Online platforms have become infrastructural to democratic life, yet their influence is often examined through narrow causal-effect models that overlook systemic socio-technical dynamics. This Forum contribution introduces the concept of platform collective behavior to conceptualize how intertwined human-AI agency generates population-level patterns with systemic risk implications for democracy and citizen well-being. Rather than focusing on isolated persuasion effects, platform collective behavior captures emergent socio-algorithmic patterns such as unequal political participation, differential exposure to misinformation, bot amplification, and subgroup information segregation. The article evaluates whether Article 40 of the European Digital Services Act (DSA), which grants vetted researchers access to platform data, enables robust investigation of such systemic risks. While the DSA represents an unprecedented regulatory intervention in research data access, its scope is limited to systemic risks and depends on platform-provided datasets. The analysis identifies blind spots, including infrastructural and economic power asymmetries, offline behavioral consequences, and data integrity concerns. The article concludes that DSA access is necessary but insufficient for fully assessing platform influence on democracy.
Using cross-lingual international X trends data and a unique role-based analytical framework well-suited for the study of granular textual data, this study examines the partisan and spatial distribution of political rhetoric in hashtag activism about the Gaza War in its early phase. Following a triadic classificatory rubric of roles composed of victim-villain-hero, I find that pro-Palestine hashtag activism was overwhelmingly focused on victimhood, but Muslim-majority countries in the Middle East uniquely exhibited comparably strong expressions of glorified sentiments of heroism. In contrast, pro-Israel hashtag activism was driven by villain-centric messages focusing on the topic of denunciation mixed with the rhetoric of victimhood. Contrary to popular characterizations of the discursive conflict as a parallel clash of victimhood or vengeful emotions, the results reveal stark asymmetries in how wartime militancy was rhetorically framed. I propose a power-centric interpretation of the observed rhetorical asymmetries, which I argue stem from diverging practical needs in a heavily unequal bilateral conflict. I also reflect on regional differences within the global Palestine solidarity movement amidst debates about the cultural praxis of subalternity that have intensified since the outbreak of the Gaza War.
Search engines like Google have become major information gatekeepers that use artificial intelligence (AI) to determine who and what voters find when searching for political information. Yet there is concern that biases in algorithmically curated representation of women politicians in online search results mirror and amplify structural inequalities and therefore further reduce women's electoral chances. We empirically assess the prevalence and consequences of algorithmic gender bias in political Google searches in a series of four studies. First, we conduct two multinational algorithm audits of political image searches (study 1: 56 countries and 6,363 images; study 2: 11 countries and 152,098 images) to assess women's quantitative representation in search results. We find that women politicians' algorithmic representation on average accurately mirrors the actual gender composition of legislative bodies while consistently remaining below absolute gender parity. Second, we show in two online experiments from three different samples (1,388 respondents in total) how more extreme cases of algorithmic underrepresentation can influence human perceptions of the political reality and actively reinforce a white and masculinized view of politics. These misperceptions act as a casual conduit through which algorithmic biases can not only diminish the perceived chances of winning an election for minoritized candidates but also result in voters feeling that their voices matter less. Together, the results highlight the risk that, under certain conditions, algorithmic systems like search engines can negatively impact electoral processes.
In democratic systems, relationships between politicians and journalists are often marked by both interdependence and tension. Prior research shows that when politicians perceive traditional news media as biased and voice these concerns, their interactions with journalists can become increasingly conflictual. However, little is known about what politicians actually mean when they blame legacy media for being slanted, or what they perceive as the underlying causes of such partisan news bias. Understanding these perceptions matters, as they shape politicians' interactions with journalists, their broader communication about the media, and, ultimately, the quality of democratic debate. Therefore, this study investigates politicians' perceptions of partisan news bias through face-to-face interviews with 110 Belgian members of parliament. Specifically, we examine how politicians define partisan bias and what they consider its main causes. We find that MPs primarily refer to selection and presentation bias when criticizing legacy media, frequently citing insufficient attention, unfair framing, and selective quoting as grievances. Both right- and left-leaning politicians attribute these issues to the presumed left-wing orientations of journalists. Right-wing MPs argue that journalists' ideological bias benefits leftist parties, whereas left-leaning MPs believe journalists disadvantage the left to avoid accusations of leftist bias. To a lesser extent, MPs also identify commercial pressures as a source of political bias in news coverage. Overall, the findings show that politicians attribute considerable power to the media and express a shared sense of dependence on its perceived biases and agendas.
Journalistic and scholarly accounts alike often depict social media platforms as key players in exposing people to misinformation. Unfortunately, existing research examining social media's role in exposing people to misinformation online implicitly operates under what we call the direct referrer paradigm, focusing only on visits to misinformation that directly follow social media. We introduce the embedded referrer paradigm, which considers social media's more comprehensive role in leading people to visit misinformation throughout their online journeys and operationalize it via counterfactual simulations. We estimate the effects of social media in exposing people to misinformation websites under both paradigms using 3 months of a large sample of Americans' (N = 1,240) web browsing behavior (21 million website visits). We explore seven different platforms' (Facebook, Instagram, Reddit, Snapchat, TikTok, Twitter, YouTube) effects in exposing people to misinformation sites with various ideological leanings. Overall, our results suggest that analyses under the direct referrer paradigm underestimate the role of social media in exposing people to misinformation, with estimated effects under the embedded referrer paradigm ranging from 1.5 to 15+ times larger than effects under the direct referrer paradigm. The difference in estimated effects under both paradigms is generally consistent across platforms, misinformation types (liberal, conservative, non-ideological), and individuals' partisanship. However, for some combinations of platforms, types of misinformation, and mobile vs. desktop devices, the discrepancy between the two paradigms is even greater.
To what extent does the YouTube recommendation algorithm push users into echo chambers, rabbit holes, or radicalization pathways? Using a novel method to estimate the ideology of YouTube videos and an original audit design, we demonstrate that YouTube users are in mild ideological echo chambers, but this is primarily driven by user behavior. We also demonstrate that the recommendation algorithm prioritizes content that is similar to what the user is currently watching, producing content rabbit holes. However, we do not find evidence of radicalization pathways where users are driven into increasingly extreme content rabbit holes. Instead, we find that YouTube pushes all users, regardless of ideology, toward moderately conservative content and an increasingly narrow range of ideological content the longer they follow YouTube's recommendations.
Misinformation is increasingly seen as a key challenge to democratic societies. Our study is one of the first to shed light onto the citizen perspective when it comes to the perceived target of misinformation during election campaigns. In doing so, we extend on a classic concept in the (political) communication literature, the hostile media effect, and examine whether this applies to misinformation as well, a so-called hostile misinformation effect. Do citizens believe that their political ingroup is being targeted more by misinformation than their political outgroup? Our argument is based on motivated reasoning and social identity theory and extends to the role of several crucial moderating factors. Using data from a panel study conducted during the 2024 European Parliament elections across Germany, the Netherlands, and Poland (N = 4,045), we find clear support for a hostile misinformation effect, as citizens believe their own political party was much more targeted than the political opponent. Moreover, we demonstrate that particularly political interest, party identity strength, ideological extremity, and being right-wing make people more susceptible to the phenomenon. Our findings demonstrate that the hostile media effect can be extended to the domain of misinformation perceptions. Moreover, they explain why people perceive to be surrounded by misinformation, and help contextualize literature suggesting that people associate misinformation with various other information disorders and threats.
Are democratic institutions under threat in the United States? We examine an understudied democratic bedrock: the judiciary and, in particular, the U.S. Supreme Court. We draw on data from "SCOTUSapp," a tool that provides a database of 100,000+ news articles across thousands of unexamined sources. We find a new media environment where the Court receives attention from thousands of ideologically fragmented outlets. Additionally, some framings of the Court are potentially corrosive. We then use two experiments to examine the effects of these corrosive news frames. The first experiment (October 2021) provides evidence that frames portraying the Court as embroiled in secrecy, hypocrisy, illegitimacy, and politics cause a decrease in support for judicial independence. These frames are influential among Republicans and Democrats. The second experiment (May 2023) follows Dobbs, which severely damaged judicial faith among Democrats, and shows that corrosive frames lead to a corresponding decline only among strong Republicans.
Despite increasing prominence in the contemporary media ecosystem, podcasts have attracted limited scholarly attention. This study examines audience characteristics of the most popular podcast in the United States, The Joe Rogan Experience (JRE), situating it within the broader cultural context of the "manosphere." Leveraging two national probability samples collected in October 2022 (N = 2,907) and November 2023 (N = 1,000), this research explores the socio-demographic profile, information environment, epistemic orientations, identity, ideology, and trust of Rogan's listeners. Across both studies, among the strongest correlates of JRE listening is conservative media consumption, with additional associations with skeptical epistemic orientations, conspiracism, political and institutional distrust, and being young and male. In contrast, partisanship is not significantly associated with JRE listening once these other factors are included in the models. Contrary to common portrayals, hostile sexism, authoritarianism, and racial attitudes do not reliably predict JRE listening once media diet is accounted for. This research underscores the importance of examining audiences' deeper cultural values and epistemic orientations alongside traditional political variables, thereby clarifying our understanding of how influential media figures such as Rogan shape and reflect broader currents in an increasingly fragmented information environment.
The Digital Services Act (DSA) creates unprecedented opportunities for research into platform-driven systemic risks, but researchers outside Europe face practical and theoretical constraints. This article examines these limitations and identifies pathways for productive engagement. For most non-European researchers, GDPR compliance creates a fundamental choice. They can work with non-personal data from anywhere, or work with personal data only within the EU - a choice with significant implications for research design. Article 40's focus on systemic risk in Europe further limits which questions qualify for access. In this article, I identify three categories of research that fit within this scope while addressing globally relevant questions: studies of core platform characteristics, research on cross-border dynamics affecting European users, and comparative work contextualizing European risk. Researchers who want to take advantage of these opportunities also have a responsibility to engage with the DSA framework in good faith, both to maintain its legitimacy and to ensure this regulatory experiment generates knowledge that genuinely serves the European public.
