The detection of political deepfakes
Open accessThe DOI matches its Crossref record; 57 of the 70 references it cites were found in a registry or an index, and 5 are not checked yet; 7 of the 54 sources its claims cite were checked for support.
Abstract
Deepfake technology, allowing manipulations of audiovisual content by means of artificial intelligence, is on the rise. This has sparked concerns about a weaponization of manipulated videos for malicious ends. A theory on deepfake detection is presented and three preregistered studies examined the detection of deepfakes in the political realm (featuring UK’s Prime Minister Boris Johnson, Studies 1–3, or former U.S. President Barack Obama, Study 2). Based on two system models of information processing as well as recent theory and research on fake news, individual differences in analytic thinking and political interest were examined as predictors of correctly detecting deepfakes. Analytic thinking (Studies 1 and 2) and political interest (Study 1) were positively associated with identifying deepfakes and negatively associated with the perceived accuracy of a fake news piece about a leaked video (whether or not the deepfake video itself was presented, Study 3). Implications for research and practice are discussed.
Verified 2026-10-06 · 5 references not checked yet
Overview
CoChat Summary
TL;DR
Three preregistered German online studies found that analytic thinking (Studies 1–2) and political interest (Study 1) predicted better detection of political deepfakes, while a deepfake embedded in a fake news article did not increase perceived accuracy or alter leadership ratings (Study 3).
Key findings
- In Study 1, 82 of 199 participants (41.2%) stated or suspected that the Boris Johnson deepfake was fake; analytic thinking (OR = 1.23, p = .013) and political interest (OR = 1.49, p = .025) were positively associated with detection.
- In Study 2, the Johnson deepfake was rated more manipulated (M = 5.34, SD = 1.70) than the Obama deepfake (M = 4.91, SD = 1.99), both above the scale midpoint; analytic thinking predicted Johnson detection (B = .13, p = .026) and content-based detection (B = .21, p = .001), but not Obama detection, and political interest was not a significant predictor.
- Study 2 found an indirect effect of analytic thinking on deepfake detection through implausible content identification (effect = .077, 95% CI [.015, .143]), but not through technological glitches (effect = .005, 95% CI [−.019, .028]); technological-glitch recognition was unrelated to analytic thinking, r(231) = .027, p = .680.
- In Study 3, adding the deepfake did not affect perceived accuracy of the fake news article (M = 3.13 vs. 2.98, p = .290) or perceived leadership of Boris Johnson (M = 3.68 vs. 3.60, p = .537); analytic thinking (B = −.14, p = .045) and political interest (B = −.40, p < .001) predicted lower accuracy ratings irrespective of deepfake exposure.
Methods
- Three preregistered online studies with German samples recruited via Clickworker.de: final N = 199 (Study 1), 233 (Study 2), and 445 (Study 3); stimuli were a Boris Johnson deepfake (Studies 1–3) and an Obama deepfake (Study 2).
- Study 1 used an open-ended thought-listing task; Study 2 used closed-ended deepfake detection and indicator ratings; Study 3 embedded the Johnson deepfake in a fake news article versus a news-only control.
- Predictors were a seven-item Cognitive Reflection Test for analytic thinking and a single political-interest item; analyses included logistic regression, linear regression, mediation, and ANOVA.
Limitations
- Stated: the studies relied on stimuli available in 2020, largely one Boris Johnson deepfake; the Obama deepfake may be perceived as satirical or entertaining; self-report detection measures may inflate detection because mentioning manipulation cues respondents; future research should use unobtrusive measures.
- Stated: context information was not an empirical focus, and the authors note that warning citizens about deepfakes could unintentionally increase distrust of real content.
- Evident: the German samples and specific political figures limit generalizability; Study 2 gave mixed support for political interest, and Study 3’s null deepfake effects contrast with prior work by Hwang et al. (2021).
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures
Verification
Claims the paper makes
6 Valid · 1 Not valid · 32 Not checked · 81 Own results and interpretations
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“Recent research has linked individual differences in analytic thinking with skepticism about various epistemically suspect information, such as fake news (Pennycook & Rand, 2019;2020), religious and paranormal beliefs (Pennycook et al., 2016;Shenhav et al., 2012), conspiracy theories (Swami et al., 2014), and pseudo-profound bullshit (Pennycook et al., 2015;Pennycook & Rand, 2020).”
§ The detection of political deepfakes > The deepfake detection model
What the sources say
- Not checked yetcites Pennycook & Rand, 2019 matched to “Lazy, not biased: Susceptibility to partisan fake news is better explained by lack of reasoning than by motivated reasoning” (2019) source
- Supportedcites Pennycook et al., 2016 matched to “Atheists and agnostics are more reflective than religious believers: Four empirical studies and a meta-analysis” (2016) source
The Introduction reports that a propensity for analytic reasoning predicts paranormal disbelief. The General Discussion reports a negative association between analytic thinking and religious belief, replicated in four studies. This citation is attached to the religious and paranormal beliefs portion of the sentence, not the other examples. (LLM, conf=0.92)
- Supportedcites Swami et al., 2014 matched to “Analytic thinking reduces belief in conspiracy theories” (2014) source
Swami et al. is cited specifically for conspiracy theories, not the other types of information in the sentence. Study 1’s Results and discussion (§2.2; page not provided) report that a stronger analytic thinking style was associated with lower belief in conspiracy theories. (LLM, conf=0.94)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Audiovisual content is met with credulity as videos are accepted almost axiomatically as accurate depictions of reality by journalists, politicians, jurisprudence, and everyday citizens (Farish, 2020;Fallis, 2021;Hancock & Bailenson, 2021).”
§ The detection of political deepfakes > The deepfake detection model
What the sources say
- Not checked yetcites Farish, 2020 matched to “Do deepfakes pose a golden opportunity? Considering whether English law should adopt California’s publicity right in the age of the deepfake” (2020) source
- Partly supportedcites Fallis, 2021 matched to “The Epistemic Threat of Deepfakes” (2021) sourceExperimental — accuracy not yet measured
Fallis's Section 2 ('The Epistemic Value of Videos') supports the general credulity half of the claim: videos are a source 'that we can simply trust without a lot of verifying' and 'we are significantly more likely to accept video evidence than other sources of information, such as testimony,' with people making significant decisions from videos and treating what they depict as what actually occurred. However, our copy does not attribute this axiomatic acceptance to the specific institutional actors the sentence lists — journalists, politicians, and jurisprudence/courts. Politicians appear only as subjects or consequences (videos leading to politicians losing elections, police officers being fired and prosecuted), reporters only as testifiers whose utterances a video records (Lester Holt on the evening news), and no passage addresses courts or legal evidentiary practice (searched: courtroom/jury/legal evidence; journalists/media trust). The 'audiovisual' framing is also narrower in the source, which concerns videos (and, in §4, recordings generally) rather than audio content being met with credulity. (LLM, conf=0.72)
- Not checked yetcites Hancock & Bailenson, 2021 matched to “The Social Impact of Deepfakes” (2021) source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“In sum, the studies provide first evidence, that users tend to ascribe accuracy to deepfakes (Ko ¨bis et al., 2021), even if the deepfakes are imperfect (Dobber et al., 2021).”
§ The detection of political deepfakes > Current evidence on deepfake detection
What the sources say
- Not checked yetcites Ko ¨bis et al., 2021
- Supportedcites Dobber et al., 2021 matched to “Do (microtargeted) deepfakes have real effects on political attitudes?” (2021) source
Dobber et al. are cited for the qualification that this can occur even with an imperfect deepfake, not for the broader accuracy claim cited to Köbis et al. In Amplification (p. 18), the authors describe flaws in their deepfake but say almost no participant questioned the video’s veracity. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Every source the paper cites for this claim was checked and supports it.
In the paper
“Individuals have a general tendency to suspend effortful cognitive processes and to rely on associative processes and intuitions (e.g., Pennycook et al., 2016, reported that 65% gave an incorrect answer to the 'bat and ball' problem).”
§ The detection of political deepfakes > The deepfake detection model
What the sources say
- Supportedcites Pennycook et al., 2016 matched to “Atheists and agnostics are more reflective than religious believers: Four empirical studies and a meta-analysis” (2016) source
The Introduction (p. 1) sets out the dual-process framework the claim paraphrases — Type 1 processes 'intuitive and autonomously cued' versus Type 2 processes 'reflective and require working memory' — and states that for the bat-and-ball problem 'The majority of participants respond "10 cents"... (e.g., 64.9%...) This is the response that typically comes to mind upon an initial read of the problem, but it is incorrect,' so that the participant 'must be willing to stop and think analytically.' The Method > Materials section reinforces the default-intuition reading: each analytic-thinking measure 'is intended to cue an incorrect intuitive response that requires additional analytic processing to override.' The cited 65% matches the paper's 64.9% figure, though the paper itself attributes that figure to its earlier report (ref. [15], Pennycook et al. 2015) rather than to new data from the four studies here; the term 'associative' is the citing authors' wording for the paper's 'intuitive/autonomously cued'. (LLM, conf=0.82)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Indeed, the higher participants' scores on analytic thinking (Studies 1-2) and political interest (Study 1), the better participants identified the deepfakes.”
§ The detection of political deepfakes > Abstract
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
47 more claims
Create a free account to see all 120 claims and to verify the 44 references still open.
- Probably not a substantial claim. 38
- It may state two separate claims as one. 37
- The quote only partly states this claim. 4
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“When faced with deepfake content of politicians, participants rely on contextual knowledge in the form of political literacy to identify spoofs [20, 21].”
From Warning: Humans cannot reliably detect speech deepfakes (2023) · cites this paper as [21]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“people with more analytic cognitive styles worldwide are more accurate at discerning between authentic and fabricated political videos62 and true and false headlines related to COVID-1963.”
From Human detection of political speech deepfakes across transcripts, audio, and video (2024) · cites this paper as [62]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Overall, political interest and partisanship do not appear to have an impact on detection [23, 32].”
From Human Performance in Deepfake Detection: A Systematic Review (2025) · cites this paper as [23]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Cognitive ability and analytical thinking are positively associated with detection accuracy [32, 23].”
From Human Performance in Deepfake Detection: A Systematic Review (2025) · cites this paper as [23]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Appel and Prietzel [23] found that an individual ' s tendency to think analytically is correlated with general skepticism, which in turn is associated with being better able to discern between genuine and manipulated information.”
From Human Performance in Deepfake Detection: A Systematic Review (2025) · cites this paper as [23]
Experimental — accuracy not yet measured
5 more claims about this paper
Create a free account to see every claim other papers make about this one.
Citations
Cites
(2013) · Cited by 44,563
Statistical power analyses using G∗Power 3.1: tests for correlation and regression analyses
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009) · Behavior Research Methods · doi:10.3758/brm.41.4.1149 · Cited by 37,615
Multiple regression: testing and interpreting interactions
(1992) · Choice Reviews Online · doi:10.5860/choice.29-3352 · Cited by 36,654
Advances In Experimental Social Psychology
(2022) · Advances in Experimental Social Psychology · doi:10.1016/s0065-2601(22)x0002-6 · Cited by 22,259
The spread of true and false news online
Vosoughi, S., Roy, D., & Aral, S. (2018) · Science · doi:10.1126/science.aap9559 · Cited by 8,847 · cited 2×
“On these platforms, research found that people are especially prone to sharing negative and novel information, with false political stories being particularly effective in being spread (Vosoughi et al., 2018).”
BackgroundBruner, J. (1986) · doi:10.4159/9780674029019 · Cited by 8,657
The Nature and Origins of Mass Opinion
Zaller, J. R. (1992) · doi:10.1017/cbo9780511818691 · Cited by 7,785
Cognitive Reflection and Decision Making
Frederick, S. (2005) · Journal of Economic Perspectives · doi:10.1257/089533005775196732 · Cited by 5,256
Central and Peripheral Routes to Advertising Effectiveness: The Moderating Role of Involvement
Petty, R. E., Cacioppo, J. T., & Schumann, D. (1983) · Journal of Consumer Research · doi:10.1086/208954 · Cited by 4,610
Dual-Process Theories of Higher Cognition: Advancing the Debate
Evans, J. S. B. T., & Stanovich, K. E. (2013) · Perspectives on Psychological Science · doi:10.1177/1745691612460685 · Cited by 4,271
Research works printed but not linked: 9
- Agarwal, S., Farid, H., Gu, Y., He, M., Nagano, K., & Li, H. (2019, June). Protecting world leaders against deep fakes. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops (pp. 38-45). IEEE. https://perma.cc/2bcs-m8j3
- Cole, S. (2018). We are truly fucked: Everyone is making AI-generated fake porn now. VICE. https://perma.cc/mwn2-z655
- Genette, G. (1987). Seuils. Edition du Seuil. (English language version:
- Genette, G. [1997] Paratexts. Thresholds of interpretation. Cambridge University Press.
- Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., & Bengio, Y. (2014). Generative adversarial nets. In Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence, & K. Q. Weinberger (Eds.), Advances in Neural Information Processing Systems 27 (pp. 2672-2680). MIT Press.
- Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux.
- Mervosh, S. (2019, May 24). Distorted videos of Nancy Pelosi spread on Facebook and Twitter, helped by Trump. The New York Times. https://perma.cc/a26h-4pf3
- Sankaranarayanan, A., Groh, M., Picard, R., & Lippman, A. (2021, August). The presidential deepfakes dataset. Paper presented at the workshop 'AIofAI: 1st workshop on adverse impacts and collateral effects of artificial intelligence technologies'. http://ceur-ws.org/Vol- 2942/paper3.pdf
- Spivak, R. (2019). "Deepfakes": The newest way to commit one of the oldest crimes. The Georgetown Law Technology Review, 3(2), 339-400. https://perma.cc/y32v-y4x9
Cited by
Items 101–133 of 133
Casás, J., Rodríguez, A. S., & Vázquez, A. I. R. (2026) · Social Sciences · doi:10.3390/socsci15040247 · Cited by 2
Beyond Seeing Is Believing: On Crowdsourced Detection of Audiovisual DeepfakesPreprint
Soprano, M., Cioci, A., & Mizzaro, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.04797
Raza, M., & Ahmad, M. R. (2026) · International Journal of Advanced Mass Communication and Journalism · doi:10.22271/27084450.2026.v7.i5a.175
Korop, O., & Baranets, V. (2026) · Mediaforum Analytics Forecasts Information Management · doi:10.31861/mediaforum.2026.18.216-233
Pasaribu, Y. M., & El-Hussaini, A. (2026) · Open Access Indonesia Journal of Social Sciences · doi:10.37275/oaijss.v9i3.326
Balcı, Ş., & Özben, K. (2026) · Kritik İletişim Çalışmaları Dergisi · doi:10.53281/kritik.1877790
Választási deepfake: szabályozási kérdések, kihívások és alkotmányos dilemmák
Szikora, T. A. (2026) · Iustum Aequum Salutare · doi:10.69695/ias.2026.2.10
Bi, C., & Zhu, Y. (2026) · Cogent Arts and Humanities · doi:10.1080/23311983.2026.2696098
Redondo, R. Q., Gonzalo, S. B., & Jiménez, V. M. (2026) · Revista de Comunicación · doi:10.26441/rc25.2-2026-4566
Digital Campaigning under Hybrid Regimes: Hungary and Serbia in Comparative Perspective
Merkovity, N., & Toma, V. Z. (2026) · Comparative Southeast European Studies · doi:10.1515/soeu-2025-0073
Villacura-Herrera, C., Zaal, F. T. J. M., & Cox, R. F. A. (2026) · Computers in Human Behavior Reports · doi:10.1016/j.chbr.2026.101221
Pratama, N., & Sangare, B. (2026) · Justicia Insight · doi:10.70716/justin.v2i2.385
Gomez, H. (2026) · VISUAL REVIEW International Visual Culture Review / Revista Internacional de Cultura Visual · doi:10.62161/revvisual.v18.6396
Kenyeres, A. Z. (2025) · Acta Universitatis de Carolo Eszterházy Nominatae Sectio paedagogica · doi:10.46436/actauniveszterhazypaedagogica.2025.101
Medovarszki, I. (2025) · Acta Biologica Plantarum Agriensis (Eszterházy Károly University, Hungary)
Legal Protection Against Deepfakes: Preventive and Punitive Tools and Their Effectiveness
Pačmag, M., & Fiala, Z. (2026) · doi:10.62768/adjuris/2026/4/11
Angelic, S. F. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7338241
MINTAKÖVETÉS ÉS/VAGY ÖNÁLLÓ VÉLEMÉNY? : A politikai érdeklődés tendenciái Magyarországon
Szabó, A. (2025) · Repository of the Academy's Library (Library of the Hungarian Academy of Sciences) · Cited by 1
Kalivoshka, N. (2026) · Universal Library of business and economics. · doi:10.70315/uloap.ulbec.2026.0301018
Kritisk informationsbedömning i en deepfake-era
Alhaidar, H., Almkawi, R., & Ramirez, A. T. (2026) · Lund University Publications Student Papers (Lund University)
Akinyetun, T. S., & SOWEMIMO, A. (2026) · Arab Media Renewal Journal · doi:10.64548/amrj.2026.28
Deepfake Web Video Detection using Deep Neural Networks and LSTM Architectures
Ridzuan, A. M. A., & Safar, N. Z. M. (2025) · Applied Information Technology And Computer Science
Multimodal Analysis of Visual Framing in Bola Tinubu’s Presidential Campaign Video
Agbedejobi, P. (2024) · Journal of Arts and Humanities · doi:10.18533/m2j6wp47
ACCURACY IN IDENTIFYING DEEPFAKE VIDEOS: EXPLORING FACTORS THAT ENHANCE AND IMPAIR DETECTION
(2026) · Issues in Information Systems · doi:10.48009/1_iis_2026_120
Papa, E. C. S. (2024) · Soft Power
Lee, J., Kim, S., Lee, H., & Hwang, J. (2026) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2026.2740422
The Impact of Contextual Information in Social Media Posts on Human Deepfake Detection Accuracy
Heikkilä, S. (2026) · UTUPub (University of Turku)
Seeing Is Believing? Most People Struggle to Detect AI-Generated Images
Hove, M. F., Hjorth, F., Golovchenko, Y., Adler‐Nissen, R., Bechmann, A., van Dalen, A., HOBOLT, S. B., & de Vreese, C. H. (2026) · Social Media + Society · doi:10.1177/20563051261483376
Kurniawan, B. D. (2026) · MANTRA Journal of Multidisciplinary Research and Management · doi:10.68594/mantra.v1i1.60
Decision Heuristics and Discernment of Photorealistic Images in the Age of AIPreprint
Wojdynski, B. W., Shivers, B. N., & Ebba, H. K. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7597673
Deepfake or Real? How Political Alignment and Believability Influence Reporting on Social Media
冯敏绮, Lioliou, E., Dickert, S., & Panzarasa, P. (2026) · Information Systems Frontiers · doi:10.1007/s10796-026-10833-0
When Seeing Is No Longer Believing: Reddit Users’ Discernment of AI-Generated Visual Content
Cheng, Z., & Li, Y. (2026) · International journal of communication · doi:10.65476/68x13815
Gueorguieva, A.-M., & Weber, N. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2611.11658
Metadata
- Authors
- Appel, M., & Prietzel, F.
- Affiliations
- University of Würzburg (DE)
- Corresponding author
- Appel, Markus
- ORCID
- Appel, Markus
- Venue
- Journal of Computer-Mediated Communication
- Publisher
- Oxford University Press (OUP)
- Published
- 2022-07-01
- Volume
- 27
- Issue
- 4
- ISSN
- 1083-6101 (ISSN-L)
- OpenAlex
- W4285246761
- Semantic Scholar
- a63a569d24370da33b1433ecb22f404f1d4dec8f
- Cited by
- 95
- References
- 64
- Citations by year
- 2026: 27 · 2025: 29 · 2024: 28 · 2023: 8
- FWCI
- 23.38
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- Locations
- Journal of Computer-Mediated Communication (publishedVersion, cc-by)
- Topics
- Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Social Sciences › Psychology › Social Psychology › Deception detection and forensic psychology
- Keywords
- deepfake detection, political deepfakes, fake news, analytical thinking, political interest, misinformation, manipulated video, perceived accuracy, Boris Johnson, Barack Obama, information processing model
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-06, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-06 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.



