Pith. sign in

REVIEW 4 major objections 5 minor 300 references

Ideologically configured LLM fact-checkers shift trust in true and false headlines even when they clash with the reader's politics; wrong or inconclusive verdicts shift trust too.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 23:33 UTC pith:JV4BH5DV

load-bearing objection Solid two-experiment demonstration that LLM fact-checkers shift trust regardless of partisan alignment, but the moderation claim rests on a mis-scaled variable and a post-treatment measure. the 4 major comments →

arxiv 2607.15364 v1 pith:JV4BH5DV submitted 2026-07-16 cs.CY

On the Effectiveness of Fact Checking Information from Politically Congruent and Incongruent Large Language Models

Jiangen He , Benjamin D Horne , Dorit Nevo This is my paper
classification cs.CY
keywords LLM fact-checkingpolitical congruencytrust in newsmisinformation correctionideologically configured chatbotswithin-subjects experimenttainted truth effectpartisanship
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tries to establish that fact-checking keeps working when the fact-checker is an ideologically configured LLM chatbot of the kind now being embedded on social platforms. In two repeated-measures experiments with 705 U.S. adults, correct chatbot verdicts raised trust in true headlines by roughly 0.9–1.1 points and lowered trust in false headlines by roughly 0.8 points on a 5-point scale, and this held regardless of whether the bot's politics matched the reader's. The paper argues that perceived political congruency plays only a narrow role: for true headlines that are politically distant from the reader, trust rose less when the bot was seen as far away and more when it was seen as moderate; no such pattern appeared for false headlines. The paper also reports that wrong or 'Unverifiable' verdicts moved trust, implying that LLM fact-checkers can both correct misinformation at scale and damage trust in true information at scale. A sympathetic reader would care because human fact-checking is being replaced by exactly these systems.

Core claim

On the paper's own terms, the central discovery is that people update their trust in news in response to an LLM fact-checker's verdict even when the bot is politically incongruent with them. Correctly labeled true headlines gained about 0.93–1.09 points of trust; correctly labeled false headlines lost about 0.78–0.85 points. The only consistent moderation was in Study 1: for true headlines that were politically distant, trust increased less when participants perceived the bot as distant and more when they perceived it as moderate; this interaction did not appear for false headlines. The paper further finds that incorrect verdicts moved trust toward the verdict (true headlines falsely called

What carries the argument

The argument is carried by a real-time conversational fact-checking bot whose political slant is manipulated in two ways: it is restricted to left- or right-leaning news domains and given an ideological persona, while still being required to open each reply with a True/False/Unverifiable verdict and cite sources. The quantitative machinery is a statistical model with random intercepts for participants that predicts trust change from the bot's verdict, the participant's perceived distance from the bot, the headline's distance from the participant, and their interaction. The distance variables are computed on a 7-point ideology scale, with headline ideology rescaled from 5 points; the interact

Load-bearing premise

The paper's key interaction finding assumes that political ideology can be captured as a single left-right number for each person and each headline, and that participants' perception of the bot's ideology, measured after they updated their trust, can be treated as an independent cause of that update.

What would settle it

Re-estimate the Study 1 interaction with the corrected headline rescaling (6/4)(HPL−1)+1 instead of the printed (6/4)HPL+1, or collect perceived bot distance before the trust rating; if the distance-by-bot interaction (β ≈ −0.051, p<0.01) becomes non-significant or reverses, the claim that congruency matters only for distant true headlines fails.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • Correct LLM verdicts can push trust in true news up and false news down by roughly one point on a 5-point scale, even for users whose politics oppose the bot's.
  • Because wrong and inconclusive verdicts also move trust, the social value of LLM fact-checking hinges on the bot's accuracy; a modest error rate repeated at platform scale can taint true information.
  • Perceived political bias is not, by itself, a strong barrier to correction, so platforms gain less than they might hope from tailoring bot ideology to users — except for politically distant true headlines.
  • Whoever configures the bot (its sources and persona) holds a lever over public trust, since users seem to accept verdicts regardless of perceived slant.
  • The results extend the existing finding that fact-check labels survive partisanship to the new setting of conversational, ideologically configured AI.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the mechanism is a 'machine heuristic' — people treating AI output as objective — then explicitly telling users the bot can err, or showing its error rate, should shrink the trust shift; the paper does not test this moderation.
  • The one-sided interaction (moderate bots beat distant bots only for distant true headlines) suggests a credibility threshold rather than a smooth distance effect; a replication varying bot persona extremity could separate the two.
  • In real deployments users may self-select into congruent bots, as in Study 2's choice condition; the paper's data suggest this does not weaken correction, but repeated self-selected echo-chamber use could affect long-run trust calibration differently.
  • Comparing the effect sizes for correct versus incorrect verdicts in this study gives a rough break-even accuracy for net benefit; a platform bot with accuracy below that threshold could do more harm than good even if it is usually right.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper reports two within-subjects experiments (total n=705) in which U.S. participants rated trust in true and false political headlines, then interacted with real-time LLM chatbots configured to lean left or right, and rated trust again. The central claims are that fact-checking by ideologically configured LLMs shifts trust in both true and false headlines; that this holds even when the bot is politically incongruent with the user; that incorrect or inconclusive bot verdicts also shift trust; and that perceived political congruency moderates trust only for politically distant true headlines, mainly via a Study 1 interaction between perceived bot distance and headline distance. The paper also reports a second study in which participants chose their bot.

Significance. If the broad findings hold, the paper makes a useful and timely contribution: it moves beyond static warning labels to real conversational LLM fact-checkers, uses recent and balanced political headlines, reports high bot correctness rates, and provides open materials via OSF. The main results about correct verdicts shifting trust—roughly +0.93/+1.09 for true headlines and -0.78/-0.85 for false headlines—are well supported by both studies, as is the concerning result that incorrect and inconclusive verdicts also move trust. However, the paper's most distinctive moderation claim is much more fragile: it rests on one interaction in Study 1, on a post-treatment perceived-distance measure, and on a headline-distance variable whose printed rescaling appears algebraically incorrect. These issues do not undermine the broad effectiveness result, but they do mean the central 'congruency matters only for distant headlines' claim requires substantial additional analysis or careful re-framing.

major comments (4)
  1. [Methods, EQ2] The rescaling of headline political leaning (HPL) from the 5-point pretest scale to the 7-point participant scale appears incorrect. The printed formula is |PL − (6/4)HPL + 1|. If HPL ranges 1–5 and PL 1–7, the affine rescaling should be (6/4)(HPL−1)+1 = 1.5·HPL − 0.5. The printed 1.5·HPL + 1 maps HPL to {2.5, 4, 5.5, 7, 8.5}, shifting every transformed value by +1.5 and exceeding the 7-point scale. Since Headline≠ enters the interaction in EQ3, this can change which headlines count as 'distant' and can create or mask the reported interaction β = -0.051 (p<0.01, Fig 4a). No robustness check with the corrected transformation or with HPL on its original scale is reported. This is load-bearing for the paper's most distinctive claim.
  2. [Methods, EQ1 and Results, Fig 4] Bot≠ is computed from perceived bot political leaning (PBL), measured after the trust-update task. It is therefore a post-treatment variable and may be endogenous to the outcome: participants who updated trust may rationalize their perception of the bot's leaning. This concern is heightened by the paper's own manipulation-check results, where the randomized designed-congruency effect is negligible (Cliff's δ = 0.11) while the regression relies on perceived congruency. The authors should report models using assigned/designed congruency, or a pre-treatment measure of perceived bot leaning, to establish that the moderation is not a post-hoc rationalization artifact.
  3. [Abstract and Results, Fig 4a vs Fig 4c] The claim that 'the perceived political congruency between the participant and the bot matters only when headlines are politically distant' is based on Study 1 alone. In Study 2, the Bot≠ × Headline≠ interaction is not significant (Fig 4c). The paper attributes this to self-selection of congruent bots, but provides no statistical support for that explanation. The abstract and conclusion should be qualified to say this pattern was found in one study, or supplemented with a pooled analysis or a choice-robustness check.
  4. [Results, Fig 5] The post-hoc tertile split of the interaction and the single significant KS contrast (KS = 0.235, p < 0.001, Cliff's δ = 0.27) are used to describe the moderation as one-sided. No correction for multiple comparisons is reported across the many KS tests in this exploratory breakdown, and the effect sizes are small. This descriptive evidence should be labeled exploratory rather than presented as strong confirmation of the interaction.
minor comments (5)
  1. [Discussion] Typo: 'chabots' should be 'chatbots'.
  2. [Limitations] Typo: 'wrong verdicts form LLMs' should be 'wrong verdicts from LLMs'.
  3. [Results, Unverifiable verdicts] The text says 'in general, inconclusive answers decreased trust in headlines, no matter their veracity,' but Study 2 shows trust in true headlines increased by +0.19 after an 'Unverifiable' verdict. This inconsistency should be acknowledged or the summary sentence adjusted.
  4. [Methods, Study 1 sample] The description of the pilot participants and how their responses map to the final n=412 is unclear. Please clarify exactly how many pilot participants' responses were retained and how they combine with the representative sample.
  5. [Results, Fig 5 caption] The caption uses 'moderate headlines' where 'politically close headlines' appears to be intended. The wording makes the figure harder to interpret.

Circularity Check

0 steps flagged

No circular derivation; only minor non-load-bearing self-citations.

full rationale

This is a behavioral intervention study, not a claimed derivation. The central estimates - trust increases of about 0.93-1.09 on correctly labeled true headlines, decreases of about 0.78-0.85 on correctly labeled false headlines, and the Study 1 interaction Bot != x Headline != (beta = -0.051, p < 0.01) - are estimated directly from pre/post trust ratings in randomized within-subjects designs; no parameter is fitted to the outcome and then reported as a prediction. EQ3 is a regression model, not a derivation, and its predictors (verdict, perceived bot distance, headline distance) are constructed from separate rating tasks. The EQ2 5-to-7 rescaling issue and the post-treatment measurement of perceived bot leaning are validity/endogeneity concerns, not cases where the result equals an input by construction: Trust_Delta remains an independently measured outcome. The manuscript cites prior work by the same authors (e.g., Horne et al. 2019; Horne 2025; Horne and Nevo 2025) for background findings on warning labels and headline selection, but the LLM fact-checking claims are tested with new experimental data and do not rest on those citations as evidence. Thus there is no load-bearing circular step; the score reflects only minor, non-load-bearing self-citation.

Axiom & Free-Parameter Ledger

1 free parameters · 4 axioms · 0 invented entities

The paper's claims do not introduce new theoretical entities. The loaded assumptions are measurement-level: ground-truth veracity, unidimensional interval ideology, exogenous post-treatment perception, and sample representativeness. The only hand-chosen number affecting the key interaction is the 5-to-7 rescaling constant in EQ2.

free parameters (1)
  • Headline leaning rescale constants = (6/4, +1) as printed
    Chosen by hand to map 5-point pretest favorability to the 7-point participant-leaning scale in EQ2. The printed formula maps HPL=5 to 8.5, exceeding the stated 7-point range, and this affects the headline-distance variable used in the key interaction.
axioms (4)
  • domain assumption Ground-truth veracity labels are correct.
    True/false status comes from fact-checking sites and reputable outlets; if any labels are wrong, the 'correct' and 'incorrect' verdict categories are mislabeled.
  • domain assumption Political ideology is unidimensional and measured on commensurate interval scales for participants, bots, and headlines.
    EQ1 and EQ2 take absolute differences between 7-point self-reports, 7-point perceived bot leaning, and a rescaled 5-point headline rating, implying equal intervals and comparability across sources.
  • domain assumption Post-interaction perceived bot distance is exogenous to trust change.
    The main regressions use perceived bot distance measured after the chatbot interaction as a predictor of trust change; if the interaction influenced that perception, estimates are biased.
  • domain assumption Prolific samples with attention checks are representative enough for U.S. political comparisons.
    Samples are described as representative across age, gender, and political affiliation, but they are opt-in panels rather than probability samples.

pith-pipeline@v1.3.0-alltime-deepseek · 17575 in / 11237 out tokens · 109697 ms · 2026-08-01T23:33:01.372558+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of On the Effectiveness of Fact Checking Information from Politically Congruent and Incongruent Large Language Models." pith.science (2026). https://pith.science/paper/JV4BH5DV

@misc{pith2026260715364,
  author       = {Pith},
  title        = {Pith review of: On the Effectiveness of Fact Checking Information from Politically Congruent and Incongruent Large Language Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JV4BH5DV}},
  note         = {Machine review of arXiv:2607.15364}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Social media companies have shifted away from human fact-checkers and instead have embedded conversational Large Language Models (LLM) on their platforms. LLM chatbots differ from human fact-checkers in many ways that may shape user responses to corrections. Of particular interest in this study is that LLM chatbots can be ideologically configured via the content emphasized in their responses, the sources cited, and the configured persona. Using data from two within-subjects experiments (n=705), this paper investigates the effectiveness of fact checking information from ideologically configured LLM chatbots. We find that LLM fact-checkers significantly shift trust in true and false political news headlines, even when the chatbot is politically incongruent with the user. The perceived political congruency between the participant and the bot matters only when headlines are politically distant. That is, trust in correctly labeled true headlines increases less when politically distant chatbots check distant headlines and increases more when moderate chatbots check distant headlines. The perceived political congruency of LLM chatbots did not impact their effectiveness at decreasing trust in false headlines. Unfortunately, LLM fact-checkers also significantly change trust in news when they are wrong or provide inconclusive answers. Our results demonstrate both the potential for LLMs to correct false information at scale but also their potential to taint the truth at scale.

Figures

Figures reproduced from arXiv: 2607.15364 by Benjamin D Horne, Dorit Nevo, Jiangen He.

Figure 1
Figure 1. Figure 1: (a) An example of the left bot interface. (b) The experiment flows for both study 1 and study 2. Both studies used [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: (a) Histograms of perceived political distance be [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Coefficient plots from four mixed effects models [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Distributions of trust change by LLM verdict for [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: (Top) Interaction between perceived bot congru￾ency and headline congruency for correctly labeled true headlines in study 1. In this plot, both perceived bot distance and headline distance were binned into three groups via ter￾tiles. (Bottom) Trust change distributions across perceived bot distance for (b) correctly labeled true politically close headlines and (c) correctly labeled true politically distant… view at source ↗
Figure 6
Figure 6. Figure 6: Distributions of trust in true headlines that came from reputable sources, true headlines that were fact-checked, and false headlines across conservatives and liberals in study 1 and study 2. In (a) and (c), we show the distributions of initial trust, while in (b) and (d) we show the distributions of trust after correct treatment (bot labeled true headline as true, labeled false headlines as false). These … view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

300 extracted references · 7 canonical work pages

  1. [1]

    and Nguyen, Thuy and Wing, Coady and

    Bento, Ana I. and Nguyen, Thuy and Wing, Coady and. Evidence from Internet Search Data Shows Information-Seeking Responses to News of Local. Proceedings of the National Academy of Sciences , keywords =. 2020 , month = may, volume =. doi:10.1073/pnas.2005335117 , chapter =

    work page doi:10.1073/pnas.2005335117 2020

  2. [2]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , volume=

    Investigating Political and Demographic Associations in Large Language Models Through Moral Foundations Theory , author=. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , volume=

  3. [3]

    How Susceptible are Large Language Models to Ideological Manipulation? , booktitle =

    Kai Chen and Zihao He and Jun Yan and Taiwei Shi and Kristina Lerman , editor =. How Susceptible are Large Language Models to Ideological Manipulation? , booktitle =. 2024 , url =. doi:10.18653/V1/2024.EMNLP-MAIN.952 , timestamp =

    work page doi:10.18653/v1/2024.emnlp-main.952 2024

  4. [4]

    Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , volume=

    Politune: Analyzing the impact of data selection and fine-tuning on economic and political biases in large language models , author=. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , volume=

  5. [5]

    PsyArXiv

    Leveraging ChatGPT for efficient fact-checking , author=. PsyArXiv. April , volume=

  6. [6]

    Communications of the ACM , volume=

    Deep learning for AI , author=. Communications of the ACM , volume=. 2021 , publisher=

    2021

  7. [7]

    Proceedings of the 17th ACM Web Science Conference 2025 , pages=

    Accuracy and political bias of news source credibility ratings by large language models , author=. Proceedings of the 17th ACM Web Science Conference 2025 , pages=

    2025

  8. [8]

    2025 8th International Conference on Advanced Algorithms and Control Engineering (ICAACE) , pages=

    Unmasking digital falsehoods: A comparative analysis of LLM-based misinformation detection strategies , author=. 2025 8th International Conference on Advanced Algorithms and Control Engineering (ICAACE) , pages=. 2025 , organization=

    2025

  9. [9]

    arXiv preprint arXiv:2403.11169 , year=

    Correcting misinformation on social media with a large language model , author=. arXiv preprint arXiv:2403.11169 , year=

    arXiv

  10. [10]

    Frontiers in Artificial Intelligence , volume=

    The perils and promises of fact-checking with large language models , author=. Frontiers in Artificial Intelligence , volume=. 2024 , publisher=

    2024

  11. [11]

    Journal of Computational Social Science , volume=

    In generative AI we trust: can chatbots effectively verify political information? , author=. Journal of Computational Social Science , volume=. 2025 , publisher=

    2025

  12. [12]

    2025 , publisher=

    @ Grok Is This True? LLM-Powered Fact-Checking on Social Media , author=. 2025 , publisher=

    2025

  13. [13]

    Gizmodo , note =

    Zeff, Maxwell , title =. Gizmodo , note =. 2023 , month =

    2023

  14. [14]

    2025 , month =

    Elizabeth Melimopoulos , title =. 2025 , month =

    2025

  15. [15]

    Proceedings of the 2022 CHI conference on human factors in computing systems , pages=

    Birds of a feather don’t fact-check each other: Partisanship and the evaluation of news in Twitter’s Birdwatch crowdsourced fact-checking program , author=. Proceedings of the 2022 CHI conference on human factors in computing systems , pages=

    2022

  16. [16]

    Journal of applied research in memory and cognition , volume=

    Searching for the backfire effect: Measurement and design considerations , author=. Journal of applied research in memory and cognition , volume=. 2020 , publisher=

    2020

  17. [17]

    Development and validation of the Misinformation Susceptibility Self-Report (MiSS) , author=

  18. [18]

    British Journal of Political Science , volume=

    Does counter-attitudinal information cause backlash? Results from three large survey experiments , author=. British Journal of Political Science , volume=. 2020 , publisher=

    2020

  19. [19]

    Journal of communication , volume=

    Emotions, partisanship, and misperceptions: How anger and anxiety moderate the effect of partisan bias on susceptibility to political misinformation , author=. Journal of communication , volume=. 2015 , publisher=

    2015

  20. [20]

    Communication research , volume=

    Misinformation and polarization in a high-choice media environment: How effective are political fact-checkers? , author=. Communication research , volume=. 2020 , publisher=

    2020

  21. [21]

    Personality and Social Psychology Bulletin , volume=

    Rumors in retweet: Ideological asymmetry in the failure to correct misinformation , author=. Personality and Social Psychology Bulletin , volume=. 2024 , publisher=

    2024

  22. [22]

    Communications Psychology , volume=

    Democrats are better than Republicans at discerning true and false news but do not have better metacognitive awareness , author=. Communications Psychology , volume=. 2023 , publisher=

    2023

  23. [23]

    Public Opinion Quarterly , volume=

    Truth and bias, left and right: testing ideological asymmetries with a realistic news supply , author=. Public Opinion Quarterly , volume=. 2023 , publisher=

    2023

  24. [24]

    Nature Human Behaviour , volume=

    Political polarization of news media and influencers on Twitter in the 2016 and 2020 US presidential elections , author=. Nature Human Behaviour , volume=. 2023 , publisher=

    2016

  25. [25]

    Journal of Experimental Social Psychology , volume=

    False memories of fabricated political events , author=. Journal of Experimental Social Psychology , volume=. 2013 , publisher=

    2013

  26. [26]

    , author=

    Effects of associative inference on individuals’ susceptibility to misinformation. , author=. Journal of Experimental Psychology: Applied , volume=. 2023 , publisher=

    2023

  27. [27]

    arXiv preprint arXiv:2507.06306 , year=

    Humans overrely on overconfident language models, across languages , author=. arXiv preprint arXiv:2507.06306 , year=

    Pith/arXiv arXiv

  28. [28]

    Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency , pages=

    Search Engines in the AI Era: A Qualitative Understanding to the False Promise of Factual and Verifiable Source-Cited Responses in LLM-based Search , author=. Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency , pages=

    2025

  29. [29]

    Nature human behaviour , volume=

    Spotting false news and doubting true news: a systematic review and meta-analysis of news judgements , author=. Nature human behaviour , volume=. 2025 , publisher=

    2025

  30. [30]

    What you need to know about Grok and the controversies surrounding it , year =

  31. [31]

    arXiv preprint arXiv:2404.09356 , year=

    Llempower: Understanding disparities in the control and access of large language models , author=. arXiv preprint arXiv:2404.09356 , year=

    Pith/arXiv arXiv

  32. [32]

    arXiv preprint arXiv:2509.12652 , year=

    Don't Change My View: Ideological Bias Auditing in Large Language Models , author=. arXiv preprint arXiv:2509.12652 , year=

    arXiv

  33. [33]

    Proceedings of the National Academy of Sciences , volume=

    Fact-checking information from large language models can decrease headline discernment , author=. Proceedings of the National Academy of Sciences , volume=. 2024 , publisher=

    2024

  34. [34]

    2021 IEEE International Workshop on Information Forensics and Security (WIFS) , pages=

    Scalable fact-checking with human-in-the-loop , author=. 2021 IEEE International Workshop on Information Forensics and Security (WIFS) , pages=. 2021 , organization=

    2021

  35. [35]

    Handbook of theories of social psychology , volume=

    A theory of heuristic and systematic information processing , author=. Handbook of theories of social psychology , volume=

  36. [36]

    , author=

    Heuristic versus systematic information processing and the use of source versus message cues in persuasion. , author=. Journal of personality and social psychology , volume=. 1980 , publisher=

    1980

  37. [37]

    The American economic review , volume=

    The economic theory of agency: The principal's problem , author=. The American economic review , volume=. 1973 , publisher=

    1973

  38. [38]

    Academy of management review , volume=

    Agency theory: An assessment and review , author=. Academy of management review , volume=. 1989 , publisher=

    1989

  39. [39]

    Corporate governance , pages=

    Theory of the firm: Managerial behavior, agency costs and ownership structure , author=. Corporate governance , pages=. 1976 , publisher=

    1976

  40. [40]

    ACM Computing Surveys (CSUR) , volume=

    A survey of fake news: Fundamental theories, detection methods, and opportunities , author=. ACM Computing Surveys (CSUR) , volume=. 2020 , publisher=

    2020

  41. [41]

    Proceedings of the Third Workshop on Fact Extraction and VERification (FEVER) , year=

    Proceedings of the Third Workshop on Fact Extraction and VERification (FEVER) , author=. Proceedings of the Third Workshop on Fact Extraction and VERification (FEVER) , year=

  42. [42]

    ACM SIGKDD explorations newsletter , volume=

    Fake news detection on social media: A data mining perspective , author=. ACM SIGKDD explorations newsletter , volume=. 2017 , publisher=

    2017

  43. [43]

    International Journal of Communication , volume=

    Automated fact-checking to support professional practices: systematic literature review and meta-analysis , author=. International Journal of Communication , volume=

  44. [44]

    arXiv preprint arXiv:2601.05050 , year=

    Large language models can effectively convince people to believe conspiracies , author=. arXiv preprint arXiv:2601.05050 , year=

    Pith/arXiv arXiv

  45. [45]

    Multidisciplinary International Symposium on Disinformation in Open Online Media , pages=

    Striking the balance in using LLMs for fact-checking: A narrative literature review , author=. Multidisciplinary International Symposium on Disinformation in Open Online Media , pages=. 2024 , organization=

    2024

  46. [46]

    Current Opinion in Psychology , pages=

    Using conversational AI to reduce science skepticism , author=. Current Opinion in Psychology , pages=. 2025 , publisher=

    2025

  47. [47]

    Nature , pages=

    Persuading voters using human--artificial intelligence dialogues , author=. Nature , pages=. 2025 , publisher=

    2025

  48. [48]

    Science , volume=

    Durably reducing conspiracy beliefs through dialogues with AI , author=. Science , volume=. 2024 , publisher=

    2024

  49. [49]

    Journal of Applied Social Psychology , volume=

    Effects of fact-checking warning labels and social endorsement cues on climate change fake news credibility and engagement on social media , author=. Journal of Applied Social Psychology , volume=. 2023 , publisher=

    2023

  50. [50]

    Political behavior , volume=

    Fake claims of fake news: Political misinformation, warnings, and the tainted truth effect , author=. Political behavior , volume=. 2021 , publisher=

    2021

  51. [51]

    New Media & Society , volume=

    Asymmetric adjustment: Partisanship and correcting misinformation on Facebook , author=. New Media & Society , volume=. 2023 , publisher=

    2023

  52. [52]

    Journal of Online Trust and Safety , volume=

    Twitter’s disputed tags may be ineffective at reducing belief in fake news and only reduce intentions to share fake news among Democrats and Independents , author=. Journal of Online Trust and Safety , volume=

  53. [53]

    Public Opinion Quarterly , volume=

    The COVID-19 infodemic and the efficacy of interventions intended to reduce misinformation , author=. Public Opinion Quarterly , volume=. 2022 , publisher=

    2022

  54. [54]

    Proceedings of the International AAAI Conference on Web and Social Media , volume=

    Does the Source of a Warning Matter? Examining the Effectiveness of Veracity Warning Labels Across Warners , author=. Proceedings of the International AAAI Conference on Web and Social Media , volume=

  55. [55]

    Journal of Quantitative Description: Digital Media , volume=

    What makes news sharable on social media? , author=. Journal of Quantitative Description: Digital Media , volume=

  56. [56]

    Frontiers in psychology , volume=

    Measuring individual differences in generic beliefs in conspiracy theories across cultures: Conspiracy Mentality Questionnaire , author=. Frontiers in psychology , volume=. 2013 , publisher=

    2013

  57. [57]

    Aubin , title =

    Galen Stocking, Amy Mitchell, Katerina Eva Matsa, Regina Widjaya, Mark Jurkowitz, Shreenita Ghosh, Aaron Smith, Sarah Naseer, Christopher St. Aubin , title =. 2022 , note =

    2022

  58. [58]

    Journalism Studies , volume=

    Feeling Misinformed? The Role of Perceived Difficulty in Evaluating Information Online in News Avoidance and News Fatigue , author=. Journalism Studies , volume=. 2024 , publisher=

    2024

  59. [59]

    arXiv preprint arXiv:2407.21592 , year=

    Does the Source of a Warning Matter? Examining the Effectiveness of Veracity Warning Labels Across Warners , author=. arXiv preprint arXiv:2407.21592 , year=

    Pith/arXiv arXiv

  60. [60]

    Despite Meta Ending Its Third-Party Fact-Checking Program, Most People Still Want Fact-Checkers on Social Media , author=

  61. [61]

    2024 , notes =

    UK examines foreign states’ role in sowing discord leading to riots , author =. 2024 , notes =

    2024

  62. [62]

    Studies in Conflict & Terrorism , volume=

    White supremacist terrorism in charlottesville: Reconstructing ‘unite the right’ , author=. Studies in Conflict & Terrorism , volume=. 2023 , publisher=

    2023

  63. [63]

    Georgetown Journal of International Affairs , volume=

    Take the Redpill: Understanding the Allure of Conspiratorial Thinking among Proud Boys , author=. Georgetown Journal of International Affairs , volume=

  64. [64]

    PolitiFact (June 30), https://www

    Misinformation and the Jan 6 insurrection: when ‘patriot warriors’ were fed lies , author=. PolitiFact (June 30), https://www. politifact. com/article/2021/jun/30/misinformation-and-jan-6-insurrection-when-patriot , year=

    2021

  65. [65]

    PNAS nexus , volume=

    Deplatforming did not decrease Parler users’ activity on fringe social media , author=. PNAS nexus , volume=. 2023 , publisher=

    2023

  66. [66]

    The New York Times , volume=

    How social media amplifies misinformation more than information , author=. The New York Times , volume=

  67. [67]

    Journal of communication , volume=

    Partisan selective sharing: The biased diffusion of fact-checking messages on social media , author=. Journal of communication , volume=. 2017 , publisher=

    2017

  68. [68]

    Nature Human Behaviour , volume=

    Fact-checker warning labels are effective even for those who distrust fact-checkers , author=. Nature Human Behaviour , volume=. 2024 , publisher=

    2024

  69. [69]

    Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems , pages=

    Birds of a feather don’t fact-check each other: Partisanship and the evaluation of news in Twitter’s Birdwatch crowdsourced fact-checking program , author=. Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems , pages=

    2022

  70. [70]

    The International Journal of Press/Politics , volume=

    Who uses fact-checking sites? The impact of demographics, political antecedents, and media use on fact-checking site awareness, attitudes, and behavior , author=. The International Journal of Press/Politics , volume=. 2020 , publisher=

    2020

  71. [71]

    Management science , volume=

    The implied truth effect: Attaching warnings to a subset of fake news headlines increases perceived accuracy of headlines without warnings , author=. Management science , volume=. 2020 , publisher=

    2020

  72. [72]

    Scientific Reports , volume=

    People adhere to content warning labels even when they are wrong due to ecologically rational adaptations , author=. Scientific Reports , volume=. 2025 , publisher=

    2025

  73. [73]

    Plos one , volume=

    Data quality in online human-subjects research: Comparisons between MTurk, Prolific, CloudResearch, Qualtrics, and SONA , author=. Plos one , volume=. 2023 , publisher=

    2023

  74. [74]

    The Journal of Politics , volume=

    Political misinformation and factual corrections on the Facebook news feed: Experimental evidence , author=. The Journal of Politics , volume=. 2022 , publisher=

    2022

  75. [75]

    Political behavior , volume=

    Real solutions for fake news? Measuring the effectiveness of general warnings and fact-check tags in reducing belief in false stories on social media , author=. Political behavior , volume=. 2020 , publisher=

    2020

  76. [76]

    Americans: Much misinformation, bias, inaccuracy in news , author=. Gallup. Retrieved from , year=

  77. [77]

    2024 , publisher=

    How news coverage of misinformation shapes perceptions and trust , author=. 2024 , publisher=

    2024

  78. [78]

    2020 , organization=

    Why so much harmful content has proliferated online-and what we can do about it , author=. 2020 , organization=

    2020

  79. [79]

    The Atlantic , volume=

    Why the past 10 years of American life have been uniquely stupid , author=. The Atlantic , volume=

  80. [80]

    Journal of Experimental Political Science , volume=

    Taking the cloth: Social norms and elite cues increase support for masks among white evangelical Americans , author=. Journal of Experimental Political Science , volume=. 2023 , publisher=

    2023

Showing first 80 references.