CoChat
Sign Up
Cited by 85Social Media and PoliticsMisinformation and Its ImpactsHate Speech and Cyberbullying Detection

Troll and divide: the language of online polarization

Open access
Simchon, Almog, Brady, William J, Van Bavel, Jay J·2022·PNAS Nexus
DOI 10.1093/pnasnexus/pgac019PMC PMC9802075

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

The affective animosity between the political left and right has grown steadily in many countries over the past few years, posing a threat to democratic practices and public health. There is a rising concern over the role that “bad actors” or trolls may play in the polarization of online networks. In this research, we examined the processes by which trolls may sow intergroup conflict through polarized rhetoric. We developed a dictionary to assess online polarization by measuring language associated with communications that display partisan bias in their diffusion. We validated the polarized language dictionary in 4 different contexts and across multiple time periods. The polarization dictionary made out-of-set predictions, generalized to both new political contexts (#BlackLivesMatter) and a different social media platform (Reddit), and predicted partisan differences in public opinion polls about COVID-19. Then we analyzed tweets from a known Russian troll source (N = 383,510) and found that their use of polarized language has increased over time. We also compared troll tweets from 3 countries (N = 79,833) and found that they all utilize more polarized language than regular Americans (N = 1,507,300) and trolls have increased their use of polarized rhetoric over time. We also find that polarized language is associated with greater engagement, but this association only holds for politically engaged users (both trolls and regular users). This research clarifies how trolls leverage polarized language and provides an open-source, simple tool for exploration of polarized communications on social media.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper develops and validates a 205-word dictionary of polarized language and applies it to social media data, finding that state-affiliated trolls from Russia, Iran, and Venezuela use more polarized language than American controls. It also finds that polarized language predicts greater retweet engagement mainly among politically engaged users, though the design cannot establish causality.

Key findings

  • The polarization dictionary discriminated polarized from nonpolarized tweets in cross-validation (Mpolarized = 6.70, SD = 9.08, N = 3,696 vs. Mnonpolarized = 4.39, SD = 6.60, N = 1,312; t(3156) = 9.79, P < 0.001, d = 0.27) and generalized to #BlackLivesMatter (d = 0.13), Reddit political communities (d = 0.31), and COVID-19 partisan opinion dynamics (r(48) = 0.45, P = 0.001; lagged 8 days r(40) = 0.67, P < 0.001).
  • Russian trolls used significantly more polarized language than American controls (M = 2.37, SD = 5.14, N = 61,413 vs. M = 1.47, SD = 5.35, N = 516,525; t(78,081) = 40.96, P < 0.001, d = 0.17), and politically matched Russian trolls also exceeded politically matched Americans (M = 5.16, SD = 8.00, N = 55,726 vs. M = 2.91, SD = 6.84, N = 55,726; t(108,836) = 50.61, P < 0.001, d = 0.30).
  • Iranian and Venezuelan trolls likewise used more polarized language than American controls (Iran: M = 2.15, SD = 4.66, N = 220,628 vs. M = 1.46, SD = 5.26, N = 929,908, d = 0.13; Venezuela: M = 1.80, SD = 4.67, N = 30,987 vs. M = 1.45, SD = 5.25, N = 953,197, d = 0.07; both P < 0.001).
  • Polarized language predicted retweets: for every polarized word in a tweet, retweets increased by 72% for Russian trolls (IRR = 1.72, CI [1.64, 1.79], P < 0.0001), and by 39% for politically engaged controls (IRR = 1.39, 95% CI [1.35, 1.48], P < 0.0001), but not for a random American sample (IRR = 1.19, 95% CI [0.80, 1.77], P = 0.390).

Methods

  • The authors built a polarization dictionary by combining data-driven exploration, manual pruning, GloVe word embeddings, and manual trimming; the final dictionary contained 205 words and was validated across four contexts: cross-validation, out-of-set #BlackLivesMatter, cross-platform Reddit, and predictive COVID-19 polling.
  • The main dataset included 2,306,233 original English tweets: 383,510 by Russian trolls, 329,453 by Iranian trolls, 85,970 by Venezuelan trolls, and 1,507,300 by American controls; troll data came from the Twitter Transparency Report, and additional data were collected via Twitter API 2.0 and academictwitteR.
  • Analyses included t-tests comparing polarization scores, weighted linear regressions of monthly polarization over time, hierarchical clustering of dictionary words into Issue and Affective components, and negative binomial generalized linear models predicting retweet counts from polarized language.

Limitations

  • The paper states the dictionary was built on data collected in 2015 and on three polarized topics, so it may be biased toward timely 2015 topics and restricted in scope; it includes some terms that may not seem strictly polarized (e.g., “people”), and the authors recommend using it comparatively with a baseline corpus.
  • The paper states that early social media accounts identified as trolls may not be authentic—some countries use hacked, purchased, or stolen accounts—so early Venezuelan or Iranian content may not have originated with the nation in question, potentially biasing over-time analyses.
  • The paper states the work is primarily quasi-experimental or correlational and cannot draw causal conclusions; it also notes that troll and control time frames were not identical, so trend differences should be treated as tentative, and some post hoc analyses had small sample sizes.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1093/pnasnexus/pgac019

Cites

60 of 60 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

0 of 41 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    The algorithm clustered the dictionary words in a manner somewhat consistent with the theoretical separation between issue and affective polarization.

    “We observed that in the highest level of division, the algorithm clustered the words in a manner that is somewhat consistent with the theoretical separation between issue and affective polarization”
    Quote in the paperNot a substantial claim
  • Findingown data

    The correlation between COVID-19 poll responses and Twitter language was strongest when Twitter language was lagged by 8 days.

    “A post hoc analysis revealed that the correlation between poll responses and twitter language was strongest when Twitter language was lagged by 8 days (i.e. pollt0, twittert8) r (40) = 0.67, P < 0.001”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The Civiqs partisan difference score validly measures public opinion polarization about COVID-19.
  • Findingown data

    Politically oriented Russian trolls use significantly more polarized language than their politically matched American sample.

    “Again, we find that politically oriented Russian trolls use significantly more polarized language than their politically matched American sample (Russian trolls: M = 5.16, SD = 8.00, and N = 55,726; American controls: M = 2.91, SD = 6.84, and N = 55,726), t(108,836) = 50.61, P < 0.001, and Cohen's d = 0.30”
    Quote in the paperSubstantial claim
    • Rests on an assumption: English-language tweets from trolls are most likely aimed at an international or American audience.
  • Findingown data

    Iranian trolls used less polarized language over time.

    “Iranian trolls used less polarized language (b = -0.03), R2 = 0.33, F(1, 85) = 41.06, and P < 0.001”
    Quote in the paperSubstantial claim
  • Findingown data

    Russian trolls used far more polarized language as time progressed.

    “We found that Russian trolls used far more polarized language as time progressed (b = 0.03), R2 = 0.46, F(1, 69) = 58.85, and P < 0.001”
    • Also stated in: § Abstract · plain
      “Then we analyzed tweets from a known Russian troll source (N = 383,510) and found that their use of polarized language has increased over time.”
    Quote in the paperSubstantial claim
  • May combine two claims. 20
  • The quote only partly states this claim. 3
  • Classified as probably not a substantial claim. 13

  • May combine two claims.

    Findingown data

    Polarized language is associated with greater engagement, but this association only holds for politically engaged users, both trolls and regular users.

    “We also find that polarized language is associated with greater engagement, but this association only holds for politically engaged users (both trolls and regular users).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In a post hoc analysis on the same time frame, Russian trolls used far more polarized language as time progressed while American control users did not.

    “Again, Russian trolls used far more polarized language as time progressed (b = 0.03), R2 = 0.51, F(1, 17) = 17.48, and P < 0.001 while American control users did not, (b = 0.006), R2 = 0.16, F(1, 17) = 3.14, and P = 0.094”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Polarized rhetoric was reduced among the Reddit community NeutralPolitics compared with other political communities, although the effect size was very small.

    “A planned comparison between the political groups revealed a significant difference in polarized rhetoric between NeutralPolitics and the other political communities (Mleft = 2.38, SD left = 4.61, N = 17,005; Mright = 2.57, SD right = 5.34, N = 15,859; and Mneutral = 2.24, SD neutral = 4.49, N = 9,772), t(42,633) = 4.12, P < 0.001, Cohen's d’ = 0.04.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: NeutralPolitics is a neutral reference of respectful political discourse.

  • The quote only partly states this claim.

    Findingown data

    Iranian trolls used significantly more polarized language than American control samples.

    “| Iran | 2.15 (4.66) | 220,628 | 1.46 (5.26) | 929,908 | 2016–11–23–2018–11–28 | 61.32 | 366,496 | 0.13 |”
    Quote in the paperSubstantial claim
    • Rests on an assumption: English-language tweets from trolls are most likely aimed at an international or American audience.

  • The quote only partly states this claim.

    Findingown data

    Venezuelan trolls used significantly more polarized language than American control samples.

    “| Venezuela | 1.80 (4.67) | 30,987 | 1.45 (5.25) | 953,197 | 2016–11–23–2018–12–07 | 12.87 | 33,588 | 0.07 |”
    Quote in the paperSubstantial claim
    • Rests on an assumption: English-language tweets from trolls are most likely aimed at an international or American audience.

  • Classified as probably not a substantial claim.

    Findingown data

    The Issue subcomponent of the polarization dictionary addresses ideological and policy keywords (e.g., liberal, conservative, socialism, gun-control).

    “while the Issue subcomponent addresses ideological and policy keywords (e.g. liberal, conservative, socialism, and gun-control)”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingcitation

    The Affective component of the polarization dictionary references negative moral-emotional words (e.g., kill, destroy, cheaters) and distinct ethnic and religious groups (e.g., Muslim and Jews).

    “the Affective component references instances of negative moral–emotional words (e.g. kill, destroy, and cheaters; (19)), and distinct ethnic and religious groups (e.g. Muslim andJews).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [19] Brady W J, Wills J A, Jost J T, Tucker J A, Van Bavel J J. 2017. Emotion shapes the diffusion of moralized content in social networks. Proc Natl Acad Sci USA. 114(28): 7313– 7318. 10.1073/pnas.1618923114 28652356 PMC5514704 source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    In all countries examined, tweets sent by trolls used significantly more polarized language than tweets sent by American control samples, both on affective and issue polarization (Ps < 0.005).

    “In all the countries we examined, the tweets sent by trolls used significantly more polarized language than tweets sent by American control samples (Ps < 0.005), both on affective and issue polarization”
    • Also stated in: § Discussion · plain
      “We found that state-affiliated trolls from Russia and other countries use more polarized language than a random sample of American users”
    • Also stated in: § Study 4: polarized language and engagement · boosted
      “Studies 2 and 3 demonstrated a link between polarized language use and troll accounts on social media.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    After adding followership as a covariate in the trolls analysis, the association between polarized language and retweets remains significant but the effect size diminishes (IRR = 1.61, 95% CI [1.57, 1.67], P < 0.0001).

    “After adding followership in the trolls analysis we find the same pattern of results, however the effect size diminishes: IRR = 1.61, 95% CI [1.57, 1.67], and P < 0.0001”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Trolls share controversial content and engage in highly polarized issues, but also use higher levels of polarized language as a tool in their discourse.

    “we found that trolls share controversial content and engage in highly polarized issues, but that they also use higher levels of polarized language as a tool in their discourse.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Rising partisan antipathy, or affective polarization, mirrors these attitudes.

    “These attitudes are mirrored in rising partisan antipathy; dislike toward members of the opposing ideology—a phenomenon known as “affective polarization” (3, 4, 5).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [3] Boxell L, Gentzkow M, Shapiro J M. 2020. Cross-country trends in affective polarization (No. 26669). Cambridge (MA): National Bureau of Economic Research. 10.3386/w26669 source

      Not checked in the free run.

    • cites [4] Finkel E J, et al. 2020. Political sectarianism in America. Science. 370(6516): 533– 536. 10.1126/science.abe1715 33122374 source

      Not checked in the free run.

    • cites [5] Iyengar S, Lelkes Y, Levendusky M, Malhotra N, Westwood S J. 2019. The origins and consequences of affective polarization in the United States. Ann Rev Polit Sci. 22(1): 129– 146. 10.1146/annurev-polisci-051117-073034 source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Previous methods for studying polarization can be computationally expensive and create a barrier for adoption for behavioral scientists who lack the required technical expertise.

    “Previous work studied polarization through network analysis or by exploring topics known to be polarized (39). These methodologies have several advantages (40, 41) but can be computationally expensive, create a barrier for adoption for behavioral scientists who lack the required technical expertise, and are most likely context-dependent which can undercut replicability (42).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [39] Demszky D, et al. 2019. Analyzing polarization in social media: method and application to tweets on 21 mass shootings. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), 2970– 3005.

      Not checked in the free run.

    • cites [40] Garimella K, Morales G D F, Gionis A, Mathioudakis M. 2018. Quantifying controversy on social media. ACM Trans Soc Comput. 1(1): 1– 27. 10.1145/3140565 source

      Not checked in the free run.

    • cites [41] Guerra P C, Meira W Jr, Cardie C, Kleinberg R. 2013. A measure of polarization on social media networks based on community boundaries. Proceedings of the 7th International AAAI Conference on Weblogs and Social Media. https://www.aaai.org/ocs/index.php/ICWSM/ICWSM13/paper/viewPaper/6104.

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    The Internet Research Agency engaged in online political tactics to sow intergroup conflict and influence US citizens during the 2016 presidential election and British citizens prior to the Brexit vote.

    “The IRA has engaged in online political tactics to sew intergroup conflict and influence US citizens during the 2016 presidential election (34) and British citizens prior to the Brexit vote (35).”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [34] Badawy A, Ferrara E, Lerman K. 2018. Analyzing the digital traces of political manipulation: The 2016 Russian interference Twitter campaign. Proceedings of the 2018 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM). 258– 265.

      Not checked in the free run.

    • cites [35] Llewellyn C, Cram L, Favero A, Hill R L. 2018. Russian troll hunting in a brexit Twitter archive. Proceedings of the 18th ACM/IEEE on Joint Conference on Digital Libraries, 361– 362. 10.1145/3197026.3203876 source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    Scholars have made a conceptual distinction between issue polarization (an ideological, policy-based political divide) and affective polarization (dislike, distrust, and general animosity of political partisans toward the other political side).

    “Scholars have made the conceptual distinction between issue polarization—an ideological, policy-based political divide, and affective polarization, i.e. dislike, distrust, and general animosity of political partisans toward the other political side (5, 61, 62).”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [5] Iyengar S, Lelkes Y, Levendusky M, Malhotra N, Westwood S J. 2019. The origins and consequences of affective polarization in the United States. Ann Rev Polit Sci. 22(1): 129– 146. 10.1146/annurev-polisci-051117-073034 source

      Not checked in the free run.

    • cites [61] Iyengar S, Sood G, Lelkes Y. 2012. Affect, not ideology: a social identity perspective on polarization. Pub Opin Quart. 76(3): 405– 431. 10.1093/poq/nfs038 source

      Not checked in the free run.

    • cites [62] Wilson A E, Parker V, Feinberg M. 2020. Polarization in the contemporary political and media landscape. Curr Opin Behav Sci. 34: 223– 228. 10.1016/j.cobeha.2020.07.005 source

      Not checked in the free run.

  • May combine two claims.

    Interpretationown data

    The polarization dictionary is composed of two subcomponents (Issue and Affective) that roughly map onto the theoretical distinction between issue and affective polarization.

    “This distinction is roughly reflected in the dictionary.”
    • Also stated in: § Results · boosted
      “we showed that the polarization dictionary is composed of 2 subcomponents that map onto theoretical elements of polarization (Issue and Affective).”
    • Also stated in: § Discussion · plain
      “We found that our data-driven dictionary taps into distinct theoretical elements of polarization”
    Quote in the paperSubstantial claim

  • May combine two claims. The quote only partly states this claim.

    Interpretationown data

    Polarized language is associated with greater traction on social media, but only in political contexts (i.e., for politically engaged users, both trolls and controls).

    “Overall, these results indicate that polarized language is associated with greater traction on social media, but only in political contexts.”
    • Also stated in: § Discussion · plain
      “this association only holds for politically engaged users—both trolls and controls.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    Polarized language is not only a marker for a static polarized state, but contributes to the polarization process and advances polarization rather than merely reflecting it.

    “we take this as evidence that polarized language is not only a marker for a static polarized state, but contributes to the polarization process.”
    • Also stated in: § Discussion · hedged
      “this finding suggests that polarized language advances polarization and not merely reflects it.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Prior work documented trolls' attempts to pollute the online environment with polarized content and sow discord among Americans.

    “These results expand on prior work documenting trolls’ attempts to pollute the online environment with polarized content and sow discord among Americans (65).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [65] Golovchenko Y, Buntain C, Eady G, Brown M A, Tucker J A. 2020. Cross-platform state propaganda: Russian trolls on Twitter and YouTube during the 2016 U.S. Presidential Election. Int J Press/Polit. 25: 357– 389. source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    The dictionary was validated using 4 strategies.

    “We validated this dictionary using 4 strategies”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    A study found that Russian trolls were trying to promote a progovernment agenda and dissolve government criticism, but were only successful at the latter.

    “In a study that explored the domestic effect of Russian trolls (i.e. messages that were targeted inwards to Russian users), it was found that trolls were trying to promote a progovernment agenda and dissolve government criticism (69); nevertheless, trolls were only successful at the latter”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [69] Sobolev A. 2018. How pro-government “trolls” influence online conversations in Russia. Los Angeles (CA): University of California, Los Angeles. http://www.wpsanet.org/papers/docs/2019W-Feb-Anton-Sobolev-Trolls-VA.pdf

      Not checked in the free run.

  • May combine two claims.

    Interpretationcitation

    Trolls win twice: once when they share the polarized content, and then again when it is echoed on other media platforms, creating a polarizing loop.

    “This way, trolls win twice: once when they share the polarized content, and then again when it is being echoed on other media platforms, creating a polarizing loop.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    There is evidence of growing polarization and partisan antipathy in the American public.

    “Given the evidence on the growing polarization and partisan antipathy in the American public (5)”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [5] Iyengar S, Lelkes Y, Levendusky M, Malhotra N, Westwood S J. 2019. The origins and consequences of affective polarization in the United States. Ann Rev Polit Sci. 22(1): 129– 146. 10.1146/annurev-polisci-051117-073034 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    A recent paper followed over 1,200 American Twitter users for 1 month in late 2017 and found that only a small fraction of users interacted with Russian trolls, and they did not observe any change in partisan attitude during that time among these users.

    “A recent paper followed over 1,200 American Twitter users for the course of 1 month in late 2017. The authors found that only a small fraction of users interacted with Russian trolls, and they did not observe any change in partisan attitude during that time among these users (68).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [68] Bail C A, et al. 2020. Assessing the Russian Internet Research Agency's impact on the political attitudes and behaviors of American Twitter users in late 2017. Proc Natl Acad Sci USA, 117(1): 243– 250. 10.1073/pnas.1906420116 31767743 PMC6955293 source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Prior research has revealed that trolls are often found in the centroids of clusters of polarized topics, driving the partisan discourse on both ends.

    “Prior research has revealed that when exploring the clusters of polarized topics, trolls are often found in the centroids of these clusters, driving the partisan discourse on both ends (37, 38, 55).”
    Quote in the paperNot a substantial claim0 of 3 sources checked
    • cites [37] Arif A, Stewart L G, Starbird K. 2018. Acting the part: examining information operations within# BlackLivesMatter discourse. Proc ACM Hum Comput Interact. 2(CSCW): 20. 10.1145/3274289. source

      Not checked in the free run.

    • cites [38] Broniatowski D A, et al. 2018. Weaponized health communication: Twitter bots and Russian trolls amplify the vaccine debate. Am J Pub Health. 108(10): 1378– 1384. 30138075 PMC6137759 source

      Not checked in the free run.

    • cites [55] Walter D, Ophir Y, Jamieson K H. 2020. Russian Twitter accounts and the partisan polarization of vaccine discourse, 2015-2017. Am J Pub Health. 110: e1– e7. PMC7144436 32191516 source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    Recent evidence suggests that troll's messages propagate to mainstream media and are represented as "the voice of the people".

    “recent evidence suggests that troll's messages propagate to mainstream media and are represented as “the voice of the people” (67).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [67] Lukito J, et al. 2019. The wolves in sheep's clothing: how Russia's Internet Research Agency tweets appeared in U.S. news as Vox Populi. Int J Press Polit. 25: 194016121989521.

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    The dictionary includes some terms that may not seem strictly polarized (e.g., people).

    “it also includes some terms that may not seem strictly polarized (e.g. people).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The work has focused primarily on quasi-experimental manipulations or correlational methodology.

    “this work has focused primarily on quasi-experimental manipulations or correlational methodology.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductnone given

    The polarization dictionary was built on data collected in 2015 and on 3 polarized topics.

    “The polarization dictionary has been built on data collected in 2015 and on 3 polarized topics.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    These results are consistent with other findings that do not find evidence for increased polarization during this brief time frame.

    “These results are consistent with other findings that do not find evidence for increased polarization during this brief time frame (70).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [70] Westwood S J, Peterson E, Lelkes Y. 2019. Are there still limits on partisan prejudice?. Pub Opin Quart. 83(3): 584– 597. 10.1093/poq/nfz034 source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductreasoning

    The dictionary is subject to bias about topics that were timely in 2015 and is potentially restricted in its scope.

    “Therefore, it is subjected to bias about topics that were timely in 2015 and is potentially restricted in its scope.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The authors employed a weighted regression analysis that downweights the relatively sparse early messages.

    “we employed a weighted regressions analysis that takes into account the relatively sparse nature of early messages (and therefore, downweights their importance).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductreasoning

    Early data may not have originated with the nation in question for some Venezuelan or Iranian content, and may have biased the polarization over time analyses.

    “Early data, therefore, may not have originated with the nation in question. While this was probably not the case with the Russian trolls dataset, it could be the case with some Venezuelan or Iranian content, and may have biased our polarization over time analyses.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Some countries use hacked, purchased, or stolen social media accounts.
    • Rests on an assumption: The Russian trolls dataset is authentic (not from hacked, purchased, or stolen accounts).

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductreasoning

    The authors attempted to mitigate the dictionary's scope limitation by expanding the lexicon using word-embeddings and testing its validation over multiple time periods.

    “We attempt to get around this limitation by expanding the lexicon using word-embeddings and testing its validation over multiple time periods.”
    Quote in the paperSubstantial claim
33 of 33

  • The paper relies on this without support.

    Ingroup bias in tweet diffusion is a valid marker of political polarization.

    “The database included tweets about contentious political topics that showed a range of ingroup bias in their spread through social networks (i.e. they either were shared with only the political ingroup or spread to 1 or more outgroup members).”

  • The paper relies on this without support.

    The Reddit communities selected as political have established ideologies that make them politically polarized.

    “We extracted up to 1,000 messages from 36 political communities with established ideologies (18 from each political side).”

  • The paper relies on this without support.

    NeutralPolitics is a neutral reference of respectful political discourse.

    “We extracted messages from the popular subreddit NeutralPolitics (www.reddit.com/r/NeutralPolitics), a reddit community devoted to factual and respectful political discourse.”

  • The paper relies on this without support.

    The Civiqs partisan difference score validly measures public opinion polarization about COVID-19.

    “We calculated a “partisan difference score” from 2020 February 25th until 2020 April 14th by subtracting the daily Republican net concern from the daily Democratic net concern, as reported by Civiqs”

  • The paper relies on this without support.

    English-language tweets from trolls are most likely aimed at an international or American audience.

    “We only used original tweets that were posted in the English language and were most likely aimed for an international/American audience.”

  • The paper relies on this without support.

    The countries examined (Russia, Iran, Venezuela) hold anti-American views.

    “We provide novel evidence that this mission spans several countries that hold antiAmerican views.”

  • The paper relies on this without support.

    Some countries use hacked, purchased, or stolen social media accounts.

    “Some countries use hacked, purchased, or stolen accounts.”

  • The paper relies on this without support.

    The Russian trolls dataset is authentic (not from hacked, purchased, or stolen accounts).

    “While this was probably not the case with the Russian trolls dataset”
8 of 8

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?

  • Polarized rhetoric about the pandemic on Twitter was highly correlated with polarized beliefs about the risks of the pandemic several days later.

    “For instance, polarized rhetoric about the pandemic on Twitter was highly correlated with polarized beliefs about the risks of the pandemic several days later (Simchon et al. 2022).”

    From Social Media and Morality. (2024) · cites this paper as Simchon et al. 2022

    The Dictionary validation section reports a strong correlation (r = 0.67) between polarized COVID-19 language on Twitter and partisan differences in pandemic concern at an eight-day offset. It does not clearly support the claimed direction of that offset: the passage specifies poll responses at t0 and Twitter language at t8, rather than beliefs occurring several days after the tweets. (LLM, conf=0.82)

    Experimental — accuracy not yet measured

  • While more voices can be heard through social media, more extreme voices may be prioritized and propaganda can foster divisive content, which might distort the information people consume.

    “Thus, while more voices can be heard through social media, more extreme voices may be prioritized (Bail 2022) and propaganda can foster divisive content (Simchon et al. 2022), which might distort the information people consume.”

    From Social Media and Morality. (2024) · cites this paper as Simchon et al. 2022

  • Different examples of misinformation often make use of the same underlying tropes.

    “different examples of misinformation often make use of the same underlying tropes (18, 19).”

    From Psychological inoculation improves resilience against misinformation on social media (2022) · cites this paper as [19]

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Consistentt(3156) = 9.79reported p < .001 · recomputed p < .001

    “t (3156) = 9.79, p < .001”

  • Consistentt(716.06) = 3.58reported p < .001 · recomputed p < .001

    “t (716.06) = 3.58, p < .001”

  • Consistentr(48) = 0.45reported p = .001 · recomputed p = .001

    “r (48) = 0.45, p = .001”

  • Consistentr(40) = 0.67reported p < .001 · recomputed p < .001

    “r (40) = 0.67, p < .001”

  • Consistentχ²(1) = 5263reported p < .0001 · recomputed p < .001

    “χ2(1) = 5263, p < .0001”

  • Consistentχ²(7) = 373563reported p < .0001 · recomputed p < .001

    “χ2(7) = 373,563, p < .0001”

  • Consistentχ²(7) = 2223reported p < .0001 · recomputed p < .001

    “χ2(7) = 2223, p < .0001”

7 of 7

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Overview attributes the full sample size to trolls

    Minor

    The overview says nearly 2.3 million tweets came from trolls and were compared with American tweets. The preprocessing section says 2,306,233 is the total *including* 1,507,300 American control tweets.

  • Stated tweet-collection period conflicts with the control sample

    Minor

    The methods say tweets were extracted starting in November 2017, but Study 2 says its US control sample includes tweets posted starting in November 2016.

2 of 2

Cites

Flagged first, then by citations

Research works printed but not linked: 9

  1. Klein, 2020, Why we're polarized
  2. Badawy, 2018, Analyzing the digital traces of political manipulation: The 2016 Russian interference Twitter campaign, 258
  3. Demszky, 2019, Analyzing polarization in social media: method and application to tweets on 21 mass shootings, 2970
  4. Twitter, 2018, Internet Research Agency (October 2018) data set
  5. Twitter, 2019, Iran (January 2019) data set
  6. Twitter, 2019, Venezuela (January 2019, set 1) data set
  7. Pennebaker, 2015, The development and psychometric properties of LIWC2015
  8. Grimmer, 2014, You lie! Patterns of partisan taunting in the US Senate
  9. Sobolev, 2018, How pro-government “trolls” influence online conversations in Russia

Cited by

Sorted by citations

Related

Authors
Simchon, A., Brady, W. J., & Van Bavel, J. J.
Affiliations
  • Ben-Gurion University of the Negev (IL)
  • Yale University (US)
  • Department of Psychology, New York University , New York, NY 10003 , USA
Corresponding author
Simchon, Almog
ORCID
Venue
PNAS Nexus
Publisher
Oxford University Press (OUP)
Published
2022-03
Volume
1
Issue
1
Pages
pgac019-pgac019
ISSN
2752-6542 (ISSN-L)
DOI
10.1093/pnasnexus/pgac019
PMC
PMC9802075
OpenAlex
W4221025338
Semantic Scholar
5450f688fee4c654c5c1d4449a9356f1c54117c6
Cited by
85
References
60
Citations by year
2026: 14 · 2025: 28 · 2024: 22 · 2023: 15 · 2022: 4 · 2021: 2
FWCI
26.28
Type
journal article
Language
en
Open access
Licence
Locations
  • PNAS Nexus (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
John Templeton Foundation (61378)
Topics
  • Social Sciences › Communication › Social Media and Politics
  • Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
  • Physical Sciences › Computer Science › Artificial Intelligence › Hate Speech and Cyberbullying Detection
Keywords
online polarization, language dictionaries, partisan bias, social media engagement, political communication, intergroup conflict, #BlackLivesMatter, Reddit, Twitter

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.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

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