Downloaded by guest on September 26, 2021 PNAS misperception strategy belief a maintaining and from forming shift in intermediaries systems. a consid- of that require roles elites one the to will from public ers the claims so to information those providing Doing on and reinforcing focused media. claims cues false the of and flow and identities forma- the the group reduce disrupt between to to linkages effective. is of more approach it tion best make the to however, and tar- Ultimately, to better how information understand corrective to these instead, get but Given effects not debunked. backfire is prevent communication been to scientific for have challenge public accu- they primary in the after less realities, persist years but typically misperceptions for congenial result, opinion more a from As promoting cues claims. media by rate overwhelmed instead, the be accumulate; and or or decay elites last to not seem do frequently review infor- often they corrective checks I fact of that like effects mation research accuracy-increasing when the the accuracy that However, belief suggests respondents. increasing by at typically effective received is emerging somewhat information an least corrective contrary, at that dura- the finds the To consensus explaining misperceptions. research in political factors other of media to bility relative and exag- role and research its generality its gerating subsequent distorting predispositions finding, how this or on show seized beliefs coverage I their controversial challenged. when a issue were about scientific misperception or a respondents political which endorsed in strongly effect” “backfire more for so-called (received 2020 a some- 8, provoke can March information times Fiske T. corrective Susan that Member indicated Board research Editorial Previous by accepted and MI, Arbor, 2019) Ann 17, Michigan, September of review University Hoffman, Andrew by Edited a Nyhan durability Brendan the misperceptions explain political not of does effect backfire the Why rvrilise nmc esatce 6.I w ftefive the “back- of a two called we In what (6). observed we articles conducted, news we that mock con- studies in about information issues corrective troversial journal to the exposure in varied randomly approach this of effects 5). misperceptions. (4, people’s reducing relevant of from effect identity information intended such conclusion, or the prevent having preferred salient potentially could is a resistance issue toward This the reason when and especially beliefs predispositions existing skep- their be contradicts or may it people when information that such is of are concern tical particular communication, A science sometimes (3). in known is model” well “deficit which the approach, as this to of referred cli- failures on The consensus scientific change. the mate of information—for evidence accurate providing providing by by instance, straight record the set opinion, to mass distort debate, public policy. public for warp to under- basis and can related factual (2), closely partisanship the often as mine such are systems mod- which belief genetically beliefs, and identities to These change (1). climate food ified from which beliefs, ranging unsupported issues or false plague over- these of to prevalence struggle the all come educators and journalists, Scholars, correct? W checking fact eateto oenet atot olg,Hnvr H03755 NH Hanover, College, Dartmouth Government, of Department n21,JsnRie n ulse natcetsigthe testing article an published I and Reifler Jason 2010, In try to is beliefs mistaken of prevalence the to response One 01Vl 1 o 5e1912440117 15 No. 118 Vol. 2021 isadsineseigys essetaddfcl to difficult and persistent poli- so in seemingly issues science contentious and about tics misperceptions are hy | akr effect backfire a,1 | misinformation oiia Behavior Political | aenews fake | which ulse pi ,2021. 9, April Published Editorial the by invited 1 editor guest the a under Published is A.H. Submission. Direct PNAS Board. a is article This interest.y competing no declares author The paper.y the wrote B.N. contributions: Author eodnso otpeettosaeaalbeo h A est at website NAS the on in available Humanities, & are Sciences, presentations video Arts, nasonline.org/misinformation and most program the complete Dame of The for Sackler. The M. recordings Foundation from Arthur husband, gift Sackler Sackler’s generous Dame M. of a memory Arthur by Dr. Mabel supported CA. February were and and From Irvine, colloquia 1995. Jillian Arnold since in 2019, PNAS the May Engineering in published at through and been have 2001 2019, Sciences and of 1991 3–4, in Academies began April colloquia National NAS held the Sphere,” of of Public Academy Center National Beckman the the in of Misinformation Science Colloquium Communication: About Sackler Science M. of Practice and Arthur Science the the “Advancing Sciences, from results paper This of 66% Conversely, election. presidential Donald help 2016 to of the stories 29% news win fake only Trump spread and but conclusion created community Democrats Russia intelligence that of the United 82% with the agreed instance, in Republicans poll be for 2018 group (8, March (10), to people’s a world States In with the seems systems. correlated belief around frequently and claims and identities are belief (7) factual beliefs that States These unsupported United 9). show the or surveys in time, false widespread over salient prevalence mea- in or their misperceptions of of census sures systematic a lack we Although Misperceptions salient of Importance to and Prevalence them The linking or identities. claims group false discouraging and political notably, promoting effective change—most from more belief elites be durable might creating suggests at evidence the communication documenting rarely that by beliefs approaches we conclude mistaken I result, However, in time. a reduction over systematic As and exposure. consistent correction rapid observe and after misconceptions effects hold who decay people reaching in In problems checks targeting issues. fact people substantial controversial revealed incline has about research to beliefs addition, flows tends congenial information partisanship holding elite like toward of a heuristics interaction to simple durably The and attributable that mind. be information their corrective to with changes likely people reach more to of is failure but persistence argue, The modest I accuracy. generates misperceptions, belief typically in expo- improvements that information significant consensus corrective emerging the to in summarize sure findings then the I how outstripped study. show quickly the first article prevalence I our article, the of this interpretations In explain misperceptions. support effects of not durability backfire conducted—does and that have interpretation I and the Reifler that research results these why below, for pre- explanation show persistent. primary so I were are the misperceptions As as that interpreted claim. groups frequently the were among believe misperception to disposed targeted belief the increased actually in exposure correction which in effect” fire mi:[email protected]. y Email: oee,tesinicltrtr—nldn subsequent literature—including scientific the However, y NSlicense.y PNAS about https://doi.org/10.1073/pnas.1912440117 science .y http://www. | f7 of 1

POLITICAL SCIENCES COLLOQUIUM PAPER Democrats endorsed the politically congenial but unsupported One potential explanation for the persistence of these mis- claim that Russia tampered with vote tallies compared with just perceptions is that the media frequently fails to aggressively 18% of Republicans. Internationally, one recent global survey of fact-check false statements by political elites or resorts to neu- 100 countries found that 32% of people who have heard of the tral “he said,” “she said” coverage of factual disputes (28, 29). To Holocaust think that it is a myth or greatly exaggerated, including examine this hypothesis, Reifler and I conducted a series of five 63% in the Middle East and North Africa and 64% of Muslims in experiments in 2005 and 2006 with undergraduates at a Catholic the region (11, 12). Similarly, supporters of populist parties and university in the Midwest (6). In each one, respondents were leaders are less likely to endorse the scientific consensus on the asked to read a realistic mock news article in which a prominent threat posed by climate change than other people in Europe and political figure made a dubious claim about a controversial issue, the United States (13). such as the war in Iraq or stem cell research. These articles were Why do people so frequently believe these false claims? experimentally manipulated so that some included a passage Research conducted to date suggests that cognitive and memory debunking the statement by the elite in question. Although it is limitations, directional motivations to defend or support some difficult to isolate the effects of directional motivations (30), the group identity or existing belief, and messages from other people evidence that we observed was consistent with such an account— and political elites all play critical roles in the spread of misin- in three of five experiments, exposure to corrective information formation (2, 14, 15). These factors may be especially difficult to was most effective among respondents who found the correction overcome in the contemporary period, which combines historic to be ideologically congenial (e.g., liberals for a correction of levels of political polarization in the United States with commu- George W. Bush on Iraq). In two of the studies, the estimated nication technology that allows false information to move farther marginal effect of exposure to corrective information was sig- and faster than ever before (16, 17). Even though evidence indi- nificant in the opposite of the expected direction—a so-called cates that most people are not trapped in “echo chambers” of backfire effect. However, this effect was not observed in the other like-minded information (18), misinformation may still be more three studies. likely to outrun society’s defenses and to be integrated into Based on these results, we concluded that “corrective informa- people’s belief systems under these circumstances. tion in news reports may fail to reduce misperceptions and can Misperceptions can also play an important role in debates sometimes increase them for the ideological group most likely over public policy. One of the most well-known examples is to hold those misperceptions” and called for further research climate change, where the United States is an outlier both on how to most effectively reduce misperceptions. In the years in the proportion of the population that believes that human since the study’s publication, the results have frequently been activity is its primary cause and in its support for measures misinterpreted as showing that that all corrections are coun- to address the crisis (19, 20). Although definitively establish- terproductive or that backfire effects are the primary cause of ing the relationship between misperception belief and national the persistence of misperceptions (our findings do not support climate policy is difficult, research indicates that this asso- either claim). ABC News, for instance, summarized our find- ciation holds at the individual level and that experimentally ings as follows: “when we encounter facts that contradict those induced changes in belief in climate change are associated beliefs, the facts are either ignored or twisted to support our with greater support for policy action (21). More recently, the positions” (31). debate over end of life care was upended by the “death panel” myth, a false claim popularized by former Alaska governor Revising the Record on Backfire Sarah Palin (22). After this claim became widespread, a provi- Our initial backfire study has often been interpreted to mean that sion was withdrawn from the Affordable Care Act that would these effects are widespread. However, subsequent research sug- have provided Medicare coverage for voluntary meetings with gests that backfire effects are extremely rare in practice. Most doctors to discuss end of life care options (a proposal that notably, an extensive replication and extension study conducted previously attracted bipartisan support). A subsequent regula- by other researchers found that no evidence of backfire effects tion to cover these meetings was again withdrawn after 2011 in response to corrective information across numerous experi- for fear of further controversy before being finally instituted in ments (32). Reifler and I collaborated with those authors in a 2015 (23, 24). subsequent study, which similarly found that exposure to correc- tive information contradicting a statement by President Trump The Backfire Effect: Findings vs. Interpretations reduced misperceptions about the prevalence of crime regardless The durability of misperceptions is especially concerning. Sur- of which candidate respondents supported (33). vey evidence indicates that false beliefs about high-profile issues Subsequent studies have shown that corrective information often persist for years or decades despite extensive efforts can also increase belief accuracy when conveyed in fact-checking by journalists, scientists, and public officials to set the record articles as well as novel presentation formats, such as graph- straight. In the aftermath of the US invasion of Iraq in 2003, ics, corrections that provide alternate causal explanations, and for example, the American government concluded that Saddam fact-checking labels on social media platforms (34–36). These Hussein’s government did not possess significant weapons of findings are consistent with meta-analyses showing that cor- mass destruction (WMDs) and was not conducting an active rections are moderately effective in improving the accuracy WMD program (25). These findings were widely publicized at of people’s factual beliefs, although effects are reduced when the time and in the years since the invasion. However, a 2015 poll the information provided concerns real-world politics and may found that 42% of Americans, including 51% of Republicans, be inflated by publication bias (37, 38). Notably, exposure to still believed that US troops found WMDs in Iraq (26). A similar information about the scientific consensus supporting anthro- pattern was observed in polling about Barack Obama’s religion. pogenic climate change leads to greater expressed belief in Even though the former president was an avowed Christian who these facts (21). Similarly, recent research has found that peo- regularly went to church, rumors circulated widely that he was ple’s self-awareness of their (lack of) knowledge is greater than a secret Muslim. Despite all of the disconfirming evidence that earlier research indicated, although the least informed peo- was available online and in media coverage, a poll conducted in ple are most likely to overestimate their performance (39–41). 2015, the seventh year of Obama’s presidency, found that only This meta-awareness of one’s lack of knowledge would simi- 39% of Americans identified Obama as Christian and that 29% larly suggest that people can in some cases recognize what they of Americans (including 43% of Republicans) said that he was do not know when presented with contradictory or uncongenial Muslim (27). information.

2 of 7 | PNAS Nyhan https://doi.org/10.1073/pnas.1912440117 Why the backfire effect does not explain the durability of political misperceptions Downloaded by guest on September 26, 2021 Downloaded by guest on September 26, 2021 h h akr fetde o xli h uaiiyo oiia misperceptions political of durability the explain not does effect backfire the Why Nyhan om fbhvo nfilssc sfiac emt ayi a in vary to seem finance as such costly fields Similarly, greatest in (56). of behavior sincere misperceptions of largely the forms are often concern, are substantive which matters, about views more that such suggesting about measured, beliefs are we when disputes factual decrease incentives what salient to misperceptions are response from in incentives reported differs modestly accuracy only Moreover, that strong where absent. manner world, strat- a typically real be guessing in the instead the in employ may observe changing they They or that true. effort egy be cognitive to necessarily more know not exerting they are finan- what respondents by studies increased that withholding prior is suggesting accuracy instance, incentives, whether For cial on unclear. evidence is mixed the findings find that these for mech- infer the anism However, in to responding. observe expressive often when reflect authors we beliefs that polarization factual the polarization partisan belief typi- of leading levels partisan studies expressed offered, of to These are levels answers 55). so-called incentives reduced correct 54, this find for isolate (46, people cally to questions response pay try knowledge to expressive to they factual is an approach that cheerleading indicate common views partisan A clearly minority express 53). that A to (52, or (51). willing rather disavow believe clearly trolling later are or sincerely true respondents they be of what to like indicating would surveys than they in what answers indicate providing that expressively, responding are dents Responding. all Expressive effort processing and factors. significant ability be political cognitive to from seem in failures cues and misperception-enhancing elites, of misinformation, consume role, flows who people major ongoing to a information, it corrective targeting play of below, in to effects problems argue the seem in I decay not over-time As do whereas evaluate responses persistence. to survey misperception conducted expressive on be not effects should are con- be research their as effects instead further should public of factors and other the causes sidered, (50), among directly about answer claims questions to easy false Although in whole? belief a more reduced translate not often into information are often corrective that mispercep- of results of studies encouraging in the persistence observed do the Why explain does? not what tions, do effects backfire If Persistence Misperception for Explanations Potential Other even countries between or polarization knowledge, 49). partisan 47, issue opinion (39, of prior encourage levels salience, in that issue differences iden- contexts on which to evidence, depend and more of may interpretations issues do motivated must than of this rather research change types on factual Future the vary 48). their 47, findings if tify 33, even although (21, change accurate, point not more opinions do become a As people’s sometimes beliefs views, and/or 46). issue one’s these (33, an reexamine like about learned to credi- processes willingness have the of dissonance-reducing they their lack in of that with distrust result information consistent expressing a is the by that of or manner bility (43–45) a in for views question responsibility political or in attitude-consistent blame facts an assigning the in by receive instance, they manner—for that inter- information but beliefs the their update pret often (42). respondents backlash opinions studies, than these in rather In groups changes ideological and parallel partisan update induces across to often exposure often that predisposi- will information indicate and factual studies misinformed Recent opinions are somewhat. beliefs who their their people to claims Second, false congenial accept tions. so are may opinions people they opin- cases, are because cause many why in necessarily polarization; not directly, ion do it misperceptions First, with polarized? presented when mation fpol ilotnudt hi eif ae nfculinfor- factual on based beliefs their update often will people If n oeta xlnto sta respon- that is explanation potential One aydet h s falwefr ersi nwihtuhis truth which in heuristic accu- low-effort their a of of exposure use perceptions prior the greater addition, to to In due leads racy 70). Cogni- headlines (69, news the ratings headlines to accuracy by news higher measured with false associated (as for are thinking test) that Reflection analytic indicate tive of studies Recent levels of information. failures lower corrective of pro- motivated of because than cessing rather misperceptions effort processing to and ability victim cognitive the fall Effort. Processing of often and may fraction Ability Cognitive per- large of Failures the a explain in help denial change might (68). allied public climate News) in of con- Fox misinformation people sistence congenial (e.g., most of outlets flows that media intense media corrective and mainstream to sume the exposure 67). sparse in con- (66, of the online information in combination and of the influential person example, or portion in For visible conducted small others especially inter- relatively with be politically versations a highly may of up but set for population makes percentage This preelec- individuals small population). the vanishingly the ested a Americans in of (vs. rest of diets (65) the 2016 diets information in news conservative period tion the most the of con- with 20% websites promote approximately Amer- By extreme—untrustworthy frequently stituted of (65). be which minority check websites, can a fact these website misperceptions, that the from untrustworthy information consume read an of icans also volume from checked the article fact contrast, Ameri- an been 3% of had the read only one that who of Moreover, the visited those websites. also half before fact-checking of website weeks than national are untrustworthy major the fewer who an the In visited people fact instance, who question. the for for cans to in audience election, matched the claims result, be 2016 the a to to As seem elites exposed 64). not political false (2, does by debunk media checks widely that social disseminated reports on news are or or often ongo- checks which receive fact claims, rarely exception prob- to people the exposure targeting controversies, With ing high-profile by information. few compounded a corrective of likely delivering are in lems effects decay and Information. Corrective unusually Targeting in an Failures in falsified been had 63). (62, by myth manner levels the immediately definitive prior though to increased was reverted even he but accordingly 2012 release that States certificate’s Belief birth United president. the after as the serve in long- to further born his eligibility providing his released Hawaii, of of Obama state proof the 2011, from April certificate birth negative In form with (61). among public resonated attitudes the that of racial myth members so- and a in the identifiers born created Republican not of which was States, claims Obama United Barack false the that the movement “birther” Consider called share 60). congenial who politically (22, elites promote that misperceptions from ideology public messages or the of partisanship par- of their flows their members ongoing with contrast, congenial receive By to are often identity. revert but that group to 59) views or tend 58, to respondents tisanship (48, or result, beliefs cases a prior some As their in time. weeks over fade for often last can (37). effects exposure misinformation after These immediately of measured Corrections prevalence are time beliefs elites. the when political even over diminish from partly dissipate cues only to by typically tendency exposure overwhelmed be the information to emphasize of or might effects beliefs the false for of durability Elites. Political the from of Cues and merely Effects not Decay are surveys belief, in in expressed (57). views differences expressive the partisan that sincere suggesting with consistent manner https://doi.org/10.1073/pnas.1912440117 hs lt messaging elite These natraeaccount alternate An ial,people Finally, PNAS | f7 of 3

POLITICAL SCIENCES COLLOQUIUM PAPER inferred from feelings of familiarity (71). Conversely, people change where a strong consensus of expert opinion and evidence who are encouraged to engage in deliberation are less likely exists (28). to believe false headlines than those who respond immediately, Instead, reporters and science communicators should empha- suggesting that greater cognitive effort helps respondents iden- size the views of nonpartisan experts. An ABC News story on the tify false claims. These findings raise important questions about death panel myth in 2009, for instance, only briefly mentioned the relative roles of analytic thinking and directionally motivated the partisan sources of the claim (77). Instead, the headline reasoning in misperception belief. One important factor may stated that “Experts debunk health care reform bill’s ‘death be the difference between news articles (as tested in the back- panel’ rule,” noting that “Doctors agree health bill has no ‘death fire effect research) and headlines. The latter, which are more panel’ requirement for the elderly.” The second paragraph of commonly encountered in the era of social media and which con- the story further noted the presence of surprising expert sources vey less information, may be especially likely to be processed contradicting the claim, stating that “even [experts] who do not using heuristics that require little cognitive effort or distorted support the version of the health care reform bill now being dis- via social transmission (72). By contrast, highly salient issues cussed” believe “these accusations are shocking, inflammatory and engaging stimuli may more commonly provoke respondents and incorrect.” to engage in effortful forms of processing, especially if heuris- tic cues suggest politically uncongenial conclusions that moti- Make Corrections Harder to Avoid or Deny. These strategies are vated respondents wish to avoid (5). The relationships between likely to prove most effective when applied to sources and levels of processing effort, analytical ability, and updating contexts in which people cannot easily avoid or disbelieve uncon- based on corrective information are important topics for future genial information. Partisans tend to diverge in how they view research. the state of the economy but to converge when it is unusu- ally strong or weak, which creates a reality that is hard for Alternate Strategies for Reducing Misperceptions either side to deny (44, 78). Similarly, most Americans have The findings described above suggest that fact checks and other been forced to recognize that extreme weather events are types of corrective information are at least somewhat effective. becoming more common, although they still diverge in whether Contrary to media coverage of the backfire effect, subsequent they believe that anthropogenic climate change is the primary research finds that people are often willing to revise mistaken cause (79). Messages drawing attention to potentially unwel- beliefs when given accurate information. However, these find- come facts like these may also be more credible when coming ings do not always cumulate into lasting reductions in many from unexpected or trusted sources like local weather forecasters prominent false beliefs. To better achieve this goal, scientists, (on climate change) (80) or internal tobacco company docu- journalists, and educators should pursue alternate communica- ments (on how smoking causes cancer). Finally, personal expe- tion approaches that research suggests might more effectively riences can be similarly powerful—many Americans attribute counter misinformation—minimizing false claims and partisan their changed views on gay marriage to personal experiences and ideological cues in discussion of factual disputes and high- with people that they know who identify as gay or lesbian lighting corrective information that is hard for people to avoid (81). It is likely that many people have also known someone or deny. with a smoking-related illness given the mortality and morbid- ity associated with tobacco use, which could contribute to the Minimize Elite Misinformation and Partisan Cues. Any effort to overwhelming consensus among the public that smoking causes counter misperceptions must first recognize the role that elites cancer (82). play in promoting false beliefs and linking them to people’s political identities. A case in point is climate change, an issue An Intermediary-Focused Approach to Fighting False Beliefs on which beliefs have polarized in a manner consistent with The strategies described above are not enough, however. Cor- a process of elite-led opinion leadership (73). Most notably, rective information seems to only rarely cause backfire effects belief polarization is highest among people who are the most among the public, but its effects are often modest, decay rela- politically attentive and thus most likely to receive and accept tively quickly, and fail to cumulate into sustained decreases in cues from elites who share their worldview (60). These mes- many common misperceptions. Fact checkers, journalists, and sages from partisan elites, which have seemingly become more science communicators should thus complement their public- polarized over time, are likely to diminish the effects of fact facing efforts with what I call an intermediary-focused approach checking either by encouraging directionally motivated reason- that targets the political elites who play a critical role in belief ing or by shaping the priors of group members in a manner and opinion formation. that reduces their responsiveness to corrective information (30, One important strategy is to increase the political costs of 74). Indeed, it is the people with the highest levels of science making false claims by sanctioning political elites who do so knowledge who are most polarized on climate change, which in a more salient and public manner. Elected officials are very suggests that the cues that these more sophisticated individu- responsive to the threat of negative news coverage (83–86). als receive about the beliefs held by the political group with Applying high-profile scrutiny to elite rhetoric can thus poten- which they identify trump the evidence that they might otherwise tially help to deter them from promoting misinformation. One consider (75). field experiment found that state legislators who were sent News coverage should, therefore, seek to avoid amplifying reminders of the reputational threat posed by fact checkers false claims and reduce the incidence of partisan and ideological in their state were less likely to make claims that were fact cues when discussing matters of fact and science—for instance, checked or whose accuracy was questioned publicly (87). There by resisting the habit of balancing messages from experts on are many potential ways of accomplishing this goal. For exam- issues where scientific consensus exists with citations to polariz- ple, providing fact-check statistics showing that a politician has ing opponents (68). Party cues have become more common, for repeatedly made false statements is more damaging to their instance, in media coverage of climate change (60, 74), which standing with the public than a fact check of a single false claim encourages people to side with adherents of their preferred (88). Conducting live fact checking during political program- party or ideology rather than evaluating the facts dispassion- ming, integrating fact checks into mainstream news coverage ately (76). These kinds of cues may be especially common in rather than relegating it to specialized websites and sidebar “balanced” reporting, which tends to overrepresent the preva- articles, and prominently featuring fact checks in online search lence of unsupported perspectives in factual disputes like climate results could similarly increase the salience of fact checks and

4 of 7 | PNAS Nyhan https://doi.org/10.1073/pnas.1912440117 Why the backfire effect does not explain the durability of political misperceptions Downloaded by guest on September 26, 2021 Downloaded by guest on September 26, 2021 h h akr fetde o xli h uaiiyo oiia misperceptions political of durability the explain not does effect backfire the Why Nyhan ecieapiain fGOtcnlg htaeappealing are that technology could GMO who professionals of and applications activists could as describe sources well as credible (95) liberal more and sympathies credibility scientific These with effec- the commentators speak concerns. public include could on liberal who polarization parties to third belief tively amplify to preventing seek and would issue GMOs and about facts tions of have summaries may than result beliefs a as on evidence. will and effects peo- elites messages durable pose—to like-minded Such more they from worldviews. come that ideological often risks and scientifically the partisan GMOs—including elites ple’s about effectively to political claims opposition not and fundamentally, link activists unsupported may Most to from public seek messages issue. of that the the flow to the of counter directly revert views people targeted as prior messages moreover, quickly, their dissipate does toward to safety audi- likely GMO websites). are about readers to information fact-checking have (e.g., corrective persuasive and that misperceptions and effects hold publications Any by to science-oriented are likely received less messages of be and are these to that unpersuasive change, ences likely often climate is most with information often As this targeted. but directly poorly public, evidence scientific the pro- and to to factual they efforts communicating notably, other on Most above. focus to described limitations information similar corrective eating vide from of suffer to risks seem the about claims become foods. unsupported initiatives GMO GMO-labeling with mandatory cli- bundled too for so to existence, support its measures of associated denial proposed may with also bundled to often but was opposition change issues often as mate policy extension Just on conflict claims. positions of factual just process public not the in changes practice, includes encour- corresponding In and make (94). parties to among policy the elites separate by process on previously efforts divide similar aging not public a did widening that a induce issues creating then identifiers, can party attentive among elites polarization of and process activists This party. similar opposition opposition adopt the offi- countervailing among generate to elected often among they positions activists elites, allied consensus and become other activists to views politicized motivates these issue After turn (93). be extension” new views in “conflict could a of which promote issue process cials, activists the a which through that lib- States in plausible United and is the Democrats in It GMO-labeling among of (92). more concentrated support them erals heavily in process link Activism is a and policy. initiatives to public misperceptions GMOs vulnerable to these of seems closely safety deepen relatively parti- quo the could are between about status that doubts views divide the in these consistent However, States but no (91). United the is (90), ques- in there eat States sans consensus and to United scientific salience safe the the low despite are in foods they partisan GMO public that of along the safety polarized of the on tions become majority issue The could an the (GMOs), lines. beliefs Consider organisms factual to science. modified which party or genetically a of evidence inside issue relevant estab- credibility giving possess the become by who communicate claims example, stakeholders messages—for false to group voice which United in by components the lished groups in ideological born and not was Obama the among Barack beliefs disavowed (89). false States that finally reduce help public Trump to In the seemed Donald it. which instance, repudiate myth, to birther for forced 2016, are politicians fall that become may costly claim repu- discredited sufficiently politicians’ a on endorsing cases, inflict some could In tations. they that damage potential the nitreir-oue praht oneigmispercep- countering to approach intermediary-focused An GMOs about misperceptions reducing to approaches Current parties within process the disrupt to valuable be also would It ino e ok atot olg,adteGrl .Fr colof Lyons, Ben School comments. Guess, Ford helpful Andy for R. thank Wood Tom I Gerald and Michigan. Reifler, the of Jason Porter, and University Ethan the College, at Dartmouth Policy York, Public New of tion ACKNOWLEDGMENTS. to vulnerable data. are that or Availability. intermediaries Data hold the who on people journal- more to require beliefs. and false credible turn public most in communicating the are on will less to so focus to directly Doing communicators sources science trust. the and from ists most often receive they people will accuracy that misperceptions that belief cues reducing the improve durably changing is but require often belief margin, information can which the corrective on and for coun- Providing issues worthwhile the to on common. generally given difficult identity is particularly find group strong polarization crisis), and of unusually over- economic of effects absent tervailing an to aware opinion (e.g., struggles become Even public deny checking people public. of that fact the inertia evidence rare, to the are flows come effects information issues— backfire controversial of if other and configuration politics the about beliefs false ing increased couples heterosexual (81). 64% parents as to same-sex parents 54% that from good beliefs while equally 2013, be in can 46% to from 2003 declined in family 56% American marriage traditional gay the that undermine Pub- beliefs would poses. shifted; correspondingly it has that opinion harm it, lic the to Republican about opposition claims in national unsupported messages threat including popular, their social abandoned increasingly no largely posed politicians becoming marriage coalition was gay a that and instance, in mounted For position. evidence emerge consensus as previously fissures by a disavow when created rare, elites and seem reverse security may national can voices to polarization these leaders threats While military disruptions. the former climate climate-related or discuss how businesses could about their speak who farm- affecting could Republican-leaning is who leaders change as advo- corporate such such and however, Republican More ers needed, activist. former climate are a turned cates Inglis, Congress Bob of and evan- member an scientist, Hayhoe, climate Katharine gelical include examples climate Notable anthropogenic doubt change. frequently with members worldviews or whose identities groups share who seek voices instead credible might amplify misperceptions to reduce for polarization to belief Efforts change, scientific issue. reduce the climate the substantially on of emphasizing to failed issue messaging when have the and consensus polarization On checks belief hold. fact corresponding taken counter instance, a already to but has taken issue, it be an could on approach emerging from polarization as perceived are who scientists than liberal.) overwhelmingly Party movement conservative Republican the the in counter- and originating in claims credible GMO more false the with be ing on will originates offered businesses be pro- it and can or right—farmers account if partisan, similar effective ideological, (A communities. their most fessional any within be from However, sources will concerns. credible checking reputational politi- fact to like present such are, sensitive accurately who often intermediaries to also cians, these incentives would among compared It the science (96). crops strengthen GMO views GMO to prior his of valuable omitted benefits be that the account an about with strength argument the a his of about perceptions of narrative increased conversion activist a anti-GMO to popula- former exposure vulnerable crop instance, for in supply For food damage tions). the environmental increasing reducing or (e.g., production values liberal to htti prahhglgt stekydnmci counter- in dynamic key the is highlights approach this What belief prevent to how describe strategies above the of All hsatcede o s ricueoriginal include or use not does article This cnweg upr rmteCrei Corpora- Carnegie the from support acknowledge I https://doi.org/10.1073/pnas.1912440117 PNAS | f7 of 5

POLITICAL SCIENCES COLLOQUIUM PAPER 1. C. Funk, L. Rainie, Public and scientists’ views on science and society. Pew Research 31. ABC News, The conversation: Why facts don’t matter. 19 July 2010. https://abcnews. Center (2015). https://www.pewresearch.org/internet/wp-content/uploads/sites/9/ go.com/WN/facts-matter-world-news-conversation/story?id=11196924. Accessed 10 2015/01/PI ScienceandSociety Report 012915.pdf. Accessed 6 September 2019. September 2019. 2. D. J. Flynn, B. Nyhan, J. Reifler, The nature and origins of misperceptions: Under- 32. T. Wood, E. Porter, The elusive backfire effect: Mass attitudes’ steadfast factual standing false and unsupported beliefs about politics. Adv. Polit. Psychol. 38 (suppl. adherence. Polit. Behav. 41, 135–163 (2019). 1), 127–150 (2017). 33. B. Nyhan, E. Porter, J. Reifler, T. J. Wood, Taking fact-checks literally but not seriously? 3. National Academies of Sciences, Engineering, and Medicine, Communicating Science The effects of journalistic fact-checking on factual beliefs and candidate favorability. Effectively: A Research Agenda (National Academies Press, 2017). Polit. Behav., https://doi.org/10.1007/s11109-019-09528-x (2019). 4. Z. Kunda, The case for motivated reasoning. Psychol. Bull. 108, 480–498 (1990). 34. K. Clayton et al., Real solutions for fake news? Measuring the effectiveness of general 5. D. M. Kahan, E. Peters, E. C. Dawson, P. Slovic, Motivated numeracy and enlightened warnings and fact-check tags in reducing belief in false stories on social media. Polit. self-government. Behav. Public Policy 1, 54–86 (2017). Behav., https://doi.org/10.1007/s11109-019-09533-0 (2019). 6. B. Nyhan, J. Reifler, When corrections fail: The persistence of political misperceptions. 35. B. Nyhan, J. Reifler, Displacing misinformation about events: An experimental test of Polit. Behav. 32, 303–330 (2010). causal corrections. J. Exp. Polit. Sci. 2, 81–93 (2015). 7. K. Frankovic, Belief in conspiracies largely depends on political identity. YouGov, 27 36. B. Nyhan, J. Reifler, The roles of information deficits and identity threat in the December 2016. https://today.yougov.com/topics/politics/articles-reports/2016/12/27/ prevalence of misperceptions. J. Elect. Public Opin. Parties 29, 222–244 (2019). belief-conspiracies-largely-depends-political-iden. Accessed 18 December 2018. 37. M.-p. S. Chan, C. R. Jones, K. H. Jamieson, D. Albarracin, Debunking: A meta-analysis 8. M. A. Gentzkow, J. M. Shapiro, Media, education and anti-Americanism in the Muslim of the psychological efficacy of messages countering misinformation. Psychol. Sci. 28, world. J. Econ. Perspect. 18, 117–133 (2004). 1531–1546 (2017). 9. B. Duffy, The Perils of Perception: Why We’re Wrong about Nearly Everything 38. N. Walter, S. T. Murphy, How to unring the bell: A meta-analytic approach to (Atlantic Books, 2018). correction of misinformation. Commun. Monogr. 85, 423–441 (2018). 10. K. Frankovic, Russia’s impact on the election seen through partisan eyes. YouGov, 39. J. H. Kuklinski, P. J. Quirk, J. Jerit, D. Schwieder, R. F. Rich, Misinformation and the 9 March 2018. https://today.yougov.com/topics/politics/articles-reports/2018/03/09/ currency of democratic citizenship. J. Polit. 62, 790–816 (2000). russias-impact-election-seen-through-partisan-eyes. Accessed 4 September 2019. 40. M. H. Graham, Self-awareness of political knowledge. Polit. Behav. 42, 305–326 11. Anti-Defamation League, ADL Poll of Over 100 Countries Finds More Than One- (2018). Quarter of Those Surveyed Infected With Anti-Semitic Attitudes. (2014). https:// 41. I. G. Anson, Partisanship, political knowledge, and the Dunning-Kruger effect. Polit. www.adl.org/news/press-releases/adl-global-100-poll. Accessed 27 March 2020. Psychol. 39, 1173–1192 (2018). 12. Anti-Defamation League, New ADL Poll Finds Dramatic Decline in Anti-Semitic Atti- 42. A. Guess, A. Coppock, Does counter-attitudinal information cause backlash? Results tudes in France; Significant Drops in Germany and Belgium. (2015). https://www. from three large survey experiments. Br. J. Polit. Sci., 10.1017/S0007123418000327 adl.org/news/press-releases/new-poll-anti-semitic-attitudes-19-countries. Accessed 27 (2018). March 2020. 43. B. J. Gaines, J. H. Kuklinski, P. J. Quirk, B. Peyton, J. Verkuilen, Same facts, different 13. J. Poushter, C. Huang, Climate change still seen as the top global threat, but interpretations: Partisan motivation and opinion on Iraq. J. Polit. 69, 957–974 (2007). cyberattacks a rising concern. Pew Research Center, February 10 (2019). https://www. 44. M. Bisgaard, Bias will find a way: Economic perceptions, attributions of blame, and pewresearch.org/global/wp-content/uploads/sites/2/2019/02/Pew-Research-Center partisan-motivated reasoning during crisis. J. Polit. 77, 849–860 (2015). Global-Threats-2018-Report 2019-02-10.pdf. Accessed 6 September 2019. 45. M. Bisgaard, How getting the facts right can fuel partisan-motivated reasoning. Am. 14. S. Lewandowsky, U. K. H. Ecker, C. M. Seifert, N. Schwarz, J. Cook, Misinformation J. Polit. Sci. 63, 824–839 (2019). and its correction: Continued influence and successful debiasing. Psychol. Sci. Publ. 46. K. Khanna, G. Sood, Motivated responding in studies of factual learning. Polit. Behav. Interest 13, 106–131 (2012). 40, 79–101 (2018). 15. D. A. Scheufele, N. M. Krause, Science audiences, misinformation, and fake news. 47. M. J. Aird, U. K. H. Ecker, B. Swire, A. J. Berinsky, S. Lewandowsky, Does truth matter Proc. Natl. Acad. Sci. U.S.A. 116, 7662–7669 (2019). to voters? The effects of correcting political misinformation in an Australian sample. 16. S. Iyengar, D. S. Massey, Scientific communication in a post-truth society. Proc. Natl. R. Soc. Open Sci. 5, 180593 (2018). Acad. Sci. U.S.A. 116, 7656–7661 (2019). 48. C. M. Dowling, M. Henderson, M. G. Miller, Knowledge persists, opinions drift: Learn- 17. B. Nyhan, Social Media, Political Polarization, and Political Disinformation: A Review ing and opinion change in a three-wave panel experiment. Am. Polit. Res. 48, 263–274 of the Scientific Literature. (2018). https://www.hewlett.org/wp-content/uploads/ (2020). 2018/03/Social-Media-Political-Polarization-and-Political-Disinformation-Literature- 49. J. Sides, Stories or science? Facts, frames, and policy attitudes. Am. Polit. Res. 44, Review.pdf. Accessed 27 March 2020. 387–414 (2016). 18. A. Guess, B. Lyons, B. Nyhan, J. Reifler, Avoiding the echo chamber about echo cham- 50. P. W. Holland, Statistics and causal inference. J. Am. Stat. Assoc. 81, 945–960 (1986). bers: Why selective exposure to like-minded political news is less prevalent than you 51. J. G. Bullock, G. Lenz, Partisan bias in surveys. Annu. Rev. Polit. Sci. 22, 325–342 (2019). think. Knight Foundation report, February 12 (2018). https://kf-site-production.s3. 52. J. Lopez, D. Sunshine Hillygus, Why so serious?: Survey trolls and misinformation. amazonaws.com/media elements/files/000/000/133/original/Topos KF White-Paper http://dx.doi.org/10.2139/ssrn.3131087 (14 March 2018). Nyhan V1.pdf. Accessed 23 October 2018. 53. B. F. Schaffner, S. Luks, Misinformation or expressive responding? What an inaugura- 19. M. Roppolo, Americans more skeptical of climate change than others in global survey. tion crowd can tell us about the source of political misinformation in surveys. Publ. CBS News, 23 July 2014. https://www.cbsnews.com/news/americans-more-skeptical-of- Opin. Quart. 82, 135–147 (2018). climate-change-than-others-in-global-survey/. Accessed 10 September 2019. 54. J. G. Bullock, A. S. Gerber, S. J. Hill, G. Huber, Partisan bias in factual beliefs about 20. Reuters, U.S. remains outlier as G20 split over tackling climate change. 29 June politics. Quart. J. Polit. Sci. 10, 519–578 (2015). 2019. https://www.reuters.com/article/us-g20-summit-climate/us-remains-outlier- 55. M. Prior, G. Sood, K. Khanna, You cannot be serious: The impact of accuracy incentives as-g20-split-over-tackling-climate-change-idUSKCN1TU0DQ. Accessed 10 September on partisan bias in reports of economic perceptions. Quart. J. Polit. Sci. 10, 489–518 2019. (2015). 21. S. L. van der Linden, A. A. Leiserowitz, G. D. Feinberg, E. W. Maibach, The scientific 56. A. J. Berinsky, Telling the truth about believing the lies? Evidence for the limited consensus on climate change as a gateway belief: Experimental evidence. PLoS One prevalence of expressive survey responding. J. Polit. 80, 211–224 (2018). 10, e0118489 (2015). 57. M. Babajide Wintoki, Yaoyi. Xi, Partisan Bias in Fund Portfolios. (2019). https://www. 22. B. Nyhan, Why the ‘death panel’ myth won’t die: Misinformation in the health care cambridge.org/core/journals/journal-of-financial-and-quantitative-analysis/article/ reform debate. Forum 8, 5 (2010). partisan-bias-in-fund-portfolios/CD950BE17F7D19F6C763B2C6C6BC1D6A. Accessed 23. R. Pear, U.S. alters rule on paying for end-of-life planning. New York Times, 4 January 27 March 2020. 2011. https://www.nytimes.com/2011/01/05/health/policy/05health.html. Accessed 10 58. A. J. Berinsky. Rumors and health care reform: Experiments in political misinforma- September 2019. tion. Br. J. Polit. Sci. 47, 241–262 (2017). 24. R. Pear, New Medicare rule authorizes ‘end-of-life’ consultations. New York 59. B. Swire, A. J. Berinsky, S. Lewandowsky, U. K. H. Ecker, Processing political misin- Times, 30 October 2015. https://www.nytimes.com/2015/10/31/us/new-medicare-rule- formation: Comprehending the trump phenomenon. R. Soc. Open Sci. 4, 160802 authorizes-end-of-life-consultations.html. Accessed 10 September 2019. (2017). 25. C. Duelfer, Comprehensive Report of the Special Advisor to the DCI on Iraq’s 60. M. Tesler, Elite domination of public doubts about climate change (not evolution). WMD. (2005) https://www.cia.gov/library/reports/general-reports-1/iraq wmd 2004/. Polit. Commun. 35, 306–326 (2018). Accessed 27 March 2020. 61. A. Jardina, M. Traugott, The genesis of the birther rumor: Partisanship, racial 26. K. Breitman, Poll: Half of Republicans still believe WMDs found in Iraq. Politico, attitudes, and political knowledge. J. Race Ethn. Polit. 4, 60–80 (2019). 7 January 2015. https://www.politico.com/story/2015/01/poll-republicans-wmds-iraq- 62. A. Berinsky, The birthers are back. YouGov, 3 February 2012. Accessed 5 September 114016. Accessed 10 September 2015. 2019. 27. S. P. Bailey, A startling number of Americans still believe President Obama is a 63. A. Berinsky, The birthers are (still) back. YouGov, 11 July 2012. https://today.yougov. Muslim. CNN, 14 September 2015. https://www.washingtonpost.com/news/acts-of- com/topics/politics/articles-reports/2012/07/11/birthers-are-still-back. Accessed 5 faith/wp/2015/09/14/a-startling-number-of-americans-still-believe-president-obama- September 2019. is-a-muslim/?utm term=.78a9322f7dbf. Accessed 10 September 2019. 64. S. Vosoughi, D. Roy, S. Aral, The spread of true and false news online. Science 359, 28. M. T. Boykoff, J. M. Boykoff, Balance as bias: Global warming and the US prestige 1146–1151, 2018. press. Global Environ. Change 14, 125–136 (2004). 65. A. Guess, B. Nyhan, J. Reifler, Exposure to untrustworthy websites in the 2016 US 29. B. Fritz, B. Keefer, B. Nyhan, All the President’s Spin: George W. Bush, the Media, and election. Nat. Hum. Behav., https://doi.org/10.1038/s41562-020-0833-x (2020). the Truth (Touchstone Books, 2004). 66. J. N. Druckman, M. S. Levendusky, A. McLain, No need to watch: How the effects of 30. J. N. Druckman, M. C. McGrath, The evidence for motivated reasoning in climate partisan media can spread via interpersonal discussions. Am. J. Polit. Sci. 62, 99–112 change preference formation. Nat. Clim. Change 9, 111–119 (2019). (2018).

6 of 7 | PNAS Nyhan https://doi.org/10.1073/pnas.1912440117 Why the backfire effect does not explain the durability of political misperceptions Downloaded by guest on September 26, 2021 Downloaded by guest on September 26, 2021 0 .Pnyok .G ad h al o aenw?Terlso bullshit of roles accuracy perceived The increases exposure news? Prior Rand, G. fake D. Cannon, for D. T. Pennycook, falls G. 71. Who Rand, G. D. is Pennycook, news fake G. partisan to 70. Susceptibility biased: not Lazy, Cov- Rand, Media G. D. Mass Pennycook, and G. Conflict 69. Balance, (News)Worthy? Experts Are Merkley, E. 68. 4 .Mrly .A tcl,Pryeie rmnfcue ob?Teinformational The doubt? manufactured or elites Party Stecula, A. D. Merkley, polarization E. and change 74. climate of politicization The Dunlap, E. R. McCright, interpersonal M. of A. consequences Informational 73. grapevine: the Through Carlson, N. T. 72. Ard A. Weeks, E. B. 67. 1 e eerhCne,Goigspotfrgymrig:Cagdmnsand minds Changed marriage: gay for support Growing Center, Research of influence Pew The experts: 81. climate dis- Local Zhao, weather X. Myers, see T. Maibach, Americans E. Bloodhart, Most B. poll: 80. inhibits) AP-NORC (and Fingerhut, facilitates H. reality Borenstein, How S. disagreement: 79. “death of bill’s Tides reform Parker-Stephen, care E. health 78. debunk Experts partisan Childs. shape D. cues Gever, party J. How culprits? Snow, as K. elites 77. Partisan Slothuus, R. Martin, problem. B. measurement the 76. and communication Climate-science Kahan, M. D. 75. h h akr fetde o xli h uaiiyo oiia misperceptions political of durability the explain not does effect backfire the Why Nyhan ffk news. fake of thinking. analytic and (2020). familiarity, overclaiming, receptivity, reasoning. motivated by than reasoning (2019). of 39–50 lack by explained better 2020. March 27 Accessed (2020). https://www.tandfonline.com/doi/full/10.1080/10584609. 2020.1713269. Consensus. Expert of erage persuasion. political (2017). and leadership, opinion otx fciaecag polarization. change climate of context 2001–2010. warming, global of (2011). views public’s American the in communication. political hnigdmgahc.2 ac 03 https://www.people-press.org/2013/03/ 2019. September 2013. 13 Accessed March 20 20/growing-support-for-gay-marriage-changed-minds-and-changing-demographics/. demographics. changing perceptions. change (2015). climate on information weather TV local 2019. September 12 Accessed 682f099a67ae43d0876b044f9f7f5b26. worsening. asters opinion. public the partisan for requirement September panel” “death 12 no Accessed has 2019. https://abcnews.go.com/print?id=8295708. bill (2009). health agree elderly. Doctors rule: panel” gaps. perceptual Psychol. –3(2015). 1–43 36, .Ep sco.Gen. Psychol. Exp. J. m .Plt Sci. Polit. J. Am. vlAru .Gld Z de Gil H. evol-Abreu, ` etme 09 https://apnews.com/ 2019. September 5 Press, Associated .Polit. J. m oi.Si Rev. Sci. Polit. Am. 0718 (2013). 1077–1088 75, 5–6 (2018). 456–469 62, 8518 (2018). 1865–1880 147, c.Commun. Sci. u ´ ia nieiflec?Sca ei use, media Social influence? Online niga, ˜ 2–3 (2019). 325–339 113, n.J ul pn Res. Opin. Publ. J. Int. 5–7 (2018). 258–274 40, LSOne PLoS oi.Q. Socio. .Pers. J. Cognition e0141526 10, 185–200 88, 214–239 29, 155–194 52, Polit. 188, 4 .Lm etrCameiwt rwagoa atcekdatabase. fact-check global a grow to ClaimReview time. better A real Lim, S. in 84. poll president real, the is Fact-checking evolution Rauch, or J. Bang 83. Big confident not Americans Press, Associated 82. 6 .A yn,A ael .Tlargd,K .Jmeo,Cneso esgsadatti- and messages Conversion Jamieson, H. K. Tallapragada, M. Hasell, A. Lyons, matters. A. B. concerns anti-GMO 96. of dismissal Tyson’s deGrasse Neil Why Klein, American the E. in extension’ ‘conflict 95. and polarization Party Carsey, M. T. conflict Layman, and C. Activists G. Cooperman, R. Herrera, 94. R. Green, C. J. Carsey, M. T. Layman, C. G. 93. too. science, with problem a have Democrats Haelle. T. 92. safety. food question labeling, GMO demand groups Consumer Gillam, C. 91. dead. from safe? far foods G.M.O. but Are diminished Brody, is J. Birtherism 90. lives. It Nyhan, B. Dropp, K. 89. U.S. on experiment field Agadjanian A elites: A. on fact-checking 88. of effect The Reifler, J. Nyhan, B. 87. Str D. Snyder, M. J. 86. Arnold, D. R. 85. i582/ cesd1 etme 2019. September 12 Accessed ai/588028/. https://www.theatlantic.com/magazine/archive/2019/06/fact-checking-donald-trump- 2019. September 13 Accessed evolution-ap-poll/. shows. uecag:Srn ruet,ntcsl signals. (2019). costly not arguments, Strong change: tude https://www.vox.com/2014/8/1/5954701/neil-degrasse-tyson-gmos- 2019. September 12 2014. Accessed dangerous-safe. August 1 electorate. politics. party American in extension 2019. September 12 Accessed with-science-too-107270?paginate=false. https://www.politico.com/magazine/story/2014/06/democrats-have-a-problem- 2014. 2019. Septem- ber 12 Accessed gmo-labeling-question-food-safety-idUSBRE82Q10820120327. https://www.reuters.com/article/us-usa-food/consumer-groups-demand- 2012. March September 2019. 12 Accessed nytimes.com/2018/04/23/well/eat/are-gmo-foods-safe.html. 2019. September 12 Accessed https://www.nytimes.com/2016/09/24/upshot/it-lives- birtherism-is-diminished-but-far-from-dead.html. 2016. September 23 Times, York politicians. of favorability and accuracy (2019). perceived 2053168019870351 on data checking legislators. state (2010). 355–408 118, 2013). Press, 2019. https://reporterslab.org/a-better-claimreview-to-grow-a-global- September 12 2019. Accessed fact-check-database/. April 18 Lab, 1Arl21.https://www.cbsnews.com/news/americans-big-bang- 2014. April 21 News, CBS m .Plt Sci. Polit. J. Am. oges h rs,adPltclAccountability Political and Press, the Congress, onigtePncho:Teefc fsmayfact- summary of effect The Pinocchios: the Counting al., et m .Plt Sci. Polit. J. Am. meg rs oeaeadpltclaccountability. political and coverage Press omberg, ¨ 8–0 (2002). 786–802 46, 2–4 (2015). 628–640 59, 3Arl2 08 https://www. 2018. 23 April 23 Times, York New m oi.Si Rev. Sci. Polit. Am. https://doi.org/10.1073/pnas.1912440117 ul nesad Sci. Understand. Publ. 2–4 (2010). 324–346 104, ue1 June 1 Magazine, Politico ue2019. June Atlantic, The PictnUniversity (Princeton PNAS ueReporter’s Duke .Plt Econ. Polit. J. e.Polit. Res. 320–338 28, 27 Reuters, | f7 of 7 New Vox, 6,

POLITICAL SCIENCES COLLOQUIUM PAPER