¹ CIE and psychological drivers (Johnson and Seifert Reference Johnson and Seifert1994). The CIE describes how false information continues to influence reasoning even after it has been explicitly corrected—a robust laboratory finding across dozens of studies. See also Ecker et al. (Reference Ecker, Lewandowsky, Cook, Schmid, Fazio, Brashier and Kendeou2022). This comprehensive review synthesizes the main mechanisms that drive misinformation persistence, including memory-based, social, and identity-related factors. It provides the empirical context within which the EAF is designed to operate.

² Gap-filling and relevance theory (Grice Reference Grice, Cole and Morgan1975). Grice’s cooperative principle holds that listeners assume speakers are contributing information that is relevant and well evidenced; in conditions of uncertainty, people extend this assumption to any source that fills the informational gap first, regardless of its accuracy. See also Sperber and Wilson (Reference Sperber and Wilson1995). The relevance framework formalizes why minds reach for the most cognitively accessible explanation rather than the most evidentially sound one—a key mechanism behind Window dynamics.

³ On corrections failing and the backfire question (Nyhan and Reifler Reference Nyhan and Reifler2010). The foundational study documenting that corrections can sometimes increase rather than reduce belief in misinformation—originally called the “backfire effect.” See also Wood and Porter (Reference Wood and Porter2019). A large replication effort that failed to reproduce the original backfire results, concluding that persistence (continued holding of a belief despite correction) is far more common than amplification (increased confidence after correction). Both findings are discussed in Part Three; the EAF framework accommodates both: it predicts persistence as the baseline outcome and amplification only when correction attempts are experienced as identity attacks.

⁴ False news spreads faster than true news (Vosoughi, Roy, and Aral Reference Vosoughi, Roy and Aral2018). Analysis of approximately 126,000 news stories shared on Twitter over eleven years, finding that false news spreads faster, farther, deeper, and more broadly than true news across all categories of information. The structural implication for the EAF is that false claims are systematically more likely to arrive within the Window than corrections are.

⁵ Institutional trust and belief updating (Zaller Reference Zaller1992). Zaller’s receive–accept–sample model demonstrates that citizens who are more politically aware and trust mainstream media are more likely to accept information from elite sources—extending to institutional credibility more broadly. See also Hallin and Mancini (Reference Hallin and Mancini2004). Cross-national analysis of how media system structure affects the degree to which citizens defer to institutional information sources.

⁶ Formal definition of the EAF. The EAF is a sequential, self-reinforcing cognitive and social mechanism through which a false belief enters the mind during a period of informational uncertainty (the Window), acquires legitimacy through repeated exposure from trusted sources (the Gate), and becomes resistant to revision because changing it would carry unacceptable social costs tied to personal identity (the Cost). The three components interact: each cycle through the system tightens the filter, making subsequent correction progressively harder. The theory predicts that successful belief correction requires intervening at the earliest stage possible, and that interventions addressed only to factual content will fail once the identity-cost threshold has been crossed. Scope and boundary conditions of the model. The EAF is primarily calibrated to explain beliefs that are socially embedded and identity-relevant—claims that people have argued for publicly, that their communities reinforce, and that carry group-membership stakes. It is less likely to apply in full to private beliefs held without community reinforcement, beliefs on topics carrying no identity stakes, or highly isolated contexts where the social cost of revision is absent. The model’s explanatory power is strongest where social embeddedness is high—the everyday cases of contested public belief that communicators and policymakers most often face. For the full spectrum discussion including clinical boundaries, see footnote Footnote 7.

⁷ The Window and the belief plasticity spectrum. Understanding where the Window sits on a spectrum of belief plasticity—from the ordinarily correctable to the clinically immovable—clarifies what communicators face in practice. Psychologists draw a careful distinction between ordinary motivated reasoning, which almost everyone does to some degree, and the rigidly held, evidence-resistant beliefs that appear in clinical settings. Research on delusional thinking has found that people who already hold one such belief are significantly more likely to acquire additional ones. Research on authoritarian personality structures—see Altemeyer (Reference Altemeyer1996)—has found that people whose identities are organized around loyalty to a hierarchical authority figure tend to show stronger resistance to belief revision, particularly when their core loyalty is challenged. This framework does not require clinical language, and applying it carelessly to political disagreement would be a mistake. But the parallel is useful: both delusional persistence and ordinary motivated reasoning operate through the same basic mechanism—just at very different intensities.

⁸ Heuristic processing, cognitive shortcuts, and fast information (Chaiken Reference Chaiken1980). The classic formulation of dual-process persuasion, distinguishing heuristic (shortcut-based) from systematic (evidence-based) processing. See also Kahneman (Reference Kahneman2011). Kahneman’s System 1/System 2 framework provides the cognitive foundation for understanding why minds default to the fastest available explanation under conditions of informational pressure—the mechanism driving Window dynamics in real-world contexts.

⁹ Source credibility, institutional trust, and the open window (Hovland, Janis, and Kelley Reference Hovland, Janis and Kelley1953). The foundational Yale studies on persuasion, establishing that credibility of the communicator is a primary determinant of attitude change—directly relevant to why trust levels modulate how long people hold their interpretive window open.

¹⁰ Legitimacy shortcuts and cognitive effort reduction. The shift from systematic to heuristic processing (see footnote 8) is the core Gate mechanism: as trust in a source increases and exposure accumulates, the cognitive cost of critical evaluation is replaced by automatic acceptance. This is not laziness—it is an adaptive economy of mental resources that becomes maladaptive when exploited by misinformation ecosystems.

¹¹ Source credibility and message repetition in belief formation (Hovland, Janis, and Kelley Reference Hovland, Janis and Kelley1953). See footnote 9. The mid-twentieth-century persuasion research identified both source credibility and message repetition as independently sufficient to lower acceptance thresholds—their combination, which is precisely what social-media echo chambers produce, is therefore especially powerful.

¹² Echo chambers, algorithmic reinforcement, and misinformation spread (Del Vicario et al. Reference Del Vicario, Bessi, Zollo, Petroni, Scala, Caldarelli and Stanle2016). Large-scale analysis of Facebook information diffusion showing that users preferentially share content within like-minded communities, creating echo chambers that provide exactly the conditions of repeated trusted-source exposure that lower Gate thresholds.

¹³ Gate erosion: source-specific to domain-general dynamics. Erosion is primarily source-specific in its early stages—repeated exposure to claims from a single trusted community lowers scrutiny for future claims from that community, without immediately generalizing to all sources on the same topic. Over time, however, the erosion can become domain-general, affecting how the person evaluates any claim bearing on the same subject regardless of its origin. The transition has an operationally visible signal: a person in the early stage can still produce objections to a nonpreferred source; a person who has crossed into domain-general erosion can no longer produce any counter-argument on the affected topic even when explicitly prompted. A fully eroded gate can also re-open a closed window for related new misinformation, bypassing the provisional uncertainty that normally characterizes the Window stage—making affected communities structurally more vulnerable to subsequent misinformation across the whole domain.

¹⁴ Social attitudes, public commitment, and belief entrenchment (Visser and Mirabile Reference Visser and Mirabile2004). Demonstrates that attitude strength—resistance to persuasion, persistence over time—increases with the degree of social reinforcement a belief receives from network contacts, providing the empirical grounding for the Cost mechanism.

¹⁵ Motivated reasoning and the social cost of public revision (Kahan Reference Kahan2013). Demonstrates experimentally that motivated reasoning—reasoning designed to protect a prior conclusion rather than find the truth—is not reduced by greater cognitive reflection ability; if anything, it is amplified. The social cost of public revision is not merely psychological discomfort but structural: the expected value of changing one’s stated position in a group that has publicly reinforced it is negative.

¹⁶ Identity-Protective Cognition: definition and evidence (Flores Reference Flores2025). A recent philosophical examination of identity-protective reasoning, arguing that it has a limited epistemic justification when group membership constitutes a genuine source of evidence—but noting that this justification breaks down when beliefs are about empirical matters where the group has no privileged access to evidence. Provides the definitional basis for Identity-Protective Cognition as used in this article. Cultural cognition, group identity, and belief formation (Kahan et al. Reference Kahan, Braman, Gastil, Slovic and Mertz2007). Empirical demonstration that individuals’ risk perceptions reliably track their cultural identity (hierarchical/individualistic versus egalitarian/communitarian worldviews) rather than objective evidence, providing the foundational experimental evidence base for identity-protective cognition theory.

¹⁷ Central versus surface beliefs: the case of charismatic political leaders. The case of charismatic political leaders illustrates this point sharply. Followers of such leaders often do not hold a fixed set of factual beliefs—they hold one core belief: loyalty to the leader. When the leader changes position, followers shift with them, sometimes overnight. What looks like inconsistency is actually consistency at a deeper level: the real central belief—loyalty—has not moved. Surface positions are flexible; the underlying loyalty is not. For communicators, this is a critical insight: you must identify which beliefs in a person’s network are truly central before deciding where to intervene. Correcting a surface claim while the deeper belief stays intact will accomplish very little.

¹⁸ Misinformation correction, persistence, and the role of repetition (Lewandowsky et al. Reference Lewandowsky, Ecker, Seifert, Schwarz and Cook2012). A comprehensive review of the conditions under which corrections succeed or fail, establishing that each failed correction attempt can function as an affirmation of the original belief. Key for understanding why the EAF predicts that late-stage correction can make the problem structurally worse, not better. Boomerang and backfire effects in science communication (Hart and Nisbet Reference Hart and Nisbet2012). Experimental evidence that science communications framed in ways that trigger identity-protective responses can increase polarization rather than reduce it—the “boomerang effect”—consistent with the EAF prediction that high-Cost-stage corrections can be counterproductive.

¹⁹ Sequential logic: why the order Window → Gate → Cost is not arbitrary. The Window is always first because it describes the initial moment of informational uncertainty—before any repeated exposure has occurred, before any belief has been formed. The Gate follows because establishing a belief through repeated use requires time and repetition that cannot precede first exposure. The Cost comes last because identity-embedding is a downstream consequence of social commitment. This ordering does not mean the components cannot exert backward influence: high identity stakes can pre-emptively narrow the Window for future claims, and an eroded Gate accelerates Cost entrenchment. Practically, interventions should be applied with the causal grain: prebunking addresses the Window; source literacy education addresses the Gate; identity-cost reduction addresses the Cost.

²⁰ “Smart idiots” and motivated numeracy (Kahan et al. Reference Kahan, Peters, Dawson and Slovic2017). Experimental study showing that greater numeracy increases rather than reduces politically motivated reasoning—more numerate individuals are better at deploying their statistical skills in the service of identity-protective conclusions rather than accurate analysis. This is the experimental basis for the “smart idiots” formulation used in the main text.

²¹ Measuring the Window component. The Window’s influence can be tracked through timing: do people who received accurate information before the window closed update their beliefs at higher rates than those who received it afterward? This requires measuring both the moment of first exposure to a claim and the moment of first correction, then tracking belief acquisition longitudinally. Tay et al. (Reference Tay, Hurlstone, Kurz and Ecker2022)—see footnote 22—provides the closest existing experimental analogue. Measuring the Gate component. The Gate’s effect can be tracked through exposure patterns: do people with higher cumulative exposure to a claim from a single trusted source show lower levels of critical scrutiny on structurally similar claims from the same source? Survey-based instruments measuring perceived source credibility and endorsement rates can approximate Gate threshold levels; natural experiments in media ecosystem changes offer a complementary approach. Measuring the Cost component. The Cost can be operationalized through established measures of identity-protective thinking: belief centrality scales, partisan identity strength measures, and implicit association tests linking self-concept to target beliefs. See footnote 16 (Flores Reference Flores2025; Kahan et al. Reference Kahan, Braman, Gastil, Slovic and Mertz2007) for the relevant measurement traditions. The key prediction is that Cost level moderates whether even well-timed corrections (Window-stage interventions) result in durable belief change.

²² Prebunking versus debunking: empirical evidence and research design implications (Tay et al. Reference Tay, Hurlstone, Kurz and Ecker2022). This study offers the closest current empirical analogue for the Window component, demonstrating that pre-exposure correction consistently outperforms post-exposure correction—a finding directly predicted by the EAF’s timing logic. Testing the full sequential interaction of Window × Gate × Cost requires longitudinal or multiwave designs that can track belief acquisition, source exposure patterns, and identity investment across the same individuals over time. No single published study has yet measured all three factors simultaneously.

²³ Intervention 1—Timing and prebunking. The most powerful intervention point is the window. Public communicators, health authorities, and journalists who can get accurate, accessible information into that window—before false explanations fill the gap—will succeed far more often than those who issue corrections after the fact. This is the logic behind prebunking: warning people in advance about the kinds of false claims likely to appear, before they encounter them. See Cook, Lewandowsky, and Ecker Reference Cook, Lewandowsky and Ecker2017). See also Lewandowsky and van der Linden (Reference Lewandowsky and van der Linden2021). Intervention 2—Source trust. A correction delivered by a source with low legitimacy in the target community will fail regardless of how well evidenced it is. This is not a failure of the audience—it is a structural feature of how the Gate operates. Communicators must invest in source credibility before they need it. Corrections issued by trusted community voices, peer networks, or figures who share the audience’s identity have a structural advantage over technically accurate corrections from distrusted institutions. See footnote 9 (Hovland, Janis, and Kelley Reference Hovland, Janis and Kelley1953) for the foundational evidence base. Intervention 3—Social safety and identity cost reduction. When the cost of changing your mind is perceived as a betrayal of community, no amount of evidence will be sufficient. Effective correction strategies must address the social cost directly—by creating contexts in which it is acceptable, even admirable, to change your position. Research on affective polarization—see Reiljan (Reference Reiljan2020)—shows that strong partisan identities raise the cost of cross-party belief revision to near-prohibitive levels. Reducing that cost often requires identity work that has nothing to do with evidence. Intervention 4—Entrenched beliefs and the limits of late-stage correction. Research on the persistence of corrected misinformation—see Thorson (Reference Thorson2016)—shows that even successful corrections often leave residual traces (“belief echoes”) that continue to influence inference. The practical implication is that prevention is dramatically more efficient than correction. Investing in pre-event source literacy education, in transparent and timely communication during uncertainty, and in reducing the social cost of updating beliefs before misinformation arrives, is one of the most durable investments a communicator, educator, or institution can make.

²⁴ Cross-reference to key empirical streams. For a comprehensive overview of the individual streams of research that this framework draws together, see: footnote 1 (Continued Influence Effect), footnote 5 (institutional trust and belief updating), footnote 6 (EAF formal definition), footnote 7 (belief plasticity spectrum), footnote 10 (legitimacy shortcuts), footnote 11 (source credibility and repetition), and footnote 12 (echo chambers). These constitute the primary empirical foundations of the framework.

²⁵ Structural explanation and moral responsibility. Structural explanation is not the same as moral exculpation. The fact that a mental filter makes false beliefs more likely does not mean individuals bear no responsibility for what they believe. History is full of cases where the structure made harmful beliefs easy to hold and costly to revise—and where that fact still did not absolve the people who held them. A doctor who holds demonstrably false beliefs about a treatment, a judge who reasons from discredited evidence, a researcher who accepts claims without scrutiny—the mental filter may explain how they got there, but it does not excuse the harm that follows. In these domains in particular, the structural account is not a reason to lower our expectations. It is a reason to raise the quality of our institutions, our training, and our habits of mind.