Social Media as Artificial Selection: The Cultural Evolution of Conspiracy Theories
Morgan Wack (Eva-Maria Vogel, Christian Pipal)
Why do conspiracy theories persist online despite sustained scrutiny from fact-checkers, moderators, and ordinary users? The dominant contagion metaphor treats misinformation as a static pathogen that can be neutralized through debunking. However, recent qualitative research tracking the participatory dynamics behind narrative construction has shown how conspiracies are collectively constructed (Starbird, DiResta & DeButts, 2023). Ordinary users evaluate claims against available evidence, refine their framing, and collaboratively produce increasingly sophisticated allegations. We argue that this participatory process constitutes a form of cultural evolution in which platforms function as selection environments where users mutate, transmit, and differentially retain narrative variants. We use an agent-based model to formalize this argument. Conspiracy narratives possess evolvable genomes (vectors of attributes including consistency, emotional valence, unfalsifiability, and explanatory scope). Users evaluate and adopt narratives according to heterogeneous epistemic styles and, upon adoption, may produce mutated variants through the same kind of collective refinement documented in studies on election fraud narratives (Prochaska et al., 2023). Community members and fact-checkers apply differential selection pressure, removing easily refuted variants while enriching the surviving population with harder-to-rebut ones. This produces a “debunking paradox“ in which the participatory dynamics of online communities inadvertently accelerate the evolution of resistant conspiratorial narratives. We test these predictions using 1.25 million Reddit posts from four conspiracy and three control subreddits. As predicted, we find that fact-check difficulty increases over time only in conspiracy communities. At the same time, within-user difficulty trajectories are flat, yet the difficulty floor encountered by newcomers rises steadily, pointing to selection on narratives rather than individual radicalization. These findings can inform platform governance. If the participatory dynamics that communication scholars have documented also function as a form of evolutionary selection, then interventions targeting individual claims may accelerate the very problem they aim to solve. We conclude by drawining on model counterfactuals comparing low-connectivity (pre-internet) and high-bridging (social media) environments to discuss evolutionarily informed approaches to content moderation.
Who Is Exposed to Migration Content on YouTube and TikTok? A Data Donation Study
Silva Heinonen (Chiara Caterina Arena, Benjamin Nangle, Brahim Zarouali)
Social media platforms play a central role in shaping contemporary information environments (Zhuravskaya et al., 2020; Kubin & Sikorski, 2021), particularly around highly politicized issues such as migration (Gethin et al., 2022; Simonsen & Bonikowski, 2022). Despite growing concerns about online polarization, misinformation, and hostility toward migrants, empirical research still lacks reliable measures of what users are actually exposed to in their day-to-day platform use. Most existing studies rely on self-reported social media use and exposure, providing only limited insight into individual-level exposure (Ohme et al., 2024; van Driel et al., 2022). This study addresses this limitation by introducing a data donation approach to capture real-world exposure to migration-related content on YouTube and TikTok among Belgian and Italian general population (N ≈ 400). The exposure data are linked to a three-wave panel survey conducted over six months. Participants voluntarily donate their platform watch histories, enabling a measure of individualized content exposure. Using a combination of computational and manual content analysis, we identify and classify migration-related and potentially harmful content within these data. We further examine how exposure varies across individuals by linking content exposure to socio-demographic characteristics, political attitudes, and psychological differences in information processing. We present preliminary findings from the first wave of data collection in Belgium and Italy (N ≈ 400), focusing on predictors of exposure to migration-related content. Methodologically, the paper demonstrates the potential of data donation as a scalable approach to studying platform influence. Additionally, it contributes to ongoing debates on selective exposure, algorithmic curation, and the role of digital platforms in structuring political information environments.
Seeking Truth with Low Trust – How Politically Alienated Citizens Navigate Mainstream and Alternative Media in Germany
Carolin-Theresa Ziemer & Pietro Leonardo Nickl
Why do people choose to abandon traditional news media and seek out alternative information diets? The growing fragmentation of information environments poses a fundamental challenge to shared perceptions of societal events and how they should be addressed. In Germany, trust in established news media has declined markedly since 2015, while engagement with alternative media has become a stable feature of the media landscape (Newman et al., 2025). While existing research maps the content of alternative media (Freudenthaler & Wessler, 2022) and typologizes their users (Schwarzenegger, 2022), we lack a deeper understanding of why people drift toward alternative media outlets, their criticism of traditional journalism as well as their perceptions of alternative media. The present research explores media practices of politically alienated individuals in Germany. Drawing on qualitative data from 50 semi-structured, in-depth interviews and two focus groups with individuals exhibiting low institutional trust, the analysis focuses on how participants engage with mainstream, alternative, and social media. We examine how individuals select and evaluate different sources of information, and what they criticize about mainstream media. In contrast, we explore how alternative and social media are perceived in terms of credibility, independence, and relevance. Some participants describe alternative media as offering “authentic“ voices and more “unfiltered“ information, particularly in live-stream formats, and emphasize their role as an independent counter-public. They criticize mainstream media for its perceived partiality, ideological bias and negativity. Particular attention is given to shifts in media use, including tendencies toward news avoidance, especially in the context of perceived turning points such as the COVID-19 pandemic. By focusing on the interplay between media environments and epistemic orientations, this study contributes to a more nuanced understanding of how alternative realities are constructed and sustained. It also speaks to debates on media trust, epistemic fragmentation, and political alienation. We situate our findings within discussions on susceptibility to misinformation and the limits of current literacy interventions. When individuals dismiss mainstream media as biased and place greater trust in alternative sources, widely promoted strategies such as lateral reading, which rely on established outlets’ credibility, may lose effectiveness.
Global Crisis or Overblown Problem? Three Tools to Clarify Contentious Issues in Misinformation Research
Mubashir Sultan & Pietro Nickl
As the field of misinformation research undergoes rapid expansion, it simultaneously faces a dual crisis that threatens its future impact. Externally, a political crisis manifests through funding cuts, anti-science sentiment, and a retreat from evidence-informed policy, with major platforms scaling back content moderation under the guise of avoiding political bias. Internally, the field is gripped by an academic crisis: a series of polarized debates regarding its own relevance and legitimacy. High-profile articles are questioning whether misinformation is a global crisis or misunderstood and overblown. We dissects three key contentious issues shaping this debate: the legitimacy of misinformation research, the prevalence of misinformation, and its causal role in societal harm. We argue that while these challenges are significant, they do not undermine the field’s validity; rather, they highlight the need for greater conceptual and methodological transparency. We introduce three conceptual tools for clarity. The (Mis)information Dimensions: This tool addresses legitimacy. Because empirical research requires classifying content as false or misleading, critics often ask why researchers should act as ""arbiters of truth."" By using the Dimensions framework, we explore the epistemological challenges of inferring truth, intent, and harm, arguing for a more transparent communication of assumptions. The (Mis)information Funnel: This tool addresses prevalence. Some studies suggest misinformation constitutes only a small fraction of online consumption, leading to claims that the problem is negligible. The Funnel traces how these estimates arise, showing how pragmatic methodological choices, such as defining misinformation solely as specific URLs, often result in ""lower bound"" estimates that obscure the true scale of the problem. The (Mis)information System: This tool addresses causality. A growing debate questions whether misinformation is a primary driver of harm or merely a symptom of deeper issues like distrust. The System framework moves beyond this false dichotomy, mapping specific causal mechanisms to determine at which level interventions, ranging from individual ""inoculation"" to systemic policy changes, will be most successful. We argue that there are fewer genuine disagreements in the field than appear at first glance. By providing a principled approach to these debates, we aim to resolve apparent contradictions and systematically map future research directions.
Strategic Social Media (Non-)Use by Diaspora During War
Mohammad Javad Imani Khoshkhoo (Arul Chib, Mona Jormand)
Diasporic communities rely heavily on social media to balance integration into host societies with an urgent need for homeland information. However, as during conflicts information scarcity increases, migrants are yet exposed to fake news environments characterized by rampant misinformation. We explore the distrust in social media faced by Iranian migrants in the Netherlands during the ongoing conflict. Such users face the dilemma of news overload and doubtful information while facing the necessity of remaining informed being concerned about their friends and family facing warfare. Conceptually, trust refers to a willingness to accept vulnerability to an information source and allow it to exert influence over the trustor. Alternatively, extreme distrust can lead to unawareness or the consumption of misleading content. The mobile media (non-)use model[i] posits that users adopt diverse strategies, such as self-control and blocking, as means to manage dis/trust. Social media (non-use) enables individuals to avoid less credible information sources. This study conducted 2 focus groups (n=9) and 20 interviews on how Iranian diaspora navigate social media-influenced distrust by adopting contextual strategies of digital (non-)use. The research study illuminates how trust is negotiated vis-a-vis general skepticism toward both formal (legacy digital news such as the BBC and Iran Intel) and informal (relatives, bloggers) social media networks. The analysis explores the role of algorithmic curation, the influence of personal networks in correcting misinformation, and the gatekeeping functions of bloggers. Furthermore, it addresses the impact of online surveillance and state-sponsored cyber-troops in intensifying systemic distrust. Ultimately, the study highlights the complex, agentic adoption of information management techniques, specifically disconnection, disengagement, disguise, and discretion. We argue that diaspora actively manage their digital disconnection to mitigate risk during sensitive geopolitical crises. Finally, we reflect upon the efficacy of silent subversive acts of (non-)use by individuals as a political strategy as opposed to more visible organized social movements on social media.
Self-selection into digital literacy interventions against misinformation
Friederike Stock (Lisa Oswald, Mubashir Sultan, Pietro Nickl, Philipp Lorenz-Spreen, Ralph Hertwig, Anastasia Kozyreva)
Research in the behavioural and social sciences has introduced several interventions aimed at improving users' competencies and behaviours to curtail the spread of misinformation. While such digital literacy interventions generally improve accuracy discernment, critical gaps remain regarding their targeted effectiveness and real-world uptake. This study addresses three key questions: For whom are digital literacy interventions most effective? Who voluntarily engages with these tools when given a choice? And does the act of self-selecting an intervention enhance its efficacy (a ""self-determination"" effect)? To answer these, we employed a large-scale, pre-post experiment utilizing a representative YouGov panel (N = 7,000) across Germany and the United States. The design crosses assignment mechanisms (randomized vs. self-selected) with three intervention types (Inoculation, Lateral Reading, Media Literacy Tips) to disentangle tool efficacy from user motivation. We estimate Conditional Average Treatment Effects (CATE) and model uptake preferences, combining standard Bayesian regression and propensity score matching. By isolating the self-determination effect and assessing preference-efficacy alignment, this research moves beyond average treatment effects to understand user engagement, offering insights for designing scalable misinformation interventions.
Inside partisan environments: The influence of congruence in news environments on veracity judgments and partisan attitudes
Ilse L. Pit (Mubashir Sultan, Lisa Oswald, Alan Tump, Ralf Kurvers)
" major finding in misinformation research is that politically congruent news headlines (i.e. aligning with one’s political beliefs) are more likely to be judged as true compared to incongruent headlines. However, previous studies have used equal numbers of congruent and incongruent headlines, which does not accurately reflect news environments on social media. This study tested the effects of the congruence of a news environment on individuals' veracity judgments of news headlines and on partisan attitudes. In an online experiment, 911 U.S. participants each observed 48 headlines, half true and half false, in one of three randomly assigned news environments: Republican-leaning (Republican-/Democrat-leaning headline ratio: 75/25%), balanced (ratio: 50/50%), or Democrat-leaning (ratio: 25/75%). Participants indicated for each headline if they believed it to be true, and whether they were familiar with it. We additionally measured partisan attitudes before and after the headlines were rated. Results replicated previous research, showing that participants were generally accurate about the veracity of the headlines, and that congruence of the headlines influenced the likelihood that participants judged them to be true. No difference on these effects was found between the different news environments, indicating stability across environments of varying congruence. We also found no difference in partisan attitudes between environments. These findings suggest that congruence effects exist at the headline rather than the environmental level and inform the use of this experimental paradigm for studying behaviour on social media.
