The Line of Argument: From Belief to Behavior, From Signal to Structure
The papers gathered here trace an arc that begins with the content of health misinformation, moves through the networks that circulate it, and arrives at the harder question of whether any of this circulation actually changes minds and bodies. Read together, they push back—repeatedly and from different methodological angles—against a naive “infodemic” story in which viral falsehood straightforwardly causes public-health harm. Instead they converge on a more structural claim: what matters is less the truth-value of any single claim than the architecture of exposure—who is positioned to see what, how often, and through which gatekeepers.
Predispositions, Pathways, and the Limits of “Fake News” as a Category
Adam2026-tz opens the empirical core of this arc by showing that conspiracy belief is not simply implanted by media exposure but co-produced by prior political predispositions—populism and mistrust—interacting with media diet. Its “marriage” model, in which alternative media contagion and mainstream media mitigation both operate but are filtered through selective engagement and counterarguing, anticipates a theme that recurs across the collection: exposure alone is a poor predictor of belief change, because audiences self-select in ways that amplify or neutralize what they encounter. Lyons2026-ca extends this logic to health content specifically, showing that low-credibility exposure is rare in aggregate but concentrated among a small, older, already-conspiracist-leaning population, and that this exposure travels through habitual referral chains between low-credibility sites rather than through search or social media discovery. Together these two papers reframe “misinformation exposure” as a symptom of pre-existing worldview and habit rather than a random contagion vector—a finding with direct implications for the causal papers discussed below.
Allen2024-av pushes this reframing furthest, empirically decoupling veracity from persuasive harm at Facebook-wide scale: flagged misinformation is more persuasive per view but reaches almost no one, while unflagged, technically-true-but-hesitancy-inducing content from mainstream outlets does 46 times more aggregate damage. This is arguably the collection’s most direct challenge to fact-checking-centric interventions, and it sets up a natural tension with Bollenbacher2026-vz, which instead builds a causal epidemiological model (SIRVA) to show that antivaccine Twitter content did measurably suppress vaccination and cost lives—albeit as a “lower bound” using platform-specific tweet geolocation rather than the survey-experimental logic of Allen et al. Read side by side, these two papers model the same underlying question—does online anti-vaccine content cause offline harm?—via incompatible epistemologies (revealed behavioral outcomes vs. controlled exposure-effect estimation), and their differing answers (modest but real epidemic effect vs. veracity-flagging is beside the point) mark the collection’s central methodological fault line.
Coordination as Infrastructure, Not Just Content
A second cluster shifts from individual exposure to the organizational machinery of circulation. Giglietto2022-0e951ac5 and Marino2023-9137f448 both build on Coordinated Link Sharing Behavior (CLSB) methodology to map how Italian covid-skeptic networks evade platform detection—through link laundering, first-comment link placement, image-macro cross-posting, and symbiotic relationships with “Alternative Influence Network” celebrities and legacy journalists. Marino’s finding that mainstream media remain the central nodes recirculating skeptic narratives (via remediation of TV and newspaper clips) directly complements Ducci’s institutional account: Ducci2022-10cb5d70 shows that Google News Italia, as an algorithmic gatekeeper, already privileges emotionally resonant, “popular” health topics (celebrity, malpractice) over technical medical-scientific content in terms of engagement, even though scientific topics dominate raw aggregation volume—suggesting that the affective architecture exploited by AIN actors is latent in “legitimate” gatekeeping infrastructure too, not just in fringe coordination.
Song2025-yh internationalizes this coordination lens, comparing pro- and anti-vaccine CLSB across the UK and US and arguing, crucially, that coordination itself is ideologically neutral—it can scale institutional public-health messaging (NHS, CDC) just as readily as it scales unverifiable anti-vaccine content. This is an important corrective to a “coordination equals bad actor” reading implicit in the Italian case studies, and it resonates with Efstratiou2026-ij, which finds that COVID science discourse on Twitter is shaped by a genuinely coordinated, contrarian retweet network amplifying a narrow set of credentialed anti-consensus experts, distinct from (and smaller than) the broader, largely pro-consensus superspreader ecosystem, with news media typically following rather than driving Twitter dynamics. Both papers thus insist that coordination and superspreading are structurally identifiable phenomena that must be disentangled from questions of truth or ideology before they can be evaluated normatively.
Ghezzi2023-8bebc91f complicates the misinformation frame yet further by examining a case—UK press coverage of GBD versus JSM signatories—where the “problematic” content is not fringe or fabricated but mainstream scientific disagreement selectively amplified along partisan lines. This paper functions as a useful boundary case for the whole topic: it shows polarized information “bubbles” (echoing the coordinated-cluster findings above) forming even among credentialed, non-fringe experts, warning against conflating political polarization of science coverage with the infodemic dynamics documented elsewhere.
Measuring the Ecosystem, Responding to the Narrative
Two papers step back from specific platforms or actors to propose ecosystem-level diagnostics and interventions. Scalco2026-bd operationalizes “information voids”—moments where public demand for vaccine information outstrips credible supply—and shows empirically that these voids coincide with drops in high-credibility content share and rises in misinformation, giving quantitative teeth to a previously qualitative concept and effectively supplying a supply-side account of why the exposure patterns documented by Adam, Lyons, and Efstratiou emerge when they do (e.g., around EMA authorizations or vaccine suspensions). Finally, Ramos2026-qo, grounded in a non-health but structurally analogous crisis-disinformation case (the Valencia floods), argues for moving beyond claim-level fact-checking toward narrative monitoring and argument-checking—an intervention-design response to the diagnosis, echoed throughout this collection, that isolated debunking cannot keep pace with adaptive, emotionally resonant narrative structures, whether those are conspiracy theories about hidden bodies or persistent vaccine-hesitancy tropes.
Synthesis
Across all twelve papers, health misinformation emerges not as a discrete category of false claims to be flagged and removed, but as an emergent property of information ecosystems: predisposition-driven selective exposure (Adam2026-tz, Lyons2026-ca), coordinated and organic amplification infrastructures (Giglietto2022-0e951ac5, Marino2023-9137f448, Song2025-yh, Efstratiou2026-ij), gatekeeping algorithms and legacy media remediation (Ducci2022-10cb5d70, Ghezzi2023-8bebc91f), supply-demand imbalances (Scalco2026-bd), and, ultimately, contested causal pathways to real-world behavior (Bollenbacher2026-vz, Allen2024-av). The collection’s implicit thesis is that veracity-centered interventions—fact-checking, labeling, source-credibility scoring—address only a thin slice of what actually shapes vaccine hesitancy and health-related belief, and that durable countermeasures must instead target narrative structures, exposure architectures, and the gatekeeping logics of both platforms and legacy media.