parallels democratic debate online including the conceptual ambiguities and challenges (Calabrese & Reich, 2024; Griffin, 2025).
Considering neural reward mechanisms in regulating social media algorithms to reduce polarizing content
Michael Cohen (Jean Decety)
Social media platforms likely bear some responsibility for the accelerating spread of false and polarizing content in recent years. Such content has poisoned public discourse in open societies and has emerged as a major risk to democracy. In a recent publication, we synthesized studies from psychology, neuroscience, and computational social science, to make that case that the nature of the brain’s reward system may lead algorithms prioritizing content evoking likes and shares to inadvertently prioritize content that harms society. Specifically, there is evidence that both giving and receiving “likes“ activates the ventral striatum, a core reward-sensitive brain region. The reward system is also critical in motivating decisions to share content. Other studies have associated reward system activation with willingness to fight and die for a sacred cause and with approval of violent protests. Furthermore, when a social media post expresses negative emotions towards the opposing political party, it is much more likely to be shared. Content from Internet domains that tend to include false stories, including domains run by the Russian government’s Internet Research Agency, have also been shown to evoke outrage and to spread more quickly than truthful content. An additional consideration emerging from the neuroscience literature is that the reward system motivates behavior while not always aligning with pleasure. This helps to explain another recent finding from survey data: while people are aware that false and polarizing content actually goes viral, they would prefer that more wholesome content would go viral. Ultimately, this analysis suggests that social media algorithms must be regulated. The EU Digital Services Act forms a good model for such regulation with its focus on reducing systemic risks to society, but a stronger evidence base is needed to support this regulatory process. I will present some possible approaches by which human neuroscience could inform regulation. For instance, the neural reward response to specific social media posts could be measured in a small sample of participants (following the tradition of “neuroforecasting“ in communication science), and this data could be used, along with behavioral ratings of anger and/or other negative emotions, to train an algorithm on what type of content is likely to be harmful. One goal of the presentation will be to stimulate discussion towards constructing future empirical research projects.
Can cross-cutting exposure and echo chambers coexist? An agent-based-modelling approach
Jens Ulrik Hansen (Eva Mayerhöffer, Frederik Henriksen)
Debates about echo chambers on digital and social media platforms persist despite extensive empirical research. Disagreement may stem less from contradictory findings than from differences in analytical approaches, operationalization, and measurement strategies (Hartmann et al. 2025). Survey-based studies frequently report that individuals encounter counter-attitudinal content (Brundidge 2010; Wojcieszak & Mutz 2009), whereas large-scale network analyses document ideological clustering in interaction and sharing systems (Cinelli et al., 2021). At the same time, research on selective exposure shows that individuals often prefer attitude-consistent information, though avoidance of counter-attitudinal sources is neither universal nor uniform (Garrett, Carnahan, & Lynch 2013; Garrett & Stroud 2014). Rather than treating these findings as contradictory, this talk examines the possibility of their co-occurrence. To this end, we construct an agent-based model that allows for cross-cutting information exposure and sharing we see in digital information environments, and test whether such environments still facilitate ideological clustering and polarization. We draw on a representative survey of Danish citizens, conducted in April 2025 (N = 1,900) to empirically ground the initialization of an agent-based model (ABM) that simulates social media users’ posting and re-sharing of aligned and cross-cutting content and attitudinal change. With this ABM we can thus investigate the complex system of information exposure, sharing and update as well as its further configurations that may lead to ideological polarization. Such further configurations will primarily include network structure, that is the patterned organization of ties between actors—such as clustering, centralization, and homophily—which shape how information flows and which viewpoints are reachable.
