Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=rics20

Information, Communication & Society

ISSN: 1369-118X (Print) 1468-4462 (Online) Journal homepage: http://www.tandfonline.com/loi/rics20

I do not believe you: how providing a sourcecorrects health misperceptions across socialmedia platforms

Emily K. Vraga & Leticia Bode

To cite this article: Emily K. Vraga & Leticia Bode (2017): I do not believe you: how providing asource corrects health misperceptions across social media platforms, Information, Communication& Society, DOI: 10.1080/1369118X.2017.1313883

To link to this article: https://doi.org/10.1080/1369118X.2017.1313883

Published online: 19 Apr 2017.

Submit your article to this journal

Article views: 460

View related articles

View Crossmark data

Citing articles: 4 View citing articles

I do not believe you: how providing a source corrects healthmisperceptions across social media platformsEmily K. Vraga a and Leticia Bodeb

aDepartment of Communication, George Mason University, Fairfax, VA, USA; bCommunication, Culture, andTechnology, Georgetown University, Washington, DC, USA

ABSTRACTSocial media are often criticized as serving as a source ofmisinformation, but in this study we examine how they may alsofunction to correct misperceptions on an emerging health issue.We use an experimental design to consider social correction thatoccurs via peers, testing both the type of correction (i.e., whethera source is provided or not) and the platform on which thecorrection ocratcurs (i.e., Facebook versus Twitter). Our resultssuggest that a source is necessary to correct misperceptionsabout the causes of the Zika virus on both Facebook and Twitter,but the mechanism by which such correction occurs differs acrossplatforms. Implications for successful social media campaigns toaddress health misinformation are addressed.

ARTICLE HISTORYReceived 14 November 2016Accepted 24 March 2017

KEYWORDSMisinformation; Twitter;Facebook; social media;health

Concerns about misinformation on a range of health, scientific, and political issues are notnew, but scholars and practitioners are increasingly concerned about the role that newcommunication platforms play in facilitating the spread of misinformation (Garrett,2011; Lewandowsky, Ecker, Seifert, Schwarz, & Cook, 2012; Rojecki & Meraz, 2016).Although misinformation can emerge from a variety of sources – ranging from deliberateattempts to manufacture or promote misinformation for political or financial gain to eliteor public misunderstandings of complicated issues – the reduced role of gatekeepers invetting quality information and increased opportunities to self-select content onlinehave both been suggested as reasons for additional concern about the potential for socialmedia to contribute to the spread and maintenance of misinformation (Lewandowskyet al., 2012; Nyhan, 2010; Radzikowski et al., 2016; Rojecki & Meraz, 2016; Sharma,Yadav, Yadav, & Ferdinand, 2016).

While social media can also serve as a source of corrective information to limit misper-ceptions on emerging health issues (Bode & Vraga, 2015), existing research is limited inthree ways. First, it only focuses on corrective information that comes from a platform-based algorithm (Facebook’s related stories function). In this paper, we consider correc-tion that comes from social sources, given that peers are a primary source of informationon social media. Second, existing research combines the correction with the provision of asource corroborating the correction. For this reason, we cannot be sure whether it is the

© 2017 Informa UK Limited, trading as Taylor & Francis Group

CONTACT Emily K. Vraga evraga@gmu.edu Department of Communication, George Mason University, 4400University Dr., MS 3D6, Fairfax, VA 22030, USA

INFORMATION, COMMUNICATION & SOCIETY, 2017http://dx.doi.org/10.1080/1369118X.2017.1313883

corrective information, the credibility that goes along with providing a source, or the com-bination of the two that effectively corrects misperceptions. Finally, existing research con-siders a single social media platform – Facebook – in isolation. Because we know that thecompeting social media platforms offer different affordances (Treem & Leonardi, 2012)and may therefore be perceived differently by their users, we think it important to considerwhether these mechanisms vary by platform as well. Here, we investigate two social plat-forms commonly used for news and information – and often implicated in the spread ofmisinformation – Facebook versus Twitter (Dredze, Broniatowski, & Hilyard, 2016;Mitchell, Gottfried, Barthel, & Shearer, 2016; Radzikowski et al., 2016; Sharma et al., 2016).

We test these mechanisms using the emerging health issue of the causes of the spread ofthe Zika virus in the Americas. In 2015, Zika started garnering international attentionwhen it was linked to microcephaly in infants, leading the World Health Organizationto declare it a Public Health Emergency of International Concern in February of 2016(2016). But as worldwide concern about Zika has increased, so has the spread of misinfor-mation about its causes and effects (e.g., Al-Qahtani, Nazir, Al-Anazi, Rubino, & Al-Ahdal, 2016; Griffiths, 2016; Worth, 2016). For example, one study found that posts con-taining misinformation about Zika are more popular on Facebook than are correct posts(Sharma et al., 2016), while another found that pseudo-scientific claims about the virusrise in tandem with public interest on Twitter (Dredze et al., 2016). The level of attentionand potential for misinformation on this issue on social media makes this an appropriatetest case for our questions.

This study moves beyond previous research to test how social correction – or correctionthat occurs via one’s social contacts – may be effective in reducing misperceptions aboutthe causes of the spread of Zika on two social media platforms: Facebook and Twitter.We use an experimental design in which we manipulate the type of social correction(i.e., whether a source is provided or not as compared to a control condition with no mis-information) and the platform on which the misinformation and correction occur (i.e.,Facebook versus Twitter). We expect that social corrections that provide an authoritativesource to substantiate the correction will be more effective than social corrections without asource, but we are uncertain as to whether social corrections absent a source is at all effec-tive compared to a control condition, or whether this process will differ on Facebook versusTwitter. This has major implications for recommendations to social media users for whenthey encounter misinformation – should they attempt to correct misinformation whereverthey encounter it, and if so, do they need to find corroborating evidence in order to do so?

Literature review

Misinformation and correction on social media

Worries aboutmisinformation (e.g., false information) have been primarily focused on thedifficulty of correcting the misperceptions – which we define as individual ‘beliefs aboutfactual matters [that] are not supported by clear evidence and expert opinion’ (Nyhan& Reifler, 2010, p. 305) – that arise from acceptance of the misinformation. In therealm of health and science issues, these misperceptions are particularly problematicbecause they can prevent individuals from engaging in appropriate behaviors to mitigaterisk (Dixon & Clarke, 2013; Rosenbaum, 2014; Tafuri et al., 2014) or prevent the public

2 E. K. VRAGA AND L. BODE

from appropriately weighing policy choices designed to address these issues (Fowler &Margolis, 2014; Reedy, Wells, & Gastil, 2014).

Research has examined the conditions under which misperceptions can be effectivelycorrected and updated. Social media seems a promising avenue for correcting mispercep-tions for several reasons. First, social media websites limit selective exposure tendenciesthat often drive other media choices (Barberá, 2014; Barberá, Jost, Nagler, Tucker, & Bon-neau, 2015; Messing & Westwood, 2014). As a result, people should have more opportu-nities to see corrective information on social media than in their regular media diet,particularly if misinformation has become associated with a worldview that aligns withpartisan preferences for news – for example, as in the case of climate change (Feldman,Myers, Hmielowski, & Leiserowitz, 2014; McCright & Dunlap, 2011).

Of course, mere exposure to corrective information is often not enough to reduce mis-perceptions. When encountering information that disputes a deeply held belief – even ifthat belief is wrong – individuals are likely to use motivated reasoning to discredit or dis-miss the new information (Bode & Vraga, 2015; Jerit & Barabas, 2012; Taber & Lodge,2006). However, the social norms and pressures that apply when individuals use socialmedia may also override initial desires to engage in motivated reasoning and selectiveexposure (Diehl, Weeks, & gil de Zuniga, 2016; Messing & Westwood, 2014; Thorson &Wells, 2016), which might also render people more amenable to corrective informationin these social spaces. Previous research supports this argument, finding that providingcorrective information via the Facebook ‘related stories’ algorithm reduced misperceptionson the issue of genetically modified foods and health, although not on the more establishedissue of autism and vaccination (Bode & Vraga, 2015).

While such an effort is promising, there are several limitations in relying on algorithmiccorrection. First, Facebook has come under fire for changes to its algorithm that may actu-ally promote misinformation (Solon, 2016), and research has suggested that posts contain-ing misinformation about Zika are more popular on Facebook than are correct posts(Sharma et al., 2016). Second, people may trust machine heuristics like an algorithmmore than social connections they may not know very well (Bode & Vraga, 2015; Wester-man, Spence, & van der Heide, 2014). Third, algorithmic correction is not available on allsocial media. Twitter is commonly studied as a source of misinformation (Castillo, Men-doza, & Polete, 2011; Oyeyemi, Gabarron, & Wynn, 2014; Starbird, Maddock, Orand,Achterman, & Mason, 2014), but currently lacks an algorithmic response that could cor-rect misinformation as the only content generated by algorithms on Twitter is promotedtweets (i.e., advertising). It is for this reason that this study focuses on social correction, orcorrection that comes from other social media users rather than from the platform itself.

Social correction and providing evidence

Social correction is likely to differ from other methods of addressing misinformation in anumber of ways. Existing research on correcting misperceptions on a range of issue oftenuses an authoritative source to rebut misinformation (e.g., Ecker, Lewandowsky, Swire, &Chang, 2011; Nyhan & Reifler, 2010; Nyhan, Reifler, & Ubel, 2013). However, social cor-rection relies on contacts on a social media network that are not necessarily credible them-selves and instead are part of a large and unknown audience (Edwards, Edwards, Spence,& Shelton, 2014; Marwick & boyd, 2011; Mitchell et al., 2016). Social correction can occur,

INFORMATION, COMMUNICATION & SOCIETY 3

in different ways, including when another user offers a statement refuting the misinforma-tion but does not provide any evidence to support their claim, or when that user offers thesame statement but also provides a link to an expert source to bolster the refutation.

There are several reasons to expect that the latter case – providing a link to a crediblethird party – will be more effective in correcting misperceptions. First, expertise is oftenlinked to credibility, which should enhance the effectiveness of corrections that rely onexperts in refuting misinformation, especially when the experts are considered unbiasedactors (Bode & Vraga, 2015; Petty & Brinol, 2008; Slater & Rouner, 1996). We expectthe external sources provided in this study – a link to the Centers for Disease Control(CDC) and to Snopes.com – should have more credibility than the anonymous othersocial media user (Edwards et al., 2014; Jahng & Littau, 2016; Mitchell et al., 2016). Second,outside of the realm of misinformation specifically, providing more evidence to support aclaim has been shown to facilitate persuasion, especially in situations where individuals arenot motivated to carefully process the information (Eagly & Chaiken, 1993; Petty & Brinol,2008; Petty & Cacioppo, 1984; Pornpitakpan, 2004). Third, offering a source to substanti-ate the correction of the misinformation can be seen in itself as a form of repeated correc-tion, as it provides a correction not only from the individual but also from the crediblethird-party sources (Lewandowsky et al., 2012).

For these reasons, we expect that social correction that includes a source will be themost effective form of correction by reducing misperceptions that may exist among thepopulation in general (e.g., compared to a control condition) as well as compared to socialcorrection that occurs without providing an external source. In addition, we expect thatthese social corrections that include a source will also be seen as more credible thanthose that do not provide a source:

H1: Social correction that provides a source will reduce misperceptions compared to a controlcondition.

H2: Social correction that provides a source will reduce misperceptions compared to socialcorrections without sources.

H3: Social correction that provides a source will be seen as more credible than social correc-tions without a source.

However, it is also worth examining how social refutation without providing a sourcecompares to a control condition in terms of correcting misperceptions that exist on thisissue in the aggregate. For an emerging issue, misinformation should be easier to address,which may suggest that a social correction may reduce misperceptions, even absent anauthoritative source (Bode & Vraga, 2015). Moreover, in this study the misinformationis immediately countered by two short, simple social corrections, which have beenshown to be effective in other domains (Bode & Vraga, in press; Lewandowsky et al.,2012). Finally, there has been a significant decline in trust in experts in scientific andhealth domains, often fueled by media choices and ideological beliefs (Gauchat, 2012;Hmielowski, Feldman, Myers, Leiserowitz, & Maibach, 2014; Kata, 2010), which may ren-der social corrections sufficiently credible to counter misinformation. However, this studyalso repeats the misinformation before providing the correction, which may serve toreinforce the misinformation if the correction is not strong enough (e.g., Berinsky,2017; Lewandowsky et al., 2012; Nyhan & Reifler, 2010). Because this difference has notbeen studied, it is considered here as a research question:

4 E. K. VRAGA AND L. BODE

RQ1: Will social correction that does not provide a source reduce misperceptions comparedto the control condition?

Platform differences

Existing research on corrections made within social media platforms has focused pri-marily on Facebook (Bode & Vraga, 2015). While Facebook is the most-used socialmedia platform in the United States (62% of adult Americans use the platform, seeDuggan, 2015), other platforms also experience high levels of use, and their affordancesmay change the way that misinformation or the correction thereof is perceived oraccepted by users. For this reason, in this study we compare Facebook to another fre-quently used platform, Twitter. Twitter is currently the fifth most-used social mediaplatform in the United States (behind Facebook, Pinterest, Instagram, and LinkedIn),with 20% of American adults using the service (Duggan, 2015). However, because ofvarious functionality issues, we think Twitter is the next most likely platform to pro-vide correction to misinformation. Pinterest and Instagram are primarily visual media,which make correction more difficult, and LinkedIn is primarily used for employmentreasons, which does not necessarily encourage discussion of controversial issues. Twit-ter also has the highest percentage of users (out of these five platforms) who reportgetting news through the platform, at 59% (Gottfried & Shearer, 2016). Finally, Twitterhas frequently been criticized for serving as a space where misinformation may flourish(e.g., Ehrenberg, 2012; Lewandowsky et al., 2012; Oyeyemi et al., 2014; Radzikowskiet al., 2016), making it important to consider whether it can also function to correctmisinformation.

Further, it is important to test differences in correcting misperceptions on Facebookversus Twitter, given that these platforms differ in a number of key ways. Scholars talkabout these differences in terms of social media ‘affordances’ – different aspects of socialmedia which transcend the individual platforms themselves (Ellison & Vitak, 2015; Gib-son, 1986). They are defined not only by technologists, who create a platform, but also byusers who may uncover new attributes of the platform (Schrock, 2015). The classic fouraffordances of social media are visibility (ease of posting, locating information), persist-ence (continuity of information after it is posted), editability (ability to craft and edit amessage), and association (connections between users and content) (Treem & Leonardi,2012; see also boyd, 2010). Facebook and Twitter have similar affordances in theseterms – the content on each is visible, easy to view, and generally persists after posting,but can be edited after the initial post is written.

However, the associations formed on Twitter differ from those formed on Facebook.On Facebook, associations require mutuality – in order to become friends, both partiesmust agree. On Twitter, the associations are one sided – I can choose to follow any publicaccount without consent required from the person I choose to follow. This represents afundamental difference in terms of the types of associations that are formed across plat-forms, which in turn may affect the way users perceive information from others, the waysin which they engage with the platform, and the norms that develop within that platform(Gruzd, Wellman, & Takhteyev, 2011) Therefore, on Twitter the imagined audience couldencompass the whole of Twitter, as compared to just a few hundred Facebook friends(Gruzd et al., 2011; Marwick & boyd, 2011).

INFORMATION, COMMUNICATION & SOCIETY 5

There are also technical functions that vary between Facebook and Twitter. Most nota-bly, Twitter first pioneered, and now uses conventions like the hashtag (#topic) muchmore than these conventions are used on Facebook (Bruns & Moe, 2014; Small, 2011),making it easier for like-minded communities to develop that extend beyond an individ-ual’s social connections (Bode, Hanna, Yang, & Shah, 2015). The greater use of hashtags,mentions, and hyperlinks on Twitter (Maireder & Ausserhofer, 2013) collectively producea ‘more inclusive mix of views (via emergent communicative norms), media objects (viahyperlinking), and actors (via @-mentioning)’ than other media platforms (Lyons, inpress). Additionally, Twitter pages are largely public, whereas Facebook posts are generallyavailable only to a chosen network. This produces more of a ‘collective network’ andencourages the sharing of community norms on Twitter (Gruzd et al., 2011).

Different affordances, more generally, can lead to different user experiences acrosssocial media platforms (Fox & Moreland, 2015; Mao, 2014; Thorson, 2014). It is thereforeeasy to imagine that the ways in which misinformation and corrective information are per-ceived on different platforms. Because we expect differences, but have no research onwhich to base expectations in this area, we pose the following research question:

RQ2: How will (a) correction and (b) evaluations of corrective information differ on Face-book compared to Twitter?

Finally, this study examines the mechanisms by which social correction may work toreduce misperceptions. Given the importance of credibility in successful persuasion(Petty & Cacioppo, 1984; Slater & Rouner, 1996), adding a source to social correctionmay mitigate misperceptions on the issue by first improving evaluations of the social cor-rection itself – and it is these heightened credibility perceptions that then reduce misper-ceptions. However, such a process might also depend on the platform on which thecorrection occurs due to different affordances on Facebook versus Twitter. Adding asource may prove more important for evaluating social corrections on Twitter, forexample, given that the platform is more commonly seen as a source of news by itsusers (Gottfried & Shearer, 2016) and is defined by a weaker and larger unknown audiencethan Facebook (Gruzd et al., 2011; Marwick & boyd, 2011). Given the lack of research inthis area, we pose a research question to test a moderated mediation model, in which plat-form moderates each of the proposed relationships (see Figure 1):

RQ3: Will the relationships among (1) including a source to social corrections, (2) evalu-ations of the social correction, and (3) misperceptions on the Zika issue differ on Facebookversus Twitter?

Methods

To test these hypotheses, we conducted an experiment embedded in an online survey inspring of 2016. Participants were recruited from a large Mid-Atlantic university andoffered course credit for their participation. The total sample included 613 valid responses,of which 271 were analyzed for this study.1 Participants in this study were roughly 20 yearsold (M= 20.17, SD = 4.28) and somewhat more male (57.6%) than female. As expected,participants also were relatively unfamiliar with the issue of the Zika virus (M= 2.99,SD = 1.49 on a five-point scale of familiarity), especially in comparison to other current

6 E. K. VRAGA AND L. BODE

issues like increased drug abuse (M= 4.03, SD = .99), the Islamic State in Iraq and Syria(ISIS) (M= 4.34, SD = .83) or climate change (M= 4.31, SD = .82).

This study used a 2 (platform: Twitter vs. Facebook) × 3 (Zika information: control,corrective response without source, corrective response with source) experimental designto test our hypotheses. After answering a short pre-test questionnaire, all participants wererandomly assigned to see either a simulated Facebook or Twitter news feed (series ofposts). Participants were asked to take their time reading through the posts becausethey would be asked questions after viewing all of the posts, and were required tospend at least five seconds on each page of the feed before the continue button wouldappear. Participants were also told that personal information about the people postingto the feed had been eliminated for privacy purposes, to maintain external validity.

Within either the Twitter or Facebook feed, participants were randomly assigned to oneof three experimental conditions. In the control condition, participants viewed three pagesof control posts, which included posts about social interactions and news. For the othertwo conditions, participants viewed the same three pages of control plus an additionalpage which contained a single news post, with an anonymous user claiming the Zika out-break was caused by genetically modified mosquitos in Brazil and posting a news storyfrom USA Today that validated that claim (in reality, this story was created by research-ers). This study manipulated the replies to this post and news story: in the social correctionwith sources conditions, two individual commenters both discredited the information andeach commenter provided a link to a debunking news stories from Snopes.com or theCDC, and in the social correction without sources, two individual posters claimed theinformation about genetically modified mosquitos was false but did not provide any out-side sources to support their claims (See Appendix for sample posts). After viewing thesimulated posts, participants answered the post-test questions before being thanked fortheir participant and debriefed, where they received information that (a) the storieswere all created by researchers for this study and (b) the scientific consensus is that geneti-cally modified mosquitos are not to blame for the Zika outbreak in Brazil (WHO, 2016b).

Figure 1. Moderated mediation model.

INFORMATION, COMMUNICATION & SOCIETY 7

Measures

Misinformation about Zika. Participants rated their level of agreement on a seven-pointscale with two statements designed to tap into their knowledge about the cause of theZika outbreak in Brazil: ‘the release of GMOmosquitos caused the Zika outbreak in Brazil’and ‘GMO mosquitos are to blame for the spread of the Zika virus in Brazil.’ These twoitems were combined into an index, with a higher score indicating greater misinformationabout the cause of the Zika virus (r = .77, p < .001, M= 3.81, SD = 1.17).

Evaluations of responses. Depending on their conditions, participants were asked toeither evaluate the ‘Facebook comments’ or ‘Twitter replies’ that appeared under the orig-inal post about the Zika outbreak as novel/new, useful, interesting, trustworthy, credible,biased (reversed), accurate, and relevant on seven-point scales. These items were all com-bined into a single index to compare across conditions (Facebook comments α = .88,M=3.68, SD = 1.28; Twitter replies α = .83, M= 3.46, SD = 1.18).

Results

To test our hypotheses, a series of two-way ANOVAs were performed. First, we examinethe effects of social correction with or without a source on misperceptions about the causesof the spread of the Zika virus. We observe a main effect of social correction type, F(2,269) = 4.74, p = .01, partial η2 = .035 (see Figure 2). H1 is supported – when the misinfor-mation is corrected and a source is provided, misperceptions are reduced (M= 3.54, SE= .12) compared to the control condition (M= 4.07, SE = .13, p = .01), in line with our pre-vious research (Bode & Vraga, in press). H2 predicted that social correction that provides asource will be more effective in reducing misperceptions compared to when no source wasprovided, and RQ1 asked whether such a correction would reduce misperceptions com-pared to the control. Our results suggest that social corrections without sources is noteffective in reducing misperceptions compared to the control (M= 3.84, SE = .12, p= .57) but neither is it significantly different from correction with sources using a Bonfer-roni correction (p = .24).

To test RQ2a, which investigated whether the process of social correction differed onFacebook versus Twitter, we examine the interaction between social correction type andplatform. This interaction is not significant, F(2, 269) = .33, p = .72, nor is the main effectof platform, F(1, 270) = .59, p = .44. Therefore, our results suggest that social correction

Figure 2. Misperceptions about the causes of Zika.

8 E. K. VRAGA AND L. BODE

requires sources to be effective to combat misperceptions and that such correction occursequally on Facebook and Twitter.

Next, we examine evaluations of the responses to the misinformation provided in theoriginal post. For these analyses, we exclude the control condition, as they did not seeeither the misinformation or corrective information and thus could not evaluate theseposts. H3 predicted that responses that provided a source for their correction would beevaluated more highly than responses without a source. The results support this hypoth-esis, F(1, 185) = 10.66, p = .001, partial η2 = .055, with social corrective responses that pro-vide a source rated significantly more highly (M= 3.86, SE = .12, p = .001) than thosewithout a source (M= 3.29, SE = .12). However, as suggested by RQ2b, this main effectis conditioned by whether the correction occurred on Facebook versus Twitter, F(1,185) = 6.60, p = .01, partial η2 = .035. The pairwise comparison using a Bonferroni correc-tion suggest that adding a source improved perceptions of the corrective replies on Face-book (p = .00) but not Twitter (p = .63; see Figure 3).2

Finally, RQ3 examines whether evaluations of the responses mediates the effects of pro-viding a source for the correction on misperceptions about the causes of the spread of theZika virus, and whether this effect is further moderated by social media platform (see Figure1). To test this moderated mediation model, we use the PROCESS macro, Model 59 (Hayes,2013). The results create a complicated picture, but do not support an overall moderatedmediation process. First, we confirm a significant interaction between platform and provid-ing sources that predicts evaluations of the response (β = .89, SE= .35, p = .01), as demon-strated in Figure 3. Next, we find that evaluations of the response to the misinformationis a significant predictor of Zika misperceptions (β =−.67, SE = .24, p = .01), with higherevaluations of the corrective replies reducing misinformation. However, a second significantinteraction term emerges between platform and evaluation of the social correction replies(β = .32, SE = .15, p = .04), such that the relationship between the evaluations of the socialcorrection is stronger for Twitter than Facebook. To parse this interaction, a separateregression analysis was performed for Twitter versus Facebook, which demonstrated a sig-nificant negative relationship between evaluations of the social correction andmisperceptionson Twitter (β =−.35, SE = .11, p < .01) but not on Facebook (β =−.04, SE = .11, p = .71). As a

Figure 3. Evaluations of corrective responses.

INFORMATION, COMMUNICATION & SOCIETY 9

result of these contrasting interactions, the index ofmoderatedmediation is not significant, asis indicated when the confidence interval includes ‘0’ (β = .00, SE = .18, lower limit confidenceinterval =−.36, upper limit confidence interval = .36). These results provide a mixedresponse to RQ3: on Facebook, providing a source improves evaluations of the correctivecomments, but these higher evaluations are not associated with reduced misperceptionson the causes of Zika, whereas on Twitter adding a source does not affect evaluations ofthe corrective responses, but higher evaluations of the social correction are related to reducedmisperceptions on the Zika issue (Table 1).

Discussion and conclusions

This study extends previous research by investigating the ways in which social correctionof misinformation functions across two prominent social media platforms: Facebook ver-sus Twitter. Our results expand on evidence that social media can provide effective correc-tion of misinformation, resulting in reduced misperceptions (Bode & Vraga, in press; Bode& Vraga, 2015) by testing the mechanisms by which other users can engage in social cor-rection, as well as whether this process differs depending on online contexts. This hasimportant implications for how we think about the role of social media in the moderninformation environment.

More specifically, it seems that when everyday users are correcting one another, theprovision of a source confirming the correction being offered is required in order to suc-cessfully address and mitigate misperceptions. Social correction absent this corroboratingevidence did not significantly reduce misperceptions compared to the control. This makessense from a theoretical perspective – provision of additional information should makethat information seem more credible and thus make the argument more persuasive(Petty & Cacioppo, 1984; Pornpitakpan, 2004). It also has important practical impli-cations. Those engaged in anti-misinformation campaigns can empower and encourageeveryday users to correct misinformation when they see it. To do so, organizations andnon-profits interested in promoting correct information can disseminate easily sharedmaterial, adapted for social media posting. This allows organizations to effectively multi-ply their influence, by encouraging others to engage in correction with them.

These results are complicated, however, when we consider the social media platform onwhich the social correction occurs. On both Facebook and Twitter, social correction with asource was successful in reducing misperceptions as compared to a control condition.However, the mechanisms that were successful for social correction with a source differed

Table 1. Results from the moderated mediation analysis.Mediator: response

evaluationsOutcome: Zikamisperceptions

β SE β SE

Constant 3.62 .40 5.84 .89Response evaluations – – −0.67** .24Comments with sources −0.77 .56 −0.24 .57Platform −0.22 .25 −0.89 .55Sources × Platform 0.89* .35 −0.10 .36Evaluation × Platform 0.32* .15R2 .097 .079

*p < .05, **p < .01, ***p < .001.

10 E. K. VRAGA AND L. BODE

in comparison to social corrections without a source. On Facebook, providing an externalsource to substantiate the correction significantly improved perceptions of the correctivecomments, while adding this source did not affect evaluations of the corrective replies onTwitter. However, the moderated mediation analysis suggested that these higher evalu-ations of the corrective replies did not directly translate into lower misperceptions onthe Zika issue on Facebook. In other words, even though individuals saw a social correc-tion that provided a source as more credible on Facebook, this heightened credibility didnot lead them to reduce their misperceptions compared to a correction without a source.In contrast, on Twitter, individuals with higher evaluations of the social correction weremore likely reduce their misperceptions regarding Zika compared to those with lowerevaluations of the correction – but adding a source to the correction was not responsiblefor these heightened credibility perceptions.

These findings have several implications for theory development and practical rec-ommendations about correcting misperceptions on social media. First, and most impor-tantly, these results reinforce the value of studying social media platforms separately. Thisstudy highlights that the mechanism underlying the effectiveness of including a source aspart of social correction differs on Facebook versus Twitter. We expect these differencesare rooted in different technical and social affordances are likely to promote some typesof information sharing over others (Lyons, in press). Most notably for this study, peopleare more likely to get news from Twitter than from Facebook (Gottfried & Shearer, 2016).Perhaps this changes their expectations of what news or information looks like in differentplatforms. If Twitter is seen as a space for news, including a source may be redundant forevaluations of the social correction, where the expectation is already that a tweet is offeringnew, and presumably credible, information. However, even if this assumption of credi-bility is not driven by including a source, seeing a social correction as credible directlytranslates into reducing misperceptions on the issue. Therefore, this study reinforcesthe importance of understanding credibility perceptions on Twitter, even if it cannotexplain what causes some people to evaluate the corrective replies on Twitter more highly.Future research should investigate what types of information or what individual charac-teristics enhance perceptions of information quality on Twitter, as it seems like this is anecessary precursor to successfully addressing misperceptions.

On Facebook, on the other hand, expectations are that content will be primarily social(Ellison, Steinfield, & Lampe, 2007; Vraga, Thorson, Kligler-Vilenchik, & Gee, 2015),which could lead users to assume that any posts are opinion rather than fact. When asource is provided, therefore, it elevates the post (or, in our case, the comment) fromsomeone’s point of view to a statement of fact, backed up by evidence. However, improv-ing evaluations of the social correction does not then lead to adjustment in attitudes aboutthe causes of Zika. As a platform whose primary purpose is social in nature (rather thannews), people may be less willing to be persuaded by information on Facebook, even infor-mation they deem credible. Alternatively, it may be that social corrections on Facebook aremore likely to have a sleeper effect (e.g., Eagly & Chaiken, 1993; Pornpitakpan, 2004), withpeople more slowly updating their attitudes about controversial issues as the source of theinformation is forgotten. Future research should examine whether social correction onsocial media is more enduring in one format or on one platform compared to others.

A final explanation for these competing findings may rest with the audiences of thesesocial media platforms. As in the general population, respondents in our sample were

INFORMATION, COMMUNICATION & SOCIETY 11

more likely to have a Facebook account (84% of our sample) than to have a Twitteraccount (44% of our sample). Simple familiarity with the platform likely changes theexpectations that people have for it, which could impact their perceptions of informationconveyed there as well, especially given likely differences in familiarity with the platformand its affordances. Supplemental analyses suggest the pattern of effects is largely similarwhen only examining those individuals who have both a Facebook and Twitter account,although the reduced sample size reduces some of the effects to insignificance.3 Most nota-bly, the interaction between platform and evaluation of the corrective comment is alsoreduced to insignificance (β = .26, SE = .25, p = .29), suggesting that among those whouse both Facebook and Twitter, higher evaluations of the corrective information had adirect impact on mitigating misperceptions. The small sample size limits our confidencein these results, however, and future research should explore how misperceptions arereduced for users and non-users of a given platform separately. This is of particular impor-tance on Twitter, as content shared on this platform often moves across channels (e.g.,Thorson et al., 2013). Future research should focus on building an understanding ofwhich affordances may be affecting the ways in which people interpret misinformationand its correction on social media.

This study is also limited by its use of an artificial social media feed, rather than sub-jects’ own native content. As a result, users had no existing knowledge of the person post-ing misinformation, nor of those offering corrective comments. This may be of particularconcern on Facebook, where friend networks (while large) often consist largely of knownothers (Hampton, Goulet, Rainie, & Purcell, 2011), whose credibility may influence thepersuasiveness of their posts. In contrast, the larger non-reciprocal networks on Twittermay make correction from unknown others more realistic. Similarly, the social correctionsin this study offer simple, clear, and repeated rebuttals of the misinformation, followingbest practices (Bode & Vraga, 2015; Lewandowsky et al., 2012). However, such correctionsmay be much less civil in practice, and future research should investigate both what socialcorrection looks like when it occurs online and the effects that differing contextual cues –including tone, number of corrections, and source – have on the effectiveness of social cor-rection across platforms.

This study reinforces the importance of studying how the evolving new media environ-ment alters information and misinformation flows online. If social media are often criticizedas a source of misinformation, this study suggests they can also serve as a corrective, whensocial peers arewilling tooffer short rebuttals of informationbackedupwith a credible source.Moreover, although the mechanisms may differ, social correction with a source was equallyeffective in reducing misperceptions on both Facebook and Twitter. These findings offerhope for planning and implementing social media campaigns to address misperceptions ona range of health and science issues: one’s peers can effectively offer social corrections, solong as they provide a source for their information, on diverse social media platforms

Notes

1. Three types of participants were excluded from this study. First, participants in several exper-imental conditions, which included a manipulation of unrelated responses to the misinfor-mation and exposure to a Facebook algorithm for correction, were excluded from thisstudy (N = 246). These conditions were not crossed with the experimental design examined

12 E. K. VRAGA AND L. BODE

here and are outside the scope of this study. Second, we exclude participants who do not passan attention check (N = 93), which asked participants to select a specific answer to a questionin the post-test to indicate they are paying attention. Finally, we included data only from thefirst time users participated in the survey to maintain internal validity (N = 43).

2. There is no main effect of platform on evaluations of the corrective responses, F(1, 185) =1.65, p = .20.

3. Although the main effects of social correction without or a with a source compared to thecontrol is reduced to insignificance, F(2, 101) = 1.61, p = .21, partial η2 = .033, the similareffect size and pattern of results suggest it may result from a substantial reduction inpower. Moreover, the interaction between platform and source remains significant for pre-diction evaluations of the social corrections, F(1, 71) = 6.18, p = .02, partial η2 = .084.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Emily K. Vraga (Ph.D., University of Wisconsin –Madison) is an assistant professor in the Depart-ment of Communication at George Mason University. Her research focuses on how individualsprocess news and information about contentious political, scientific, and health issues, particularlyin response to disagreeable messages in digital media environments. She is interested in testingmethods to limit biased processing and misinformation and to encourage attention to more diversecontent online. [email: evraga@gmu.edu, or on Twitter at @ekvraga].

Leticia Bode (Ph.D., University of Wisconsin –Madison) is an assistant professor in the Communi-cation, Culture, and Technology master’s program at Georgetown University. Her work lies at theintersection of communication, technology, and political behavior, emphasizing the role communi-cation and information technologies may play in the acquisition and use of political information.She can be contacted via mail at 3520 Prospect St NW Suite 311, Washington DC 20057 [email:lb871@georgetown.edu or on Twitter at @leticiabode].

ORCID

Emily K. Vraga http://orcid.org/0000-0002-3016-3869

References

Al-Qahtani, A. A., Nazir, N., Al-Anazi, M. R., Rubino, M. R., & Al-Ahdal, M. N. (2016). Zika virus:A new pandemic threat. The Journal of Infection in Developing Countries, 10(3), 201–207.

Barberá, P., Jost, J. T., Nagler, J., Tucker, J. A., & Bonneau, R. (2015). Tweeting from left to right: Isonline political communication more than an echo chamber? Psychological Science, 26(10),1531–1542.

Barberá, P. (2014). How social media reduces mass political polarization: Evidence from Germany,Spain, and the U.S. working paper. Retrieved from http://pablobarbera.com/static/barbera_polarization_APSA.pdf

Berinsky, A. J. (2017). Rumors and health care reform: Experiments in political misinformation.British Journal of Political Science, 47(2), 241–262. doi:10.1017/S0007123415000186

Bode, L., Hanna, A., Yang, J. H., & Shah, D. V. (2015). Candidate networks, citizen clusters, andpolitical expression: Strategic Hashtag use in the 2010 midterms. The ANNALS of theAmerican Academy of Political and Social Science, 659(1), 149–165.

Bode, L., & Vraga, E. K. (2015). In related news, that was wrong: The correction of misinformationthrough related stories functionality in social media. Journal of Communication, 65, 619–638.

INFORMATION, COMMUNICATION & SOCIETY 13

Bode, L., & Vraga, E. (in press). See something, say something: Correction of global health misin-formation on social media. Health Communication.

boyd, d. (2010). Social network sites as networked publics: Affordances, dynamics and implications.In Z. Papacharissi (Ed.), Networked self: Identity, community, and culture on social network sites(pp. 39–58). New York: Routledge.

Bruns, A., & Moe, H. (2014). Structural layers of communication on Twitter. In K. Weller, A. Bruns,J. Burgess, M. Mahrt, & C. Puschmann (Eds.), Twitter and society (pp. 15–28). New York, NY:Peter Lang.

Castillo, C., Mendoza, M., & Polete, B. (2011). Information credibility on Twitter. In WWW ’11proceedings of the 20th international conference on World Wide Web (pp. 675–684). NewYork: ACM.

Diehl, T., Weeks, B. E., & gil de Zuniga, H. (2016). Political persuasion on social media: Tracingdirect and indirect effects of news use and social interaction. New Media & Society, 18(9),1875–1895.

Dixon, G., & Clarke, C. (2013). The effect of falsely balanced reporting of the autism–vaccine con-troversy on vaccine safety perceptions and behavioral intentions. Health Education Research, 28(2), 352–359.

Dredze, M., Broniatowski, D. A., & Hilyard, K. M. (2016). Zika vaccine misperceptions: A socialmedia analysis. Vaccine, 34, 3441–3442.

Duggan, M. (2015). The demographics of social media users. Pew Research Center. Retrieved fromhttp://www.pewinternet.org/2015/08/19/the-demographics-of-social-media-users/

Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX: Harcourt, Brace,Jovanovich.

Ecker, U. K., Lewandowsky, S., Swire, B., & Chang, D. (2011). Correcting false information in mem-ory: Manipulating the strength of misinformation encoding and its retraction. PsychonomicBulletin & Review, 18(3), 570–578. doi:10.3758/s13423-011-0065-1

Edwards, C., Edwards, A., Spence, P. R., & Shelton, A. K. (2014). Is that a bot running the socialmedia feed? Testing the differences in perceptions of communication quality for a humanagent and a bot agent on Twitter. Computers in Human Behavior, 33, 372–376.

Ehrenberg, R. (2012). Social media sway: Worries over political misinformation on Twitter attractscientists’ attention. Science News, 182(8), 22–25. doi:10.1002/scin.v182.8

Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Facebook “Friends”: Social capitaland college students’ use of online social network sites. Journal of Computer-MediatedCommunication, 12(4), 1143–1168. doi:10.1111/jcmc.2007.12.issue-4

Ellison, N., & Vitak, J. (2015). Social media affordances and their relationship to social capital pro-cesses. In S. Sundar (Ed.), The handbook of psychology of communication technology (pp. 205–227). Boston, MA: Wiley-Blackwell.

Feldman, L., Myers, T. A., Hmielowski, J. D., & Leiserowitz, A. (2014). The mutual reinforcement ofmedia selectivity and effects: Testing the reinforcing spirals framework in the context of globalwarming. Journal of Communication, 64, 590–611.

Fowler, A., & Margolis, M. (2014). The political consequences of uninformed voters. ElectoralStudies, 34(1), 100–110. doi:10.1016/j.electstud.2013.09.009

Fox, J., & Moreland, J. J. (2015). The dark side of social networking sites: An exploration of the rela-tional and psychological stressors associated with Facebook use and affordances. Computers inHuman Behavior, 45, 168–176.

Garrett, R. K. (2011). Troubling consequences of online political rumoring. HumanCommunication Research, 37, 255–274.

Gauchat, G. (2012). Politicization of science in the public sphere: A study of public trust in theUnited States, 1974 to 2010. American Sociological Review, 77(2), 167–187.

Gibson, J. J. (1986). The ecological approach to visual perception. Hillsdale, NJ: Lawrence ErlbaumAssociates.

Gottfried, J., & Shearer, E. (2016). News use across social media platforms 2016. Pew Research Center.Retrieved fromhttp://www.journalism.org/2016/05/26/news-use-across-social-media-platforms-2016/

14 E. K. VRAGA AND L. BODE

Griffiths, E. (2016). Was Zika outbreak caused by release of genetically modified mosquitoes inBrazil? The Mirror. Retrieved from http://www.mirror.co.uk/news/world-news/zika-outbreak-caused-release-genetically-7281671

Gruzd, A., Wellman, B., & Takhteyev, Y. (2011). Imagining Twitter as an imagined community.American Behavioral Scientist, 55(10), 1294–1318.

Hampton, K. N., Goulet, L. S., Rainie, L., & Purcell, K. (2011). Social networking sites and our lives.Pew Research Center. Retrieved from http://www.pewinternet.org/2011/06/16/social-networking-sites-and-our-lives/

Hayes, A. (2013). Introduction to mediation, moderation, and conditional process analysis. NewYork: Guilford Press.

Hmielowski, J. D., Feldman, L., Myers, T. A., Leiserowitz, A., & Maibach, E. (2014). An attack onscience? Media use, trust in scientists, and perceptions of global warming. Public Understandingof Science, 23(7), 866–883.

Jahng, M. R., & Littau, J. (2016). Interacting is believing: Interactivity, social cue, and perceptions ofjournalistic credibility on Twitter. Journalism and Mass Communication Quarterly, 93, 38–58.

Jerit, J., & Barabas, J. (2012). Partisan perceptual bias and the information environment. TheJournal of Politics, 74(3), 672–684.

Kata, A. (2010). A postmodern Pandora’s box: Anti-vaccination misinformation on the Internet.Vaccine, 28(7), 1709–1716. doi:10.1016/j.vaccine.2009.12.022

Lewandowsky, S., Ecker, U. K. H., Seifert, C. M., Schwarz, N., & Cook, J. (2012). Misinformationand its correction: Continued influence and successful debiasing. Psychological Science in thePublic Interest, 13(3), 106–131.

Lyons, B. (in press). From code to discourse: Social media and linkage mechanisms in deliberativesystems. Journal of Public Deliberation.

Maireder, A., & Ausserhofer, J. (2013). Political discourses on Twitter: Networking topics, objects andpeople (pp. 291–341). Twitter and Society. New York, NY: Peter Lang.

Mao, J. (2014). Social media for learning: A mixed methods study on high school students’ technol-ogy affordances and perspectives. Computers in Human Behavior, 33, 213–223.

Marwick, A. E., & boyd, D. (2011). I tweet honestly, I tweet passionately: Twitter users, context col-lapse, and the imagined audience. New Media & Society, 13, 114–133.

McCright, A. M., & Dunlap, R. E. (2011). The politicization of climate change and polarization inthe American public’s views of global warming, 2001–2010. The Sociological Quarterly, 52(2),155–194.

Messing, S., & Westwood, S. J. (2014). Selective exposure in the age of social media: Endorsementstrump partisan source affiliation when selecting news online. Communication Research, 41(8),1042–1063.

Mitchell, A., Gottfried, J., Barthel, M., & Shearer, E. (2016). The modern news consumer: News atti-tudes and practices in the digital era. Pew Research Center. Retrieved from http://www.journalism.org/2016/07/07/the-modern-news-consumer/

Nyhan, B. (2010). Why the ‘death panel’ myth wouldn’t die: Misinformation in the health carereform debate. The Forum, 8(1). doi:10.2202/1540-8884.1354

Nyhan, B., & Reifler, J. (2010). When corrections fail: The persistence of political misperceptions.Political Behavior, 32(2), 303–330. doi:10.1007/s11109-010-9112-2

Nyhan, B., Reifler, J., & Ubel, P. A. (2013). The hazards of correcting myths about health carereform. Medical Care, 51(2), 127–132. doi:10.1097/MLR.0b013e318279486b

Oyeyemi, S. O., Gabarron, E., &Wynn, R. (2014). Ebola, Twitter, and misinformation: A dangerouscombination? BMJ, 349, g6178–g6178.

Petty, R. E., & Brinol, P. (2008). Persuasion: From single to multiple to metacognitive processes.Perspectives on Psychological Science, 3, 137–147.

Petty, R. E., & Cacioppo, J. T. (1984). The effects of involvement on responses to argument quantityand quality: Central and peripheral routes to persuasion. Journal of Personality and SocialPsychology, 46, 69–81.

Pornpitakpan, C. (2004). The persuasiveness of source credibility: A critical review of five decades’evidence. Journal of Applied Social Psychology, 34, 243–281.

INFORMATION, COMMUNICATION & SOCIETY 15

Radzikowski, J., Stefanidis, A., Jacobsen, K. H, Croitoru, A., Crooks, A., & Delamater, P. L. (2016).The measles vaccination narrative in Twitter: A quantitative analysis. JMIR Public Health andSurveillance, 2(1), e1. doi:10.2196/publichealth.5059

Reedy, J., Wells, C., & Gastil, J. (2014). How voters become misinformed: An investigation of theemergence and consequences of false factual beliefs. Social Science Quarterly. Online first.doi:10.1111/ssqu.12102

Rojecki, A., & Meraz, S. (2016). Rumors and factitious informational blends: The role of the web inspeculative politics. New Media & Society, 18, 25–43.

Rosenbaum, L. (2014). Invisible risks, emotional choices – mammography and medical decisionmaking. New England Journal of Medicine, 371, 1549–1552.

Schrock, A. R. (2015). Communicative affordances of mobile media: Portability, availability, locat-ability, and multimediality. International Journal of Communication, 9, 1229–1246.

Sharma, M., Yadav, K., Yadav, N., & Ferdinand, K. C. (2017). Zika virus pandemic – analysis ofFacebook as a social media health information platform. American Journal of InfectionControl, 45(3), 301–302.

Slater, M. D., & Rouner, D. (1996). Howmessage evaluation and source attributesmay influence credi-bility assessment and belief change. Journal and Mass Communication Quarterly, 73, 974–991.

Small, T. A. (2011). What the hashtag? A content analysis of Canadian politics on Twitter.Information, Communication, and Society, 14, 872–895.

Solon, O. (2016). In firing human editors, Facebook has lost the fight against fake news. TheGuardian. Retrieved from https://www.theguardian.com/technology/2016/aug/29/facebook-trending-news-editors-fake-news-stories

Starbird, K., Maddock, J., Orand, M., Achterman, P., & Mason, R. M. (2014). Rumors, false flags,and digital vigilantes: Misinformation on Twitter after the 2013 Boston Marathon Bombing. IniConference 2014 proceedings (pp. 654–662). Berlin, Germany.

Taber, C. S., & Lodge, M. (2006). Motivated skepticism in the evaluation of political beliefs.American Journal of Political Science, 50(3), 755–769.

Tafuri, S., Gallone, M. S., Cappelli, M. G., Martinelli, D., Prato, R., & Germinario, C. (2014).Addressing the anti-vaccination movement and the role of HCWs. Vaccine, 32(38), 4860–4865. doi:10.1016/j.vaccine.2013.11.006

Thorson, K., Driscoll, K., Edgerly, S., Ekdale, B., Schrock, A., Swartz, L.,…Wells, C. (2013).YouTube, Twitter, and the occupy movement: Connecting content and circulation practices.Information, Communication, & Society, 16, 421–451.

Thorson, K. (2014). Facing an uncertain reception: Young citizens and political interaction onFacebook. Information, Communication, and Society, 17, 203–216.

Thorson, K., & Wells, C. (2016). Curated flows: A framework for mapping media exposure in thedigital age. Communication Theory, 26, 309–328.

Treem, J. W., & Leonardi, P. M. (2012). Social media use in organizations: Exploring the affordancesof visibility, editability, persistence, and association. Communication Yearbook, 36, 143–189.

Vraga, E. K., Thorson, K., Kligler-Vilenchik, N., & Gee, E. (2015). How individual sensitivities todisagreement shape youth political expression on Facebook. Computers in Human Behavior,45(1), 281–289. doi:10.1016/j.chb.2014.12.025

Westerman, D., Spence, P. R., & Van Der Heide, B. (2014). Social media as information source:Recency of updates and credibility of information. Journal of Computer-MediatedCommunication, 19(2), 171–183. doi:10.1111/jcc4.12041

World Health Organization. (2016a). WHO Director-General summarizes the outcome of theEmergency Committee regarding clusters of microcephaly and Guillain-Barré syndrome.Retrieved from http://www.who.int/mediacentre/news/statements/2016/emergency-committee-zika-microcephaly/en/

World Health Organization. (2016b). Dispelling rumours around Zika and complications. WorldHealth Organization. http://www.who.int/emergencies/zika-virus/articles/rumours/en/

Worth, K. (2016). As Brazil confronts Zika, vaccine rumors shape perceptions. Frontline. Retrievedfrom http://www.pbs.org/wgbh/frontline/article/as-brazil-confronts-zika-vaccine-rumors-shape-perceptions/

16 E. K. VRAGA AND L. BODE

Appendix. Sample posts

Figure A1. Facebook social correction with sources.

Figure A2. Twitter social correction without sources.

INFORMATION, COMMUNICATION & SOCIETY 17