Information Disorder & Disinformation Narratives
From discrete falsehoods to circulating narratives
The intellectual spine of this collection is a sustained argument against treating “fake news” and “misinformation” as discrete, classifiable objects—false claims to be flagged, scored, and corrected. Foundational statements of this critique recur across the corpus: Giglietto2019-e9be81c1 dismantles creator-centric typologies of fake news, arguing via second-order cybernetics that truth judgments are observer-dependent and that attention must shift to propagators and the emergent, composite cascades their judgments produce. Marwick2026-qd pushes this further, reconceptualizing disinformation as cross-platform, identity-affirming, cathartic story—with characters, catchphrases, and moral stakes—that domain-level classification cannot capture. Sadler2025-vu supplies the theoretical vocabulary this move has lacked, borrowing Ricoeur’s hermeneutic realism and Phelan’s narrative ethics to show that content can be referentially and ethically suspect without being straightforwardly “false,” thereby troubling the true/false binary itself. Goel2025-iq, Waight2025-al, and Suau_Martinez2026-lv operationalize this narrative turn empirically: mainstream, factually accurate articles are shown to be repurposed into misleading narrative structures at scale, a phenomenon invisible to domain-blocklist and even exact-text-reuse methods, and repeated cross-outlet exposure to a narrative—not single false claims—predicts belief. Starbird2025-jj and Prochaska2025-ef extend the point to election-integrity rumors: the “misinformation problem” lives as much in the political framing applied to true evidence as in fabrication, and audiences collaboratively perform evolving “deep stories” whose meaning depends on how events are told, not merely what is claimed. Prochaska2026-pr formalizes this into a “grammar” of recurring Actor-Object-Action tropes spanning election cycles, while Marwick2025-vx shows an analogous processual logic in far-right “redpilling,” where disinformation functions as cumulative “evidence” sustaining a rationalist self-image rather than a single persuasive event. Frameworks papers—Frischlich2025-vn, Thiele2025-ol—argue that even our root metaphors (virus, warfare, pollution) and definitions (coordinated inauthentic behavior) are conceptually inadequate to this complexity, each illuminating one level of a multi-level system while obscuring others.
The hybrid media system as structuring condition
Underlying nearly every empirical study is Andrew Chadwick’s hybrid media system thesis (Chadwick2013-ns, Chadwick2015-ii), which frames contemporary disinformation not as a social-media-native pathology but as the product of interdependent old/new media logics and shifting “political information cycles.” This framework licenses the collection’s recurring finding that mainstream journalism is itself an engine of disinformation circulation. Tsfati2020-uo synthesizes why: direct exposure to fake-news sites is tiny and concentrated, so most belief in notable false stories is mediated by mainstream coverage attempting to debunk it—repetition that paradoxically increases perceived truth. Donovan2025-ws historicizes this as “misinformation-at-scale,” a structural feature of engagement-driven platform business models rather than a bug; Hurcombe2025-cs shows how Meta’s own “Newsroom” discourse discursively displaces responsibility onto “inauthentic” outside actors while erasing the role of hyperpartisan verified accounts and Facebook Groups. Rodarte2026-dk and Knupfer2025-vt trace how political elites and factions exploit this hybridity directly—parliamentarians waging epistemic-authority contests through derivative reporting and direct-authority claims, or factional Republicans building “connective” ties to hyper-partisan media ecosystems that reward ideological extremity with attention. Copland2025-em and Bosch2024-hj document platform-specific propaganda infrastructures (Sky News Australia’s YouTube/Facebook dual strategy; TikTok’s sonic affordances) feeding transnational right-wing and war-disinformation pipelines, while Bouchafra2026-ts shows a populist party deliberately tailoring securitizing visual rhetoric to each platform’s audience and affordances. Marino2023-9137f448 and Ducci2022-10cb5d70 extend the hybridity argument to health information, showing legacy media and algorithmic gatekeepers (Google News) remain central nodes even where “alternative influence networks” and IDW figures generate outsized engagement.
Measuring prevalence, concentration, and reach
A parallel empirical thread interrogates how much misinformation actually circulates and who is exposed. The foundational US studies—Allcott2017-yz, Grinberg2019-ua, Guess2020-rr, Allen2020-nj, Vosoughi2018-at—collectively established that fake-news consumption is a tiny, heavily concentrated fraction of the overall media diet, dominated by a small “supersharer” fringe, even as falsehood diffuses farther and faster than truth when it does spread. Gonzalez-Bailon2024-rq and Rossi2023-847d5a9f update this picture for Facebook’s 2020 election and the European newsfeed respectively, finding stable or declining exposure to untrustworthy content even as shares of untrustworthy URLs rise, with older cohorts disproportionately responsible for both exposure and sharing. Scalco2026-bd reframes the prevalence question structurally, operationalizing “information voids”—supply-demand imbalances that persist longer than overabundance and correlate with lower-credibility content—as a measurable ecosystem pathology rather than a simple volume metric. Hourigan2026-oc’s diary study complicates the entire prevalence debate by showing that lay publics’ perceived misinformation is overwhelmingly mainstream-sourced, text-based, and about everyday topics like cost-of-living, rarely matching what researchers flag as objectively false—a sociocultural-trust problem as much as a veracity problem. Humprecht2025-ml and Nenno2025-xa push against the WEIRD-centrism of this literature, the latter showing that news-value signatures of flagged misinformation diverge meaningfully (if modestly) across 24 countries, with the US and Brazil as outliers.
Partisan asymmetry and the “who spreads it” question
A persistent empirical question is whether misinformation sharing is symmetric across the political spectrum. Benkler2018-lw and Bennett2018-ys set the terms: disinformation in the contemporary US is asymmetric, rooted in a radicalized right-wing media ecosystem insulated from journalistic correction, itself a symptom of declining institutional trust rather than technology per se. Osmundsen2021-et grounds this in individual psychology, showing partisan animus (not ignorance) drives sharing and that Republicans share more fake news partly because of differential supply of politically useful out-party-derogating content. Tornberg2025-ir generalizes the claim cross-nationally across 26 countries’ parliamentarians, finding that neither populism nor right-wing ideology alone predicts misinformation sharing—only their conjunction in radical-right populism does. Renault2025-uh extends the asymmetry finding into the post-fact-checking era, showing Republican posts are flagged 2.3x more often even under X’s crowdsourced Community Notes, directly undercutting claims that professional fact-checking was ideologically biased. Tai2026-qk and Balluff2026-bv probe institutional and economic mechanisms—elected officials’ institutional roles as “checks,” and covert media-capture-via-advertising in Austria—that shape elite-level disinformation behavior. Iris2026-pg documents a structural, cross-national visibility bias favoring the Radical Right in 2024 EP election coverage regardless of outlet orientation, complicating radical-right “censorship” narratives, while Gattermann2025-yx shows far-right electoral performance itself shapes voter disinformation concern.
Belief, correction, and psychological mechanisms
A cognitive-psychological layer explains why corrections so often fail. Lewandowsky2012-vn remains the touchstone: retractions typically only halve reliance on misinformation, can backfire, and succeed only via specific techniques (alternative causal explanations, repeated retraction, pre-exposure warnings). Berlinski2023-rj applies this directly to electoral fraud claims, finding fact-checks fail to mitigate confidence damage among susceptible partisans—a finding echoed by Rossini2026-jn and Rossini2026-mj in the Brazilian context, where belief in electoral misinformation predicts both rising political intolerance and declining trust in results, even as mainstream news use bolsters resilience. Van_Erkel2026-mk extends hostile-media-effect theory to misinformation perception itself, showing partisans believe their own side is disproportionately targeted by misinformation—a bias amplified by interest, identity strength, and right-wing orientation. Adam2026-tz offers the most granular mechanistic account of conspiracy belief formation, tracing both “contagion” (alternative media) and “mitigation” (mainstream media) pathways alongside populism-driven selective engagement and counterarguing/backfire dynamics—directly operationalizing Zaller’s information-plus-predispositions model for the COVID moment. Mahl2026-hc’s Delphi study aggregates expert judgment across these micro/meso/macro levels, concluding “no silver bullet” exists and platform/journalism-level interventions outrank individual fixes.
Coordination, bots, and platform-level manipulation
A distinct empirical tradition examines how campaigns are organized and detected. Keller2019-nk provides rare ground-truth evidence (the NIS Twitter campaign) showing coordination signatures—not bot-like individual traits—best reveal astroturfing, while Thiele2025-ol builds a rational-choice typology for attributing covert principals from observable scale/elaborateness/disguise tradeoffs. Starbird2019-qv reframes disinformation as collaborative work, blurring orchestrated and organic participation across RU-IRA, White Helmets, and conspiracy cases. The Italian CLSB research program—Giglietto2022-0e951ac5, Marino2023-9137f448—and its extensions (Song2025-yh on vaccine discourse, Rodriguez_Farres2025-sg on Bluesky, Zhao2025-ny on hashtag hijacking) collectively show coordination is agnostic as to content and politics, serving both misinformation and legitimate public-health amplification depending on actor alignment. Efstratiou2026-ij traces a parallel contrarian-coordination dynamic in COVID science communication, finding credentialed anti-consensus experts disproportionately amplified by a small coordinated network, with news coverage following rather than leading superspreader activity. Bollenbacher2026-vz closes the causal loop for one platform-health link, estimating antivaccine tweets’ causal contribution to reduced vaccination and excess COVID mortality via a novel epidemic-plus-exposure model. Kim2026-wg documents geo-racially targeted voter-suppression advertising in 2016, showing average-treatment-effect designs mask severe, concentrated harms—a methodological lesson echoed by Graham2025-gp and Rogers2026-cy, which show propaganda exploits rather than evades platforms’ trust-infrastructure (#IStandWithPutin; Moldovan “post-truth spaces” bridging fringe and mainstream clusters).
War, geopolitics, and state propaganda
The Ukraine war and great-power information competition form a recurring case cluster. Waight2025-al’s biolabs-narrative tracing, Sadler2025-vu’s NATO-blame case study, Bosch2024-hj’s TikTok sound analysis, and Makeev2026-ma’s dual domestic/foreign propaganda model collectively show how state-aligned narratives are engineered for differentiated audiences and diffused asymmetrically into Western information ecosystems, often via “narrative laundering” through ostensibly independent or low-quality outlets rather than direct falsification. Waight2026-tz pushes this into the generative-AI era, demonstrating that state-coordinated Chinese media is memorized by commercial LLMs and causes measurably more pro-regime outputs when prompted in languages of low-media-freedom countries—recasting “media capture” as now extending into AI training pipelines. Hameleers2026-mc complicates assumptions about visual/AI disinformation’s unique power, finding video-based decontextualization (not AI-generated stills) outperforms text only for the already-salient, footage-rich Ukraine case—suggesting panic about generative AI may outrun the evidence relative to “low-tech” recontextualization.
Fact-checking, platform governance, and the regulatory turn
The collection closes on an increasingly institutional and adversarial note. Farkas2026-lr and Cazzamatta2026-lo document fact-checkers’ precarious rhetorical and material position as Meta dismantles its Third-Party Fact-Checking Program, showing content removal was always rare and censorship narratives largely unfounded, while Community Notes is viewed with professional skepticism. Ramos2026-qo proposes moving fact-checking practice itself toward “narrative monitoring” and “argument-checking,” operationalizing the paper’s broader narrative-turn argument for frontline newsroom use. Kuznetsova2026-hw shows platforms performing “governance-washing” around climate misinformation specifically, favoring soft interventions and discursively redistributing responsibility to users and outside experts. Lewandowsky2026-ob and de-Vreese2026-zx bookend the corpus with a structural diagnosis and a regulatory prescription: platforms bear responsibility for democratic backsliding, and the DSA’s Article 40 data-access provisions—however turbulent in implementation—represent the first serious constraint on platform informational power, a fight researchers (including this project) are urged not to abandon. DeVerna2025-dl and Altoe2026-wu point toward the emerging automation frontier for this fight, showing LLM fact-checking requires curated retrieval rather than raw capability, and that narrative-cluster-level, pre-peak virality triage can close the gap between misinformation’s ~1.5-day viral peak and fact-checkers’ multi-day response time—operationalizing, at the level of infrastructure, the same narrative-centric reorientation that organizes this entire line of inquiry.