Exposure to opposing views on social media can increase political polarization

Christopher A. Baila,1, Lisa P. Argyleb, Taylor W. Browna, John P. Bumpusa, Haohan Chenc, M. B. Fallin Hunzakerd, Jaemin Leea, Marcus Manna, Friedolin Merhouta, and Alexander Volfovskye

aDepartment of Sociology, Duke University, Durham, NC 27708; bDepartment of Political Science, Brigham Young University, Provo, UT 84602; cDepartment of Political Science, Duke University, Durham, NC 27708; dDepartment of Sociology, New York University, New York, NY 10012; and eDepartment of Statistical Science, Duke University, Durham, NC 27708

Edited by Peter S. Bearman, Columbia University, New York, NY, and approved August 9, 2018 (received for review March 20, 2018)

There is mounting concern that social media sites contribute to challenges for the study of social media echo chambers and political polarization by creating “echo chambers” that insulate political polarization, since it is notoriously difficult to establish people from opposing views about current events. We surveyed whether social media networks shape political opinions, or vice a large sample of Democrats and Republicans who visit Twit- versa (27–29). ter at least three times each week about a range of social Here, we report the results of a large field experiment designed policy issues. One week later, we randomly assigned respon- to examine whether disrupting selective exposure to partisan dents to a treatment condition in which they were offered information among Twitter users shapes their political attitudes. financial incentives to follow a Twitter bot for 1 month that Our research is governed by three preregistered hypotheses. The exposed them to messages from those with opposing political first hypothesis is that disrupting selective exposure to parti- ideologies (e.g., elected officials, opinion leaders, media orga- san information will decrease political polarization because of nizations, and nonprofit groups). Respondents were resurveyed intergroup contact effects. A vast literature indicates contact at the end of the month to measure the effect of this treat- between opposing groups can challenge stereotypes that develop ment, and at regular intervals throughout the study period to in the absence of positive interactions between them (30). Stud- monitor treatment compliance. We find that Republicans who ies also indicate intergroup contact increases the likelihood of followed a liberal Twitter bot became substantially more con- deliberation and political compromise (31–33). However, all of servative posttreatment. Democrats exhibited slight increases these previous studies examine interpersonal contact between in liberal attitudes after following a conservative Twitter bot, members of rival groups. In contrast, our experiment creates although these effects are not statistically significant. Notwith- virtual contact between members of the public and opinion lead- standing important limitations of our study, these findings have ers from the opposing political party on a social media site. significant implications for the interdisciplinary literature on polit- It is not yet known whether such virtual contact creates the ical polarization and the emerging field of computational social science. Significance political polarization | computational social science | social networks | social media | sociology Social media sites are often blamed for exacerbating political polarization by creating “echo chambers” that prevent people from being exposed to information that contradicts their pre- olitical polarization in the United States has become a central existing beliefs. We conducted a field experiment that offered Pfocus of social scientists in recent decades (1–7). Americans a large group of Democrats and Republicans financial com- are deeply divided on controversial issues such as inequality, gun pensation to follow bots that retweeted messages by elected control, and immigration—and divisions about such issues have officials and opinion leaders with opposing political views. become increasingly aligned with partisan identities in recent Republican participants expressed substantially more conser- years (8, 9). Partisan identification now predicts preferences vative views after following a liberal Twitter bot, whereas about a range of social policy issues nearly three times as well Democrats’ attitudes became slightly more liberal after fol- as any other demographic factor—such as education or age (10). lowing a conservative Twitter bot—although this effect was These partisan divisions not only impede compromise in the not statistically significant. Despite several limitations, this design and implementation of social policies but also have far- study has important implications for the emerging field of reaching consequences for the effective function of democracy computational social science and ongoing efforts to reduce more broadly (11–15). political polarization online. America’s cavernous partisan divides are often attributed to “echo chambers,” or patterns of information sharing that rein- Author contributions: C.A.B., L.P.A., T.W.B., J.P.B., H.C., M.B.F.H., J.L., M.M., F.M., and force preexisting political beliefs by limiting exposure to oppos- A.V. designed research; C.A.B., L.P.A., T.W.B., H.C., M.B.F.H., J.L., M.M., and F.M. per- ing political views (16–20). Concern about selective exposure formed research; C.A.B., T.W.B., H.C., J.L., and A.V. contributed new reagents/analytic tools; C.A.B., L.P.A., T.W.B., H.C., M.B.F.H., J.L., M.M., F.M., and A.V. analyzed data; and to information and political polarization has increased in the C.A.B., L.P.A., T.W.B., M.B.F.H., M.M., F.M., and A.V. wrote the paper. age of social media (16, 21–23). The vast majority of Ameri- The authors declare no conflict of interest. cans now visit a social media site at least once each day, and a This article is a PNAS Direct Submission. rapidly growing number of them list social media as their primary This open access article is distributed under Creative Commons Attribution- source of news (24). Despite initial optimism that social media NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND). might enable people to consume more heterogeneous sources Data deposition: All data, code, and the markdown file used to create this report of information about current events, there is growing concern will be available at this link on the Dataverse: https://dataverse.harvard.edu/dataverse. that such forums exacerbate political polarization because of xhtml?alias=chris bail. social network homophily, or the well-documented tendency of 1 To whom correspondence should be addressed. Email: [email protected] people to form social network ties to those who are similar to This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. themselves (25, 26). The endogenous relationship between social 1073/pnas.1804840115/-/DCSupplemental. network formation and political attitudes also creates formidable Published online August 28, 2018.

9216–9221 | PNAS | September 11, 2018 | vol. 115 | no. 37 www.pnas.org/cgi/doi/10.1073/pnas.1804840115 Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 ale al. et Bail nuneteaon fepsr oteitreto edescribe we intervention the to exposure would of reasoned amount we the to which influence according interest use, Twitter assignment of of randomized level frequency also (ii) respondents’ We and polarization: party events. political political current to to in linked attachment of been level have (i) that according ran- variables respondents block two stratified a to further used we that group, design each domization within and, respondents, our lican in participating before Review consent Institutional research. informed the respon- provided All by Twit- University. York dents on approved New and follow University was Duke they at Boards research accounts mine including the Our to behavior, of ter. used background online we partisan which their the ID, about Twitter information their additional report to indi- asked demographic in were of and describe range we media that a social a cators and of on sources, use media attitudes, issues conventional political policy other about about statements mation or of agree (α vari- range scale to a seven-point respondents a via outcome period asked with key study that disagree the the and instrument during 2017 measure survey ideology 10-item mid-October political surveys in in These change survey able: each later. 10-min times mo Republi- three a least 1.5 self-identified complete at hired Twitter recruit to We visit week to design. who Democrats research firm and our cans survey of professional overview an a provides 1 Fig. examine Design that Research any media social of on aware public—and particular. not broader in are the among we found liberal dynamics have trend, such in studies this of increase of number no a evidence or Although (42). little positions exhibit voting Democrats voting conserva- whereas call more substantially roll tive upon become in build have polarization also Republicans asymmetric wherein we Finally, observe examine sus- 26). that and that (34, studies create differences sociology that value also narratives such cultural and We tain schemas in 41). cultural studies tradition, (40, deeper the diversity recent and and upon certainty con- change build prioritize indicate value that that liberals studies whereas values liberals. recent than hold conservatives upon servatives among builds occur hypothesis to This likely more be will counterattitudinal media. to social exposure on prolonged messages of assess evaluate impact to two to aim broader we designed in the Instead, not inaccuracies. is replicate factual study of correct to our to invasion attempts Yet failed US 39). (38, 2003 have studies the recent findings to these prior factual destruction Iraq—although mass corrects of possessed that Hussein weapons Saddam that information notion the to as inaccuracies—such where exposed experiments are survey this via respondents in effects studies their backfire Many observe increases (34–37). literature and beliefs preexisting groups to to accentu- between commitment prone which differences are reasoning, perceived attitudes motivated ates own using exposed their them with are sev- counterargue conflict who spans that people messages political now to disciplines—indicates politi- exacerbate literature—which academic opposing that This eral effects with backfire (34–37). those create polarization to may exposure views cal suggest that nonverbal studies other partic- and be cues could offline facial in arguments incivility of of settings. escalation Such the absence prevent might manner. the that gestures people uncivil in emboldens rife an forums ularly the in media whether act social understanding—or to of mutual anonymity positive relative of type same ernsprt edeprmnsfrDmcai n Repub- and Democratic for experiments field separate ran We effects backfire that is hypothesis preregistered third Our of wave recent more a upon builds hypothesis second Our = 1) u uvyas olce infor- collected also survey Our (10). .91) ial,alrespondents all Finally, Appendix. SI ramn pnflycmlatrsodnsidctsti effect this indicates respondents compliant a of fully effect on the upon points estimates treatment that 0.12 model our increased although Republicans scale, seven-point cautious treated most that level Our significant. is with highly estimate increase are they also but effects compliance, of These posttreatment. conserva- views more Treated tive substantially significant. exhibited statistically contrast, were by effects Republicans, compli- these of of level none with size ance, in increase that posttreatment attitudes categories two last These questions. exclusive. survey these mutually success- the are of who during all those answered week one are fully each Respondents” of Compliant administered content “Fully answer the tweets period. to about bots’ “Partially able all—questions the period. not were of one—but study who least those entire at are the Respondents” throughout fol- Compliant bot who those our describes lowed Respondents” confidence 95% ways. Compliant three and in “Minimally 90% treatment with describe compliance 3 measured We Fig. intervals. and in estimates, lines con- point horizontal more the unstandardized became describe they Circles indicate to servative. scores response in positive Negative liberal and effects. more period treatment, became of respondents influence indicate the in scores liberal/conservative mitigate described pretreatment to covariates score respondents’ scale other for 12 control as We well this as on liberal/conservative scores scale pretreatment the predict for controlling that above, on described scale models scores multivariate posttreatment via These respondents’ produced observed. we were respondents among estimates differen- compliance the of for rates account tial Complier which the (CACE) as Effects well Causal as condi- Average effects, treatment (ITT) the Intent-to-Treat to the assigned or being tion, of effect the reports 3 Fig. Results the identify to day. able each were retweeted 50.2% picture and animal week, mes- of each content retweeted the were about sages questions bots substantive the all followed answer Approximately to who able Republicans invitation. and our Democrats of accepted and 62% Democrats Republicans of 64.9% of survey. bot, Twitter (pretreatment) 57.2% a initial follow to the invited survey those from Of final questions a same the complete the of to deleted with asked conclusion but were the bot respondents At the period, survey. by study day weekly a the picture twice before a tweeted identify immediately was and that content bots animal the Twitter an the about of by questions produced answer tweets the to of them asked to that (up surveys incentives financial additional bots, study’s the of design to the ideolo- refer about please political details follow opposing further to with For those (44). likely gies than more Twitter are on ideologies other assumes each political that similar technique with network-sampling those a via accounts lead- These identified opinion groups). were nonprofit officials, and 4,176 elected organizations, of media (e.g., These ers, list accounts experiments. a from our Twitter sampled of political randomly each messages for retweeted and bot bot bots Twitter Respondents Twitter liberal a mo. conservative would created bots we a 1 the illustrates, messages 2 for the Fig. of As day content retweet. the told each of were informed messages not they were 24 that account, retweet Twitter offered would automated were or condition respon- bot, an time, treatment Twitter using this the At thus in (43). condition, dents design treatment survey unrelated” a “ostensibly randomly to we later, respondents week assigned One Republicans). 751 and Democrats of likelihood overall the thereby change. and opinion paragraph following the in lhuhtetdDmcasehbtdsihl oeliberal more slightly exhibited Democrats treated Although offered were respondents compliance, treatment monitor To (901 survey pretreatment our to responses 1,652 received We IAppendix. SI PNAS | etme 1 2018 11, September 8 ocmlt weekly complete to $18) | o.115 vol. 1t olwa follow to $11 | IAppendix. SI o 37 no. | 9217

SOCIAL SCIENCES Initial Survey Randomization Weekly Surveys Post-Survey

Respondents were offered $11 One week later, respondents Respondents in treatment Respondents were to provide their Twitter ID and were assigned to treatment conditions informed they are offered $12 to repeat complete a 10-minute survey and control conditions within eligible to receive up to $6 each the pre-treatment strata created using pre- week during the study period about their political attitudes, survey one month treatment covariates that for correctly answering social media use, and describe attachment to party, questions about the content of after initial survey. media consumption habits frequency of Twitter use, messages retweeted by Twitter (demographics provided by and overall interest in .Bots. survey firm). current events.

TreatmentTreatment TreatmentTreatment OfferedOffered $$1111 t oto followfollow TTwitterwitter botbot thatthat rretweetsetweets 2424 mmessagesessages fromfrom RepublicansRepublicans liberalliberal aaccountsccounts eeachach ddayay fforor 1 monthmonth

ControlControl ControlControl

TreatmentTreatment TreatmentTreatment OfferedOffered $$1111 t oto followfollow TTwitterwitter botbot thatthat rretweetsetweets 2424 mmessagesessages fromfrom conservativeconservative aaccountsccounts eeachach ddayay fforor DemocratsDemocrats 1 monthmonth

ControlControl ControlControl

Fig. 1. Overview of research design.

is substantially larger (0.60 points). These estimates correspond Discussion and Conclusion to an increase in conservatism between 0.11 and 0.59 standard Before discussing the implications of these findings, we first note deviations. important limitations of our study. Readers should not interpret

9218 | www.pnas.org/cgi/doi/10.1073/pnas.1804840115 Bail et al. Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 ale al. et Bail might views opposing Twit- to use exposure US how not unclear entire do also the is Americans It to of (24). generalized ter majority be a not because date, should to population, media above social findings on among conducted the is experiments ours field Although largest settings. the all in views political polarization opposing increase to will exposure that evidence as findings our ilr Clinton Hillary Clinton Liberal (Small NetworkComponentPictured) Parenthood Planned ilr Clinton Hillary Clinton 0 Ivanka TrumpIvanka ieEz RW) Sntrni Mri ’aly @martinomalley @CarlyFiorina @realDonaldTrump @Lessig O’Malley Martin @RealBenCarson Lawrence Donald Lessig Trump Fiorina Carly @SenatorEnzi @SenJohnBarrasso BenCarson ...etc @stevedaines @SenatorTester John Barrasso(R-WY) Enzi(R-WY) Mike Steve Daines(R-MT) @lisamurkowski Jon Tester(D-MT) ...etc Don ...etc (R-AK) Lisa Murkowski ...etc Young (R-AK) @repdonyoung Hillary Clinton @HillaryClinton Bot #1 Tim Kaine Mike PenceMike .14 (Small NetworkComponentPictured) E l e c oadTrump Donald Trump t e d O Sarah Sanders Mike PenceMike f f i c i iaMurkowski Lisa Murkowski a .28 l s CNN Sarah Sanders Liberal/Conservative Scale Mike PenceMike .42 Tim Kaine Tim Kaine i.2. Fig. Sarah Sanders FOX Tucker Carlson eino td’ wte bots. Twitter study’s of Design .57 tv Daines Steve Daines Ivanka TrumpIvanka P r e s Foundation Heritage i d e n A= a aiiyo u nig,bcuew fee u respondents our offered we exter- because the findings, evaluate such our further of intervention to validity needed an nal are individuals to studies Such Future reactions ours. infrequently. or different as so independents, quite do as exhibit but identify might Twitter who addi- use people In who study world. those not the did of parts we other tion, in polarization political shape t i Mike PenceMike a l C .71 a n tv Daines Steve Daines d i Ivanka TrumpIvanka d a t e

.... s v oadTrump Donald Trump iaMurkowski Lisa Murkowski 0 0 11 Bot #2 PNAS Sarah Sanders .85 01 0 1 Conservative | 0 Mike PenceMike etme 1 2018 11, September .... 4 6 3 5 2 all Twitteraccounts thenamesof Extract (n=4,176). originate outsidetheU.S. and accounts that U.S. government agencies, Twitter accounts from less than15aswell as dropping non-elected they follow; everyone candidates, and network ofallelected Create directed (n=636,738). candidates follow that these563elected candidates. and presidential handles TwitterCollect distribution . quantiles ofthe and 5-7(conservative) from the1-3(liberal) sample oftweets tweet arandom Create botsthat 4,176 opinionleaders. ideology score for liberal/conservative component to create (see Supp. Materials). large no. offollowers scores ofaccounts with Analysis. Adjust Correspondence leaders” andconduct the 4,176 “opinion following patternsof matrix thatdescribes Create adjacency | o.115 vol. | o 37 no. | 9219

SOCIAL SCIENCES More Liberal More Conservative More Liberal More Conservative

Fig. 3. Effect of following Twitter bots that retweet messages by elected officials, organizations, and opinion leaders with opposing political ideologies for 1 mo, on a seven-point liberal/conservative scale where larger values indicate more conservative opinions about social policy issues, for experiments with Democrats (n = 697) and Republicans (n = 542). Models predict posttreatment liberal/conservative scale score and control for pretreatment score on this scale as well as 12 other covariates described in SI Appendix. Circles describe unstandardized point estimates, and bars describe 90% and 95% confidence intervals. “Respondents Assigned to Treatment” describes the ITT effect for Democrats (ITT = −0.02, t = −0.76, p = 0.45, n = 416) and Republicans (ITT = 0.12, t = 2.68, p = 0.008, n = 316). “Minimally-Compliant Respondents” describes the CACE for respondents who followed one of the study’s bots for Democrats (CACE = −0.04, t = −0.75, p = 0.45, n of compliant respondents = 271) and Republicans (CACE = 0.19, t = 2.73, p < 0.007, n of compliant respondents = 181). “Partially-Compliant Respondents” describes the CACE for respondents who correctly answered at least one question, but not all questions, about the content of a bot’s tweets during weekly surveys throughout the study period for Democrats (CACE = −0.05, t = −0.75, p = 0.45, n of compliant respondents = 211) and Republicans (CACE = 0.31, t = 2.73, p <.007, n of compliant respondents = 121). “Fully-Compliant Respondents” describes the CACE for respondents who answered all questions about the content of the bot’s tweets correctly for Democrats (CACE = −0.14, t = −0.75, p = 0.46, n of compliant respondents = 66) and Republicans (CACE = 0.60, t = 2.53, p < 0.01, n of compliant respondents = 53). Although treated Democrats exhibited slightly more liberal attitudes posttreatment that increase in size with level of compliance, none of these effects were statistically significant. In contrast, treated Republicans exhibited substantially more conservative views posttreatment that increase in size with level of compliance, and these effects are highly significant.

financial incentives to read messages from people or organiza- psychology, communications, and information science. Although tions with opposing views. It is possible that Twitter users may we found no evidence that exposing Twitter users to opposing simply ignore such counterattitudinal messages in the absence views reduces political polarization, our study revealed signif- of such incentives. Perhaps the most important limitation of our icant partisan differences in backfire effects. This finding is study is that we were unable to identify the precise mechanism important, since our study examines such effects in an exper- that created the backfire effect among Republican respondents imental setting that involves repeated contact between rival reported above. Future studies are thus urgently needed not only groups across an extended time period on social media. Our to determine whether our findings replicate in different popula- field experiment also disrupts selective exposure to informa- tions or within varied social settings but to further identify the tion about politics in a real-world setting through a combina- precise causal pathways that create backfire effects more broadly. tion of survey research, bot technology, and digital trace data Future studies are also needed because we cannot rule out collection. This methodological innovation enabled us to col- all alternative explanations of our findings. In SI Appendix, we lect information about the nexus of social media and politics present additional analyses that give us confidence that our results with high granularity while developing techniques for measuring are not driven by Hawthorne effects, partisan “learning” pro- treatment compliance, mitigating causal interference, and veri- cesses, variation in the ideological extremity of messages by party, fying survey responses with behavioral data—as we discuss in SI or demographic differences in social media use by age. At the Appendix. Together, we believe these contributions represent an same time, we are unable to rule out other alternative explana- important advance for the nascent field of computational social tions discussed in SI Appendix. For example, it is possible that our science (46). findings resulted from increased exposure to information about Although our findings should not be generalized beyond party- politics, and not exposure to opposing messages per se. Similarly, identified Americans who use Twitter frequently, we note that increases in conservatism among Republicans may have resulted recent studies indicate this population has an outsized influence from increased exposure to women or racial and ethnic minori- on the trajectory of public discussion—particularly as the media ties whose messages were retweeted by our liberal bot. Finally, itself has come to rely upon Twitter as a source of news and a our intervention only exposed respondents to high-profile elites window into public opinion (47). Although limited in scope, our with opposing political ideologies. Although our liberal and con- findings may be of interest to those who are working to reduce servative bots randomly selected messages from across the liberal political polarization in applied settings. More specifically, our and conservative spectrum, previous studies indicate such elites study indicates that attempts to introduce people to a broad are significantly more polarized than the general electorate (45). range of opposing political views on a social media site such as It is thus possible that the backfire effect we identified could be Twitter might be not only be ineffective but counterproductive— exacerbated by an antielite bias, and future studies are needed to particularly if such interventions are initiated by liberals. Since examine the effect of online intergroup contact with nonelites. previous studies have produced substantial evidence that inter- Despite these limitations, our findings have important impli- group contact produces compromise and mutual understanding cations for current debates in sociology, political science, social in other contexts, however, future attempts to reduce political

9220 | www.pnas.org/cgi/doi/10.1073/pnas.1804840115 Bail et al. Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 1 ce H atl M(2016) LM Bartels CH, Achen 11. increas- How public: American the in polarization Political (2014) D Carroll M, Dimock 10. 3 oelL etkwM hpr M(07 rae nentuei o soitdwith associated not is use Internet Greater (2017) JM Shapiro M, Gentzkow L, Boxell 23. polarization. political and (2011) E Media Pariser 21. (2013) M Prior expression 20. public activate media news the How (2017) A White B, Schneer G, King 18. changing The effects? minimal of era new A (2008) S Iyengar WL, Bennett 14. (2011) JS Fishkin (1993) 13. JP McIver GC, Wright RS, Erikson 12. 4 ernA(05 oilmdauae 0521 PwRsCn,Wsigo,DC). Washington, Cent, Res (Pew 2005-2015 usage: media Social (2015) A Perrin 24. twitter. on polarization Political (2011) M Francisco J, Ratkiewicz M, Conover 22. (2013) S Sobieraj JM, Berry 19. ideologically to Exposure science. (2001) Political C Sunstein (2015) LA 17. Adamic S, Messing E, Bakshy 16. (2002) C Sunstein 15. sdwti hssuya ela ik oorpeeitainstatement, preregistration our to links as well as study this within used See Methods and Materials effective divides. more partisan deliv- be America’s settings—might bridge others—perhaps offline to whether vehicles in or and nonelites by learning effects ered likely require backfire most are likely create positions issue most to or tactics, will messages, of media types which social on polarization ale al. et Bail .MsnL(2018) L Mason 9. lattes? drink (2009) liberals M do Why Levendusky (2015) 8. M Macy Y, Shi D, public. DellaPosta American the 7. in polarization Political (2008) SJ polarization. Abrams political MP, Fiorina of Dynamics 6. (2007) P Bearman D, Baldassarri 5. (2015) DJ Hopkins J, Sides and 4. polarization Political constraint: without Partisans (2008) on A evidence Gelman New D, lines: Baldassarri party across 3. loathing and Fear more (2015) become SJ Westwood attitudes S, social Iyengar American’s Have 2. (1996) B Bryson J, Evans P, DiMaggio 1. States Government Responsive and compromise, DC). politics, Washington, affect Cent, Res antipathy (Pew partisan life everyday and uniformity ideological ing Chicago). Press, Republicans Became Conservatives 1473–1511. Sci Polit 72:784–811. London). opinion. public American in trends polarization. group polarized? atrgot npltclplrzto mn sdmgahcgroups. demographic USA us Sci among polarization political in growth faster 133:89–96. Think We How and Read We 101–127. Incivility New agendas. national influence and Princeton). Press, Univ communication. political of foundations Consultation ies esadoiino Facebook. on opinion and news diverse IAppendix SI CmrdeUi rs,Cmrde UK). Cambridge, Press, Univ (Cambridge 114:10612–10617. 11:563–588. mJSociol J Am Ofr nvPes Oxford). Press, Univ (Oxford Ofr nvPes Oxford). Press, Univ (Oxford h itrBbl:HwteNwProaie e sCagn What Changing Is Web Personalized New the How Bubble: Filter The nii gemn:HwPltc eaeOrIdentity Our Became Politics How Agreement: Uncivil o ealddsrpino l aeil n methods and materials all of description detailed a for coCabr,Bs .Gr,Ipahet n Beyond and Impeachment, Gore, v. Bush Chambers, Echo Republic.com hntePol pa:Dlbrtv eorc n Public and Democracy Deliberative Speak: People the When mJPltSci Polit J Am 102:690–755. h atsnSr:HwLbrl eaeDmcasand Democrats Became Liberals How Sort: Partisan The PictnUi rs,Princeton). Press, Univ (Princeton oiia oaiaini mrcnPolitics American in Polarization Political h urg nuty oiia pno ei n the and Media Opinion Political Industry: Outrage The Pnun e York). New (Penguin, eorc o elss h lcin oNtProduce Not Do Elections Why Realists: for Democracy PictnUi rs,Princeton). Press, Univ (Princeton Science 59:690–707. Ui hcg rs,Chicago). Press, Chicago (Univ mJSociol J Am Science 358:776–780. Commun J ulcOiinadPlc nteAmerican the in Policy and Opinion Public 114:408–446. 348:1130–1132. 58:707–731. nuRvPltSci Polit Rev Annu mJSociol J Am mSco Rev Sociol Am rcNt Acad Natl Proc Ui Chicago (Univ (Bloomsbury, (Princeton nuRev Annu ICWSM 120: 16: 7 ae ,Rbna ,Cekvc ,Kt ,NboM(00 h ovlto of coevolution The (2010) M Neblo N, Katz C, Chetkovich B, Rubineau networks. in D, distinction Lazer cultural and 27. resources social Cultural in (2014) Homophily S Vaisey feather: A, a Edelmann of Birds 26. (2001) JM Cook L, Smith-lovin M, McPherson 25. 6 ae ,e l(09 iei h ewr:Tecmn g fcmuainlsocial and media computational Online disinformation: of and age Propaganda, Partisanship, coming (2017) The al. et network: R, the Faris in 47. Life (2009) al myth?. et a D, polarization Is Lazer (2008) KL 46. Saunders AI, Abramowitz 45. practice. in Gr theory democratic 33. Testing networks: social Cross-cutting (2002) DC Mutz (2004) 32. J Sprague PE, theory. contact Johnson intergroup R, of test Huckfeldt meta-analytic 31. A (2006) LR Tropp composition?. TF, network health Pettigrew influence of worldviews 30. cultural adoption Can the (2010) in O Lizardo homophily S, of Vaisey study 29. experimental An (2011) D Centola 28. 3 roka ,KlaJ(06 ual euigtashba edeprmn on experiment field A transphobia: Barber reducing 44. Durably (2016) J Kalla D, with Broockman associated 43. threat (2016) and DA uncertainty Hopkins manage M, to Grossmann needs 42. Are sets (2007) different al. on et rely JT, conservatives and Jost Liberals 41. (2009) BA Nosek J, factual Haidt steadfast J, attitudes’ Graham Mass effect: 40. backfire elusive The (2018) factual E steadfast Porter attitudes’ T, Wood Mass political effect: 39. backfire of elusive The (2016) persistence beliefs. E political Porter The of T, Wood evaluation fail: the 38. in skepticism corrections Motivated (2006) When M Lodge CS, (2010) Taber J 37. Reifler polarization: attitude B, and Nyhan assimilation 36. Biased (1979) MR Lepper L, Ross CG, Lord 35. (2015) C Bail 34. ovrain bu hssuypirt u eerh u okwssup- the was and work Foundation, Sage Our Russell research. the Foundation. our Foundation, Science helpful National to Carnegie for the Luks prior by Samantha study ported and this Nyhan, Brendan about Lupia, York conversations Arthur Li, New Fan and King, University ACKNOWLEDGMENTS. Duke at Boards Review approved University. Institutional was research the Our findings. our by of discus- explanations extended alternative an of and sion checks, robustness additional materials, replication esScPsychol Soc Pers attitudes. political and networks Poetics networks. oilForces Social behavior. h 06US rsdnileeto,(eka li etItre o avr Univ, Harvard Soc Internet Cent Klein (Berkman MA). election, Cambridge, presidential U.S. 2016 the science. Rev Sci Polit Am Networks Communication Within UK). Opinions Cambridge, Diverse of siainuigtitrdata. twitter using estimation canvassing. door-to-door Democrats Interest Group extremity? ideological or conservatism political foundations. moral of adherence. adherence. Sci Polit J Am misperceptions. evidence. considered subsequently on 37:2098–2109. theories prior of effects The Princeton). Press, Univ (Princeton Behav Polit nudK en ,Set K, Herne K, onlund ¨ 21)Brso h aefahrtettgte:Bysa da point ideal Bayesian together: tweet feather same the of Birds (2014) P a ´ 46:22–37. Science Science nuRvSociol Rev Annu oi Behav Polit Behav Polit 37:995–1020. 88:1595–1618. erfid o niMsi rneOgnztosBcm Mainstream Became Organizations Fringe Anti-Muslim How Terrified: 50:755–769. 323:721–723. oi Behav Polit 90:751–783. 96:111–126. PNAS 334:1269–1272. esSca Psychol Social Pers J 10.1007/s11109-018-9443-y. , 1–29. , etakPu iago usieHlyu,Gary Hillygus, Sunshine DiMaggio, Paul thank We | Science al ¨ Ofr nvPes Oxford). Press, Univ (Oxford 27:415–444. 21)De nlv eieainplrz opinions? polarize deliberation enclave Does (2015) M a ¨ etme 1 2018 11, September 32:303–330. oi Anal Polit oi Commun Polit 352:220–224. smercPltc:IelgclRpbiasand Republicans Ideological Politics: Asymmetric 23:76–91. 96:1029–1046. oiia iareet h Survival The Disagreement: Political esScPyhlBull Psychol Soc Pers 27:248–274. | o.115 vol. CmrdeUi Press, Univ (Cambridge Polit J | 70:542–555. esScPsychol Soc Pers J o 37 no. 33:989–1007. | 9221 J

SOCIAL SCIENCES