Platforms attach warning labels to misinformation, and some labels cite an external assessing entity such as a fact-checking organization, news outlet, or university research center. Habib, Elsharawy and Rahman (arXiv:2410.00866, 2024) showed with 655 US survey participants that trust in the cited entity varies substantially and conditions the label's expected effectiveness, and they explicitly left open how such entity-referenced labels behave beyond their original setting. No study to date has measured whether stated sharing reductions translate into actual sharing behavior, has replicated the entity-perception pattern outside the US survey context, has tested label wording that prevents the label from being misread as an endorsement of the content, or has tested multi-entity corroboration or personalization of the cited entity on behavioral or perception outcomes. The problem is to settle these five linked external-validity and design dimensions of the same label-design object — the entity-referenced warning label.

Background

Online misinformation — false or misleading content spread through social platforms — has prompted platforms to attach warning labels to questionable posts. Some of these labels cite an external entity that assessed the content's accuracy, for example professional fact-checking organizations (many of them verified by the International Fact-Checking Network, IFCN, a body that sets standards for fact-checkers), news outlets, government agencies, or university research centers. Whether such an 'entity-referenced' label deters sharing plausibly depends on how much the user trusts the named entity.

Habib, Elsharawy and Rahman (2024) studied exactly this. In an online survey experiment with 655 US participants recruited through Prolific (a crowdsourced-participant platform), they measured perceptions of eight categories of fact-checking entities across two misinformation topics. They found that self-reported trust differed markedly across entity categories and across participants' demographic and political characteristics, that some participants reported they would be less deterred by a label citing an entity they distrust, that less familiar entities (such as regional IFCN fact-checkers) were expected to be less effective, and that a small share of participants misread the label — interpreting the citation of an entity as an implicit endorsement of the flagged content rather than a warning. All outcome measures were self-reported perceptions and stated sharing intentions.

Adjacent literature clarifies what is and is not settled. A large meta-analysis of 21 survey experiments (N = 14,133) by Martel and Rand (Nature Human Behaviour, 2024) found that fact-checker warning labels reduce belief in false headlines by about 27.6% and self-reported willingness to share them by about 24.7%, with sizable reductions even among participants who most strongly distrust fact-checkers — so distrust does not appear to eliminate label effectiveness on stated outcomes. Qi, Wang, Liu and Metzger (Journal of Media Psychology, 2025) compared credibility perceptions of different fact-checking source types (professional, news, platform, automated, crowdsourced) in US samples, again with self-report measures and no label-design manipulation. Lukumon, Byrd and Strickland (2024) found that US partisans prefer and trust bipartisan fact-checking groups, supporting the perception-side case for multi-entity labels without deploying one. Aruguete, Calvo and Ventura (Research & Politics, 2025) showed in Argentina that fact-checks improve the fact-checker's reputation but shift its perceived ideology. Studies of label design and source attribution more broadly (Tang et al. 2025; Lim and Perrault 2024; King, Phillips and Carley 2025) vary label format and source but do not test the specific questions above on behavioral outcomes. Finally, the deployment context shifted in January 2025 when Meta ended its US third-party fact-checking program in favor of Community Notes — a system, pioneered on X (formerly Twitter), in which volunteer raters rather than professional fact-checkers write and rate short notes attached to disputed posts. This shift is documented by Horne and Craig (2025) and by Borenstein, Warren, Elliott and Augenstein (2025), who note that Community Notes themselves rely on professional fact-checking sources. The shift makes large-scale platform tests of entity-referenced labels harder, but it does not answer the scientific questions above.

Problem Statement

The source paper establishes, in a US survey sample with self-reported outcomes, that user trust in the entity cited by a misinformation warning label conditions the label's expected deterrent effect. The problem is to settle, with evidence a solver can produce from this package, the five linked external-validity and design questions arising from the paper's stated limitations and recommendations, all concerning the same design object — the entity-referenced warning label:

  1. Generalization: do the entity-perception patterns (differences in trust and expected effectiveness across entity categories and user groups) replicate beyond the US, Prolific-recruited sample — for example, in other countries and cultural contexts, other age cohorts, or other online communities?
  2. Behavioral efficacy: do the observed reductions in stated sharing intention translate into reduced actual sharing of flagged content, rather than only self-reported intentions?
  3. Wording and format: how do alternative phrasings and presentations of the entity citation — in particular, wording that makes explicit what the named entity concluded — affect user comprehension and persuasion, and can they reduce or eliminate the misinterpretation in which users read the label's citation of an entity as an implicit endorsement of the flagged content?
  4. Multi-entity citation: does citing multiple entities that reached the same conclusion — including reliable entities with perceived biases on different sides of the political spectrum — reduce perceived label bias?
  5. Personalization: does customizing which entity is cited to better resonate with a particular user's trust profile (for example along age or political preference) improve label effectiveness?

The five questions are the population, outcome, wording, multiplicity, and selection dimensions of one overarching objective: evidence-based design and external validation of entity-referenced misinformation labels. They are jointly posed: settling a single dimension resolves neither the remaining questions nor the overarching objective.

References

  1. Hana Habib, Sara Elsharawy, Rifat Rahman. "I don't trust them": Exploring Perceptions of Fact-checking Entities for Flagging Online Misinformation. arXiv:2410.00866, 2024. DOI: 10.48550/arxiv.2410.00866. https://arxiv.org/abs/2410.00866
  2. Cameron Martel, David G. Rand. Fact-checker warning labels are effective even for those who distrust fact-checkers. Nature Human Behaviour, 2024. DOI: 10.1038/s41562-024-01973-x. https://doi.org/10.1038/s41562-024-01973-x
  3. Li Qi, Laurent H. Wang, Xingyu Liu, Miriam J. Metzger. Facts About Fact-Checkers: Comparing Credibility Perceptions, Usage, and Sharing of Different Fact-Checking Sources. Journal of Media Psychology, 2025. DOI: 10.1027/1864-1105/a000484. https://doi.org/10.1027/1864-1105/a000484
  4. Gafari Abiodun Lukumon, Nick Byrd, Brent Strickland. Adversarial collaboration among fact-checkers: bipartisanship outweighs partisanship in preferences and trust towards news. PsyArXiv preprint, 2024. DOI: 10.31219/osf.io/gp9w7. https://doi.org/10.31219/osf.io/gp9w7
  5. Natalia Aruguete, Ernesto Calvo, Tiago Ventura. The fact-checking dilemma: Fact-checking increases the reputation of the fact-checker but creates perceptions of ideological bias. Research & Politics, 2025. DOI: 10.1177/20531680251323120. https://doi.org/10.1177/20531680251323120
  6. Huiyun Tang, Björn Rohles, Yuwei Chuai, Gabriele Lenzini, Anastasia Sergeeva. More Than Just Warnings: Exploring the Ways of Communicating Credibility Assessment on Social Media. arXiv:2503.10445, 2025. https://arxiv.org/abs/2503.10445
  7. Gionnieve Lim, Simon T. Perrault. Effects of Automated Misinformation Warning Labels on the Intents to Like, Comment and Share Posts. arXiv:2403.12916, 2024. https://arxiv.org/abs/2403.12916
  8. Catherine King, Samantha C. Phillips, Kathleen M. Carley. Public support for misinformation interventions depends on perceived fairness, effectiveness, and intrusiveness. arXiv:2508.05849, 2025. https://arxiv.org/abs/2508.05849
  9. Benjamin Horne, Matthew J.A. Craig. Despite Meta Ending Its Third-Party Fact-Checking Program, Most People Still Want Fact-Checkers on Social Media. 2025. DOI: 10.7290/iii037fjd. https://doi.org/10.7290/iii037fjd
  10. Nadav Borenstein, Greta Warren, Desmond Elliott, Isabelle Augenstein. Can Community Notes Replace Professional Fact-Checkers? arXiv:2502.14132, 2025. https://arxiv.org/abs/2502.14132