Can we change your mind about Assad? : processing counter-attitudinal information: an experiment using the Syrian conflict as a case study

! 


Master Thesis

Can we change your mind about Assad?

Processing counter-attitudinal information: An experiment using the Syrian conflict as a case study

Heidi Hindam 11260394 June 2018

Introduction 1

Theoretical framework 3

A. Effects of exposure to cross-cutting information and prior knowledge 3 B. Involvement and differential processing 5

C. Differential processing and attitude change 7 D. Case study 8

Methodology 10

Design 10

Sample 11

Counter-attitudinal treatment variable 12 Attitude amplification variable 12

Prior attitude strength variable 14 Involvement 14

Systematic & heuristic processing 15 Results 16 Randomization checks 16 Manipulation check 17 Testing H1 17 Testing H2a 18 Testing H2b 19

Testing H3a & H3b 20

Discussion and conclusion 21 Limitations and future research 25 Bibliography 28

Appendix 1: tables and figures 34 Appendix 2: stimulus sample 43 Appendix 3: Questionnaire 44

INTRODUCTION

Political communication scholars interested in the effect of exposure to counter-attitudinal information have often hypothesised that the most politically sophisticated citizens were the most likely to receive a counter-attitudinal message, but also the least likely to be per-suaded by it (Vaccari, 2008). In contrast, the least politically sophisticated were supposed-ly the least likesupposed-ly to be exposed to heterogeneous information, while being the most likesupposed-ly to be persuaded by it. This would leave only the moderately politically interested citizens as a viable demographic with a potential for deliberation and opinion change (Vaccari, 2008). This is especially relevant in a media landscape where the low cost of news pro-duction and distribution has made a diverse range of views highly accessible online (Flaxman, Goel, & Rao, 2016). Whether through social networking website or search en-gines, citizens today can definitely access a broader range of opinions and narratives than those found in traditional media (Flaxman, Goel, & Rao, 2016). Given this diversified landscape, it can be understood that investigating the effect of exposure to dissonant in-formation can be extremely relevant for political communication scholarship. Especially in an era where “fake news”, a term signifying a fabricated story, makes it into popular culture (Allcott & Gentzkow, 2017, p. 212).

Traditionally, exposure to heterogeneous information has often been studied with classical democratic indicators such as political cynicism and interest (Hampton, Shin, & Lu, 2017; Lu, Heatherly, & Lee, 2016; Torcal & Maldonado, 2014), voter turnout (Mc-Clurg, 2006; Schuck, Vliegenthart, & De Vreese, 2014), and attitudes during election time (Christin, Hug, & Sciarini, 2002; Chung, & Waheed, 2016; Vaccari, 2008). This associa-tion is usually due to the fact that exposure to a diverse range of opinions is highly prized

in the normative frameworks of most democratic models (Ferree, Gamson, Gerhards & Rucht, 2002; Strömbäck, 2005). Whether for representative, participatory or deliberative democracy ideals, the supposition is that the role of the media in such political systems is to include heterogeneous points of view in the public discourse in order to adequately equip the citizen with the information necessary to optimally perform their civic duty (Ferree et al., 2002; Strömbäck, 2005).

Accurate information is also important when consider foreign policy issues. For instance, in most democratic societies, leaders do seek the mobilisation of public support before engaging in military action abroad. This was for instance very visible in the case of the American invasion of Iraq in 2003 sanctioned by the Bush administration (Dutta-Bergman, 2005; Luther & Miller, 2005; Schwalbe et al., 2004). In fact, to garner and maintain support for the invasion, the Bush administration had to engage in a costly per-suasion campaign (Hiebert, 2003; Kull, Ramsay, & Lewis, 2003). The political cam-paign’s aim was to convince the public of two contentions. The first claim was that the war would be a neat and bloodless surgical intervention. The second assertion was that Saddam Hussein possessed weapons of mass destruction (WMD) which put their own na-tional interests at stake (Hiebert, 2003; Kull, Ramsay, & Lewis, 2003). Interestingly, re-search has found that individuals who believed the statement that Iraq possessed WMDs to be true were more likely to support President Bush (Kull et al., 2003). This finding highlights the pivotal role of accurate factual information about foreign policy issues es-pecially in the context of support or opposition to military intervention. When considering exposing citizens to accurate information, it is important to note that their initial state can be uninformed or misinformed (i.e., believing erroneous information to be true; Nyhan & Reifler, 2010). Studies in the field of political heuristics have demonstrated the difficulty

of overcoming confidently held false beliefs and concluded that misinformation represent-ed an obstacle in represent-educating the public (Nyhan & Reifler, 2010). Perhaps the field of politi-cal heuristics can offer some insights into the study of exposure to cross-cutting narratives which hold the potential for correcting such misinformation. As such, I would like to vestigate the relationship between exposure to cross-cutting information and political in-volvement on the potential for attitude change. This study will attempt to do so by using theories of differential cognitive processing.

THEORETICAL FRAMEWORK

A. Effects of exposure to cross-cutting information and prior knowledge

The extant literature on the effects of exposure to heterogeneous information is divided. Some scholars believe that, exposure to counter-attitudinal information dissuades the citi-zens from being politically engaged (Hampton, Shin & Lu, 2017; Kim & Chen, 2017; Lu, Heatherly & Lee, 2016; Shi, 2016; Torcal & Maldonado, 2014). Others have found expo-sure to cross-cutting information to rather stimulate political interest (McLeod, Scheufele, Moy, Horowitz, et al. 1999; Scheufele, Nisbet, Brossard & Nisbet, 2004). Regarding the potential for changing attitudes, diverse effects were also detected: While most studies concluded that dissonance only amplified the individual’s initial attitude (Holbert, Garrett & Gleason, 2010; Nyhan & Reifler, 2010), some scholars managed to observe some opin-ion change within their sample (Gilens, 2001; Miller, Krochik & Jost, 2009). Yet others found no variation at all in the attitudes measured (Sides & Citrin, 2007). A factor poten-tially explaining part of the inconsistent results is the lack of consistency in the opera-tionalization of information heterogeneity. Taking the definition of Nir (2011),

heterogene-ity can be understood as either (a) the presence of competing points of view, or (b) a con-flict with one’s own belief. This distinction, although rarely made in the literature (Kim & Chen, 2015), is of particular relevance in explaining why certain scenarios are likely to prompt or delay political participation. As Nir (2011) argues, exposure to competing opin-ions within one’s network is likely to draw the individual’s attention to their lack of knowledge on the issue at hand. This, in turn, promotes ambivalence and a lack of confi-dence in one’s own attitude, which then dissuades the citizen from expressing their unripe attitude. The notion of confidence in one’s issue position would also explain why Torcal and Maldonado (2014) found the demobilization effect to be even more pronounced among the less politically knowledgeable and why, on the other hand, McClurg (2006) found that having an expert in one’s political discussion network promoted participation by increasing confidence in the information at hand. Corroborating this explanation is Gilens’ (2001) experiment: The researcher manipulated the initial level of issue-relevant factual knowledge held by participants at the beginning of the experiment. As expected, attitude change was only observed after presenting participants whose initial attitude was not based on solid knowledge with relevant information. As such, it appears that the most relevant attribute to retain from the operationalization of dissonance as (a) exposure to competing views within one’s network is that it does not assume that the individual ex-posed to the heterogeneity possesses a strong or factual initial belief. It is worth nothing that, operationalizing dissonance as (b) conflict with one’s own beliefs presupposes an ex-isting attachment to the attitude at hand from previous encounter with the polarizing issue or because the attitude derives from an individual’s political preference (Nyhan & Reifler, 2010). This explanation is consistent with Sides and Citrin’s (2007) findings where the researchers exposed American citizens to the real figures of immigration in their country,

which conflicted with their expectations, but were still unable to change their attitude to-wards the topic of immigration. The authors interpreted the discounting of relevant but cross-cutting information as a result of the participants’ party identification. In sum, it ap-pears that when confronted with heterogeneous views, an individual’s potential for politi-cal participation or attitude change greatly depends on the strength of their initial attitudes, if present. Bullock (2007) argues that when an individual is exposed to a cross-cutting message they have previously encountered, the individual is quicker to remember the ar-guments supporting that prior belief rather than the arar-guments opposing it. The individual then uses one’s own pre-generated arguments to counter-argue the dissonant information even more strongly. This creates a backfire effect, explaining why the aforementioned studies observed an amplification in the participants’ initial attitudes (Holbert, Garrett & Gleason, 2010; Nyhan & Reifler, 2010). And further explaining why attitude change is only observed when the participants do not possess enough issue-relevant knowledge (Gilens, 2001), or in other words, are not equipped with an extensive repertoire of counter-arguments. As such, we pose the following hypothesis:

—> H1: For individuals that hold stronger prior attitudes, cross-cutting messages will amplify initial attitudes compared to individuals that hold weaker prior attitudes.

B. Involvement and differential processing

Not all counter-attitudinal information is processed equally: difference in the cognitive processing routes exist. This is explained by the dual processing theories of persuasion such as Petty and Cacioppo’s (1986) Elaboration-Likelihood Model or Chaiken’s (1987) Heuristic-Systematic Model. Both theories are based on the premise that information can

be processed either via a central (Petty & Cacioppo, 1986) or systematic (Chaiken, 1987) route, or through a peripheral (Petty & Cacioppo, 1986) or heuristic (Chaiken, 1987) route. The central or systematic route, more careful and analytic, is dedicated to informa-tion that needs to be scrutinized and where persuasive arguments are closely evaluated (Chaiken, 1987; Petty & Cacioppo, 1986). According to Petty and Cacioppo (1986), how-ever, for the message to be processed centrally, the individual must possess the cognitive ability as well as motivation to do so. In contrast, the heuristic or peripheral route is rather used to to skim through the information at hand by resorting to cues or “shortcuts” sur-rounding the persuasive message. Such shortcuts include judging the validity of the mes-sage by its source, rather than devoting considerable cognitive resources to process and evaluate the information for its content (Chaiken, 1987; Petty & Cacioppo, 1986). Accord-ingly, the individual will attempt to balance both systems and spend the least amount of cognitive effort necessary while achieving enough confidence in their judgement (Bohner, Moskowitz & Chaiken, 1995). In line with this reasoning, Chaiken (1980) argues that in-dividuals are naturally more inclined to resort to the more economic heuristic route. How-ever, the author argues that in order for individuals to be motivated to carefully process information (by using the systematic route), they need to be involved in the issue at stake (Chaiken, 1980). This link between motivation, involvement and elaborative processing is confirmed by Hampton and colleagues’ (2017) study. The research demonstrated that, in online settings, only highly involved individuals holding strong opinions about a political issue were willing to engage when exposed to cross-cutting information. Willingness to engage evidently being associated with systematic processing. As such, we pose the fol-lowing hypotheses:

•

•

C. Differential processing and attitude change

As previously mentioned, politically involved individuals, assuming their involvement translated into political knowledge, can react to counter-attitudinal information in two manners: Their initial attitudes cam be amplified if their counter-arguing leads to an in-crease in salience of their pre-existing pro-attitudinal arguments. Alternatively, the new information will decrease their confidence in their initial attitude by, not changing their opinion, but introducing doubt and a “intellectual paralysis” (Barker & Hansen, 2005, p. 322) that discourages engagement without necessarily incurring attitude change. As such, politically sophisticated individuals are arguably more resilient to opinion change com-pared to their less knowledgeable counterparts. This can be explained by the information retention style of each individual. In fact, it can be argued that the more politically en-gaged citizens are more likely to expose themselves to more factual knowledge that influ-ences long-term memory. The less engaged citizenry is more likely to acquire knowledge via more heuristic routes that do not translate into long-term knowledge (Baum, 2003). Lending support to this mechanism is Chaiken’s (1980) research demonstrating that atti-tude change mediated by content-related cues (e.g. number of arguments) lasted longer

than attitude change mediated by heuristic shortcuts such as source cues. The authors fur-ther explain that, although heuristic and systematic processing routes can be activated si-multaneously, highly involved individuals are more likely to be persuaded by content cues generated during the systematic processing task, while less involved citizens are more likely to be persuaded by cues generated during a more heuristic processing task. In light of this research, the following hypotheses are formulated:

•

•

D. Case study

Studying conceptions and misconceptions is particularly interesting in the case of the Syr-ian conflict due to the overwhelming documentation surrounding this war. In fact, most scholars consider the Syrian war to be the most documented conflict in the history of digi-tal communication (Lynch, Freelon & Aday, 2014), and some scholars even dubbed the bulk of information emanating from the conflict the Syrian “data glut” (Powers & O’Loughlin, 2015, 1). The authors attributed this unprecedented flow of information to the rising levels of digital literacy among the Syrian population attempting to satisfy utilitari-an needs (e.g. giving their own account of the events; attempting to garner sympathy) in

times of conflict (Powers & O’Loughlin, 2015). In contrast, the utilitarian needs for As-sad’s government would be winning the “PR battle”. Concretely, this means successfully depicting their opponents as criminals, terrorists or foreign conspirators. In turn, this would support the government’s narrative that the civil uprisings are no more than a for-eign conspiracy aimed at achieving regime change in Syria (Powers & O’Loughlin, 2015). Emphasising evidence that would support this narrative should lend the regime more legit-imacy, shielding the latter against foreign criticism, and demobilise its opponents and their supporters. For the opposition and their sympathisers, the interest in documenting the con-flict lies in their ability to portray their loss of human lives at the hands of Assad’s gov-ernment. This narrative, if successfully circulated, should draw sympathy from the in-ternational community in the form of military or material support (Berman and Matanock 2015; Humphreys & Weinstein 2008). When an avalanche of visual and audio-visual ma-terial documenting the consequences of the chemical attack on Syrian civilians surfaced in 2013, the Syrian conflict became an issue abroad. Western parliaments started debating whether or not to resort to military intervention in Syrian (Cozma & Kozman, 2015; Kaarbo & Kenealy, 2016). One of the central questions leading the debate was to deter-mine who was responsible for the chemical attack: the Assad regime or the opposition. A following UN investigation caused disagreement and polarised reactions around the world (Cozma & Kozman, 2015). From a political communication perspective, this gives us the opportunity to test theories of exposure to counter-attitudinal information and political heuristics.

METHODOLOGY Design

The hypotheses were tested via an experiment administered through an online ques-tionnaire (included in Appendix). The experiment can be divided in three sections. First the participants were asked about their attitudes towards taking action to get As-sad out of power. Then, they were randomly assigned to a condition and shown a news item that either supports the government or the opposition (see Appendix 2 for stimuli). A hidden timer was placed on the stimulus page to record the amount of time spent by each participant on the stimulus. After the manipulation check, the partici-pants were asked to answer the same battery of questions about their attitudes for a second time, leaving the socio-demographic details to the end of the questionnaire.

Prior to the experiment, a pre-test was conducted to choose one stimulus pair among three possible pair options (1a/1b, 2a/2b, 3a/3b). The stimuli pairs were news items, each pair presenting a different event in the Syrian conflict.The aim was that the pair items (a/b) present narratives that are as antagonistic as possible for each event (1, 2, 3), in other words, that they clearly show bias towards the government or the oppo-sition. Six oneway ANOVAs were conducted with sympathy as the grouping variable (pro-government, neutral, pro-opposition). Three of them were conducted with expect-edness of the stimulus as a dependent variable in order to gauge the extent to which the stimulus presented counter-attitudinal information. The other three ANOVAs were conducted with the credibility of the stimulus as dependent variable. The results are summarized in Table 1 of the appendix. The stimulus pair 1a/1b was selected as it was the stimulus that showed the most significant difference for expectedness between

groups while maintaining no significant difference in credibility (see Table 1 of the appendix for values).

Sample

The sample was a convenience sample recruited via social media. I received a total of 197 responses, out of which 87 were deleted for being incomplete. Two more respons-es were excluded for not having accepted the consent form and one more rrespons-esponse was excluded because the respondent entered his name in the open-ended questions intend-ed to measure recollection of information presentintend-ed. No straight-lining behaviour was detected. The last step in the data cleaning process was to exclude outliers in duration time. The average time taken to complete the survey was 737 seconds, which is about 12 minutes (SD = 448.77, or about 7.5 minutes). The six responses that took more than 2387 seconds (approximately 40 minutes, according to the extreme values flagged by SPSS) to complete the questionnaire were excluded leaving us with a final sample of N = 101 responses. Out of those 101 respondents, 39 were female, 61 were male and one respondent did not indicate their gender. 83 respondents were highly educated (had a Bachelor’s diploma or higher) while the other 18 respondents either held a high school or professional diploma. Due to the considerable number of Egyptian respondents, a new variable was created to identify Egyptian from non-Egyptian respondents. There were 47 Egyptian respondents in the sample and 54 non-Egyptian respondents. An analysis of the measures of central tendency and dispersion showed that respondents were on average 41.5 years old (SD = 14.12), while the median age was 45 years, and the most represented age was 55 years, the mode which occurred 14 times.

Counter-attitudinal treatment variable

This variable was computed by first recoding Q9_1 “In the Syrian conflict, who do you sympathise with more?” which was initially an 8-point scale with no mid-point (0,1,3,4 = “pro-government"; 6,7,9,10 = “pro-opposition”) into a dichotomous variable labelled side favored (1 = “pro-opposition”; 2 = “pro-government”). Then, a new vari-able counter-attitudinal treatment was computed based on whether the side favored matched the condition (1 = “pro-opposition treatment”; 2 = “pro-government treat-ment”). If the side favored matched the condition, then the respondent was considered to have received a pro-attitudinal treatment (counter-attitudinal treatment = 0). If the side favored did not match the condition, then the respondent was considered to have received a counter-attitudinal treatment (counter-attitudinal treatment = 1). After com-puting this new variable, it appeared that 57 respondents received a pro-attitudinal treatment and 44 participants received a counter-attitudinal treatment.

Attitude amplification variable

Participants were shown a battery of six questions attempting to measure their attitude towards taking action against the Syrian president Assad twice, once before the stimu-lus and once after. The questions were measured on a 7-point Likert scale (1 = Strong support for action against Assad; 7 = Strong opposition to action against Assad) and were based on the research by Clemons, Peterson and Palmer (2016). The attitude am-plification variable is operationalised as the difference between the scores prior to and post exposure to the stimulus. Before computing this variable, a reliability analysis on the six items in question was conducted. The reliability analysis showed that the scale

was very reliable (Cronbach’s alpha = .90) and that there was no room for improve-ment if any item was deleted.

Then, two scales, prior attitude sum and post attitude sum, were constructed based on the sum of scores on the respective batteries of questions. For those new variables, scoring the minimum of 6 points (scoring the minimum 1, on each of the six questions) meant that the respondent showed a “strong support to removing Assad” and scoring the maximum of 42 points (7 points on each question) indicated that the respondent showed a “strong opposition to removing Assad”.

Finally, the variables prior attitude sum and post attitude sum were used to compute the attitude amplification variable, understood as the number of points gained post exposure to the treatment, in the same direction as the prior attitude. As such, for participants that are initially rather supportive of the removal of Assad from power (prior attitude sum ranging from 6 to 24 points), then the attitude amplification would be computed as: prior attitude sum - post attitude sum. In this case, an attitude amplifi-cation score of +1 would mean that the respondent scored even lower in the post expo-sure questions than the prior expoexpo-sure questions by one point, or in other words, that he became more supportive of the removal of Assad from power after seeing the stim-ulus (closer to the minimum of 6 points in total) by one point in total. As for partici-pants that are initially rather opposed to the removal of Assad from power (prior atti-tude sum ranging from 25 to 42 points), then the attiatti-tude amplification would be com-puted as: post attitude sum - prior attitude sum. In this case, an attitude amplification score of +1 would mean that the respondent scored even higher in the post exposure questions than the prior exposure questions by one point, or in other words, that he

be-came more opposed of the removal of Assad from power after seeing the stimulus (closer to the maximum of 42 points) by one point in total.

The maximum score for attitude amplification observed among the sample was +8 and the maximum score for attitude change observed among the sample — understood as the number of points gained post exposure to the treatment, in the opposite direction as the prior attitude — was -8. Among the sample, 34 participants showed no attitude change or amplification (attitude amplification = 0). For more descriptives, refer to Table 2 of the Appendix.

Prior attitude strength variable

This variable was computed based on the scores on item Q9_1 of the survey available in the Appendix 3 which is an 8-point Likert scale with no mid-point (1 = strong sym-pathy for the government"; 8 = strong symsym-pathy for the opposition). For the prior atti-tude strength variable, the “weaker” attiatti-tudes, values 4 and 5 — which accounted for 33,7 percent of the sample — were recoded as 0 “weak prior attitude”. For the more moderate attitudes, values 3 and 6 — which accounted for 37,8 percent of the sample — were recoded as 1 “moderate prior attitude”. And finally, for the stronger prior atti-tudes, values 1, 2, 7 and 8 — which accounted for 27,7 percent of the sample — were recoded as 2 “strong prior attitude”.

Involvement

Involvement was measured using a battery of three questions (see Q5 in the survey in Appendix 3) each asking the degree of involvement in following the Syrian conflict on the news on a 6-point Likert scale with no mid-point (1=highest involvement,

6=low-est involvement). After reverse coding the variables so that an increase in points equates an increase in involvement, a scale was constructed by computing the mean score for all three recoded questions. A reliability analysis showed that the scale was very reliable (Cronbach’s alpha = .91) and that there was no room for improvement if any item was deleted. For more descriptives, refer to Table 2 of Appendix 1.

Systematic & heuristic processing

To measure the extent and type of information processing the respondents devoted to the stimulus, participants were asked to recollect information about the news item they were exposed to in four open-ended questions. Two of those questions were intended to measure attention to source cues (Q19 & Q22 of the questionnaire in Appendix 3) indicative of heuristic processing (Chaiken, 1980) and the other two were intended to measure attention to content cues (Q20 & Q21 of the questionnaire in Appendix 3) in-dicative of systematic processing (Chaiken, 1980). When participants were able to ful-ly recollect the information, they were awarded one point on the question, while a par-tial recollection was awarded 0.5 points, and an erroneous answer or no answer was awarded no points. A variable labelled content cues score was then computed as the sum of scores on both content questions, and a variable labelled source cues score was computed as the sum of the scores on the source cues questions. Additionally, because systematic processing is associated with more motivation to seek new information (Chaiken, 1980), a question asking the participant whether they would like to read the rest of the article was added (Q15 of the survey in Appendix 3). This was translated by the variable information seeking where the participant is only awarded 1 point if they answer that they would like to read the rest of the article immediately (1 = information

seeking present, 0 = information seeking not present). The last variable included to measure processing is the time the the participant took to examine the news item, the longer the time taken being associated with a more careful processing. As such, the scores of the variable processing time (number of seconds taken by each participant to press “next” after seeing the stimulus) was recoded into six equal groups. The groups were created according to the relevant percentiles and starting at a score of 0 and mov-ing up by 0.5 increments (0=quickest to dismiss the stimulus, 2.5=slowest to dismiss the stimulus).

Finally, the different processing routes indices were calculated using the following formulas, guided by Chaiken’s (1980) theory, as stated in the section above:

—> Heuristic processing index = source cues score + information seeking reverse coded

—> Systematic processing index = content cues score + information seeking + pro-cessing time recoded

For the descriptives associated with those indices, please refer to Table 2 of Appendix 1.

RESULTS

Randomization checks

To check whether the participants’ age was comparable over the conditions, a Oneway ANOVA was conducted with age as the dependent variable and the condition (1 = “pro-opposition treatment”; 2 = “pro-government treatment”) as independent variable. The ANOVA showed that respondent’s mean age in the pro-opposition treatment

con-dition (M = 39.4 years, SD = 14.5) did not significantly differ from the respondents’ mean age in the pro-government treatment condition (M = 43.8 years, SD = 13.5), F (1, 99) = 2.56, p = .11. Furthermore, Levene’s test revealed that the assumptions of homogeneity of variances was not violated (F = 1.624, p = .206), thus we can trust the p-value of the ANOVA. As such, participants’ age can be considered successfully ran-domly distributed across the conditions and should not bias the analyses.

As for the non-continuous socio-demographic variables gender, nationality and and educational attainment, chi square tests showed that all three variables were equal-ly distributed among the two conditions (see Table 3 of Appendix 1 for values).

Manipulation check

To check whether the participants had correctly perceived the bias in the news item they were exposed to, a oneway ANOVA was conducted with condition (pro-govern-ment treat(pro-govern-ment, pro-opposition treat(pro-govern-ment) as a grouping variable, and the scores on question Q16_1 of the questionnaire “Which side do you think this news items favours” (10-point Likert scale; 0=the government, 10=the opposition) as a dependent variable. The results show that the evaluation of the pro-government treatment (M = 7.39, SD = 2.62) significantly differed from the evaluation of the pro-opposition treat-ment (M = 4.22, SD = 3.57) F(1, 99) = 25.963, p = .000. As such, the manipulation can be considered successful.

Testing H1

For individuals that hold stronger prior attitudes, counter-attitudinal messages will am-plify initial attitudes compared to individuals that hold weaker prior attitudes.

To test the first hypothesis, a two-way ANOVA was conducted, with prior attitude strength (weak, moderate, strong) and counter-attitudinal treatment (pro-attitudinal, counter-attitudinal) as independent variables and attitude amplification as dependent variable. The assumption of equal variances between groups has been met (F = 1.197, p = .317). Examining the plot showing the interaction between prior attitude strength and exposure to counter-attitudinal information on attitude amplification (Figure 5, Appendix 1), does not appear promising for the hypothesis. The graph shows that par-ticipants that hold moderate prior attitudes show more attitude amplification when ex-posed to counter-attitudinal information than participants who hold stronger initial atti-tudes. Indeed, the findings did not show that the main effect of counter-attitudinal treatment on attitude amplification was sufficiently significant F(1, 517) = 3.816, p = . 054, eta2 _{= .038. The effect of prior attitude strength on attitude amplification was also}

not shown to be significant F(2, 517) = .352, p = .704, eta2 _{= .007. And the interaction}

of prior attitude strength and counter-attitudinal treatment on attitude amplification was also not significant F(2, 517) = .712, p = .493, eta2 _{= .014.}

Testing H2a

Individuals with higher levels of involvement in an issue are more likely to process counter attitudinal information systematically compared to individuals will lower lev-els of involvement.

To test this hypothesis, a two-way ANOVA was conducted, with involvement recoded into three groups according to the 33rd and 66th percentiles (low, moderate, high) and counter-attitudinal treatment (pro-attitudinal, counter-attitudinal) as independent vari-ables and systematic processing index as dependent variable. Looking at the graph of

estimated marginal means (Figure 6 of Appendix 1), it appears that participants highly involved in following the Syrian conflict generally processed counter-attitudinal in-formation considerably less systematically than pro-attitudinal inin-formation. Whereas their less involved counterparts appeared to devote more systematic processing to counter-attitudinal information than pro-attitudinal information. This is the opposite of what the hypothesis predicted. The statistical evidence shows a significant but small main effect of exposure to counter-attitudinal information F(1, 95) = 4.715, p = .032 , eta2 _{= .047 on systematic processing. The main effect involvement was not found to be}

significant F(2, 95) = 1.386, p = .255 , eta2 _{= .028. However, the interaction visible in}

the plot (Figure 6 of Appendix 1) was found to have a medium and significant effect on the dependent variable F(2, 95) = 3.471, p = .035 , eta2 _{= .068. A more detailed}

in-terpretation is provided in the discussion.

Testing H2b

Individuals with lower levels of involvement in an issue are more likely to process counter attitudinal information heuristically compared to individuals will higher levels of involvement.

To test this hypothesis, a two-way ANOVA was conducted, with involvement (low, moderate, high) and counter-attitudinal treatment (pro-attitudinal, counter-attitudinal) as independent variables and heuristic processing index as dependent variable. How-ever, in this case, neither one of the main effect exposure to counter-attitudinal infor-mation F(1, 95) = .980, p = .325, eta2 _{= .010, nor involvement F(2, 95) = .208, p = .}

Similarly, their interaction also did not seem to significantly affect their use of heuris-tic processing F(2, 95) = .963, p = .386 , eta2 _{= .020.}

Testing H3a & H3b

For individuals with higher levels of involvement, the effect of exposure to counter attitudinal messages on attitude change is mediated by systematic processing and H3b: For individuals with lower levels of involvement, the effect of exposure to counter atti-tudinal messages on attitude change is mediated by heuristic processing

To test the aforementioned hypotheses, we used Hayes’ PROCESS macro Model 7 (Hayes, 2013), which allows testing of a moderated mediation relationship (with pos-sibility for multiple mediators) with bootstrap confidence intervals for an indirect ef-fect (see Figure 8 of Appendix 1 for values).

Hypothesis 3a predicted that systematic processing moderated by involvement would mediate the relationship between exposure to counter-attitudinal information and attitude change (See Figure 4 of Appendix 1 for conceptual model). Results showed that the model generated bias corrected 95% bootstrap confidence intervals for the indirect effects using 5,000 bootstrap samples.

Notably, the confidence intervals surrounding the indirect effect of stress did span zero, which indicates that no significant indirect effect has been found at high levels of involvement (β = .0081, 95% Conf. Interval: -.1387 to .1301), moderate levels of volvement (β = -.0290, 95% Conf. Interval: -.1254 to .2550), and high levels of in-volvement (β = -.0660, 95% Conf. Interval: -.2899 to .5190). As zero is present in the confidence intervals, the results show no evidence of conditional indirect effect which is different from zero with 95% confidence interval. Therefore, the association

be-tween exposure to counter-attitudinal information and attitude change through system-atic processing does not significantly increase when an increase in involvement occurs. Consequently, hypothesis 3a is not supported.

Hypothesis 3b predicted that heuristic processing moderated by involvement would mediate the relationship between exposure to counter-attitudinal information and attitude change (See Figure 4 of Appendix 1 for conceptual model). Results showed that the model generated bias corrected 95% bootstrap confidence intervals for the indirect effects using 5,000 bootstrap samples.

Notably, the confidence intervals surrounding the indirect effect of stress did span zero, which indicates that no significant indirect effect has been found at low levels of involvement (β = .0279, 95% Conf. Interval: -.4634 to .4364), moderate levels of volvement (β = .0916, 95% Conf. Interval: -.2128 to .3770), and high levels of in-volvement (β = .1553, 95% Conf. Interval: -.2762 to .6126). As zero is present in the confidence intervals, the results show no evidence of conditional indirect effect which is different from zero with 95% confidence interval. Therefore, the association be-tween exposure to counter-attitudinal information and attitude change through heuris-tic processing does not significantly increase when a decrease in involvement occurs. Consequently, hypothesis 3b is not supported.

DISCUSSION AND CONCLUSION

As news consumption is becoming increasingly electronic, more and more heteroge-neous narratives are flooding the online media scene (Powers & O’Loughlin, 2015). This study attempted to provide a theoretical examination of the cognitive inner

work-ings of individuals exposed to dissonant information using the Syrian conflict as a case study. Previous research investigated the effect of exposure the heterogenous informa-tion but without always operainforma-tionalizing heterogeneity in a harmonious manner (Kim & Chen, 2015). This is why I wanted to be specific in the operationalization of hetero-geneity. More specifically, since prior scholarship had flagged unexpected information as a determinant of the processing route used (heuristic vs. systematic) (Smith & Petty, 1996s), I wanted to operationalize heterogeneity as counter-attitudinal information, because I believe this definition implies a certain expectation due to prior exposure. As such, the aim of this study was to construct an integrated theoretical framework that combines the potential effects of exposure to counter attitudinal information on differ-ential processing routes known to affect attitude change.

Assuming that individuals holding strong attitudes about a specific issue have arrived to this strong attitude due to repeated exposure to arguments for and against their stance (Bullock, 2007), the first hypothesis (H1) posited that: for individuals that hold stronger prior attitudes, counter-attitudinal messages will amplify initial attitudes compared to individuals that hold weaker prior attitudes. However, hypothesis H1 was not supported. Not only did the analysis of variance conducted in the previous section show that individuals holding stronger prior attitudes were not significantly more like-ly to consolidate their prior stance compared to their less opinionated counterparts, but also, the estimated marginal means plot (Figure 5, Appendix 1) appeared to move in an unexpected direction. Although not significant, the results rather showed that the group most likely to consolidate their initial stance after facing counter-attitudinal informa-tion were those holding moderate views, not strong views. The insignificant results and the direction of the effect could potentially be explained by the data collection method

of the prior attitude strength variable. The variable in question has been inferred by recoding the values obtained on an 8-point slider, with no mid-point, asking the partic-ipant which side they favoured. As such, it is important to acknowledge that some par-ticipants, maybe a considerable amount of them, did not have an opinion on the con-flict and were forced to pick a side by the lack of availability of a neutral option. The decision to force an answer was taken after examining the results from the pretest and realising that a considerable amount of data was “lost” to mid-point values. However, the findings from this experiment point to the possibility that the backfire effect ex-plained in the literature (Bullock, 2007; Nyhan & Reifler, 2010) may not have been observed because the distribution of responses amongst the recoded groups was not highly representative of their true attitudes.

The second pair of hypotheses take a closer look at the relationship between the exposure to information that contradicts one’s opinion and how carefully or superfi-cially one processes this kind information, depending on how involved they are in the issue. The second hypothesis which posited that individuals with higher levels of in-volvement in an issue were more likely to process counter attitudinal information sys-tematically compared to individuals will lower levels of involvement, H2a was not supported. While the interaction between involvement and counter-attitudinal infor-mation on systematic processing was found to be significant (see Results section), the interaction was observed in the opposite direction than predicted by the hypothesis (Figure 6, Appendix 1). Individuals that were most involved in following the Syrian conflict appear to carefully analyze counter-attitudinal significantly less often than in-dividuals that did not consider themselves involved in following the conflict. This can be explained by the familiarity principle simply stating that individuals that are already

familiar with a specific topic do not feel the need to allocate considerable cognitive resources to process the message (Garcia-Marques & Mackie, 2001). It is possible that individuals that rated themselves as highly involved in following the news concerning the Syrian conflict were already familiar with the issue and its pro and con arguments, as such they did not deem it necessary to scrutinize already-known arguments. Perhaps the type of motivation can also explain the discrepancy between the observed and ex-pected findings. According to this motivated reasoning theory (Kunda, 1990), an indi-vidual’s approach to processing information can be guided by two types of goals: ac-curacy goals and directional goals. Individuals concerned with acac-curacy are typically equally open to disconfirming or confirming both pro-attitudinal and counter-attitudi-nal information, in an attempt to get as close as possible to “the truth”. In contrast, in-dividuals satisfying directional goals attempt to reduce dissonance with their own prior beliefs (intrinsic motivation) or with what they believe is expected from them in such situation (impression or extrinsic motivation). While accuracy-seeking individuals might not be motivated to carefully process the familiar information, defence-motivat-ed individuals might display signs if systematic processing while they sift through their counter-arguments repertoire (Bullock, 2007). Therefore it is possible that the hy-pothesis was based on research that failed to differentiate between the motivational goals of their subjects.

The third hypothesis which expected that individuals with lower levels of in-volvement in would be more likely to process counter attitudinal information heuristi-cally compared to individuals with higher levels of involvement, H2b, was also not supported. Although not significant this time, the results show a similar deviation from the expectation as observed in the previous hypothesis. It appears that in order to

process counter-attitudinal information, highly involved individuals could resort to heuristic cues more often than less involved individuals (Figure 7, Appendix 1). Even though it is important to keep in mind that this effect was not significantly proven, this tendency would go hand in hand with the unexpected results from the H2a. In fact, as-suming that the more politically involved are also likely to be the more politically knowledgable, research has shown that the political experts very often make use of their knowledge to employ heuristic shortcuts when making political decisions (Lau & Redlawsk, 2001). Moreover, because of their expertise they are able to do so without compromising on the quality of their decisions (Lau & Redlawsk, 2001).

The final part of the study attempted to integrate the effects of exposure to counter-attitudinal information (x) on attitude change (y) via two moderated media-tions (m1, m2). The first effect being mediated by systematic processing (m1) which is moderated by involvement (w), and the second effect being mediated by heuristic pro-cessing (m2), also moderated by involvement (w) (see conceptual model in Figure 4, Appendix 1). Contrary to the expectations, none of the two moderated mediations (H3a & H3b) were found to be significant (see Results section).

LIMITATIONS AND FUTURE RESEARCH

The discrepancy between the effects observed in the literature and the lack of corrobo-ration in this study could be due to several shortcomings that need to be addressed.

Starting with the most obvious, the high dropout rate severely constrained the sample size. However, I believe the main issue can be attributed to the measurement of the central concepts of systematic and heuristic processing. Upon consulting the

litera-ture, it appeared that thought-listing (Cacioppo, Harkins, & Petty, 1981; Cappella, Price & Nir, 2010) and argument quality (Cyr, Lim, Head, & Stibe, 2015) were the most reliable techniques to measure cognitive processing style. And while manipulat-ing the argument quality would have been an arguably more reliable manner of testmanipulat-ing systematic processing, this strategy would have required the stimulus to become longer in text length, and because of the additional manipulation it would have required a bigger sample size. Unfortunately, given the already elevated drop-out rate for a head-line stimulus, it is quite probable that if I had lengthened the news item, the completed response rate would have been even lower when the sample size should have been re-quired to become even higher. The same argument can be made for the thought-listing technique. Unfortunately I did not possess the resources necessary to provide partici-pants with incentive to engage in a lengthy and daunting process such as thought-list-ing. Although this technique would have most likely provided a more reliable way of measuring both systematic and heuristic processing as well as even gone one step fur-ther and coded for potential indices of defensive or accuracy motivated arguments.

In order to better investigate the effects of exposure to counter-attitudinal infor-mation and political issue involvement on processing route and the resulting attitude change, future research should employ techniques that are more sensitive to measuring the differential processing mechanisms and the motivations behind them. Especially, given the counter-intuitive and unexpected finding that the more politically involved were significantly less likely than the less politically involved to carefully analyze in-formation that contradicts their views. A phenomenon potentially attributable to their reliance or heuristic shortcuts instead. Parallel to that is the literature pointing to heuristic processing as being less effective than systematic processing at achieving

long-lasting attitude change (Chaiken, 1980). As such, expanding this line of research and combining it with motivated reasoning would help us gain a better insight on the conditions under which more or less engaged citizens are willing to update their stances on political matters.

BIBLIOGRAPHY

Allcott, H., & Gentzkow, M. (2017). Social media and fake news in the 2016 election. Journal of Economic Perspectives, 31(2), 211-36.

Barker, D. C., & Hansen, S. B. (2005). All things considered: Systematic cognitive processing and electoral decision‐making. Journal of Politics, 67(2), 319-344. Baum, M. A. (2003). Soft news and political knowledge: Evidence of absence or

ab-sence of evidence?. Political Communication, 20(2), 173-190.

Berman, E., & Matanock, A. M. (2015). The empiricists' insurgency. Annual Review of Political Science, 18, 443-464.

Bohner, G., Moskowitz, G. B., & Chaiken, S. (1995). The interplay of heuristic and systematic processing of social information. European review of social psy-chology, 6(1), 33-68.

Bullock, J. G. (2007). Experiments on partisanship and public opinion: Party cues, false beliefs, and Bayesian updating (Doctoral dissertation, Stanford Universi-ty).

Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39(5), 752.

Chaiken, S. (1987). The heuristic model of persuasion. In Social influence: the ontario symposium (Vol. 5, pp. 3-39).

Christin, T., Hug, S., & Sciarini, P. (2002). Interests and information in referendum voting: An analysis of Swiss voters. European journal of political research, 41(6), 759-776.

Chung, S., & Waheed, M. (2016). “Biased” Systematic and Heuristic Processing of Politicians’ Messages: Effects of Source Favorability and Political Interest on Attitude Judgment. International Journal of Communication, 10, 20.

Clemons, R. S., Peterson, R. D., & Palmer, C. L. (2016). Priming Assad: An experi-ment of ethnic priming and attitudes toward military action in Syria. Foreign Policy Analysis, 12(3), 433-449.

Cozma, R., & Kozman, C. (2015). The Syrian crisis in the news: How the United States' elite newspapers framed the international reaction to Syria's use of chemical weapons. Journalism Practice, 9(5), 669-686.

Dutta-Bergman, M. J. (2005). Civil society and public relations: Not so civil after all. Journal of Public Relations Research, 17(3), 267-289.

Ferree, M. M., Gamson, W. A., Gerhards, J., & Rucht, D. (2002). Four models of the public sphere in modern democracies. Theory and Society, 31(3), 289–324. https://doi.org/10.1023/A:1016284431021

Flaxman, S., Goel, S., & Rao, J. M. (2016). Filter bubbles, echo chambers, and online news consumption. Public Opinion Quarterly, 80(Specialissue1), 298–320.

https://doi.org/10.1093/poq/nfw006

Gilens, M. (2001). Political ignorance and collective policy preferences. American Po-litical Science Review, 95(2), 379-396.

Hampton, K. N., Shin, I., & Lu, W. (2017). Social media and political discussion: when online presence silences offline conversation. Information Communica-tion and Society, 20(7), 1090–1107. https://doi.org/10.1080/1369118X. 2016.1218526

Hayes, A. F. (2013). Mediation, moderation, and conditional process analysis. Intro-duction to Mediation, Moderation, and Conditional Process Analysis: A Re-gression-Based Approach edn. New York: Guilford Publications, 1-20.

Hiebert, R. E. (2003). Public relations and propaganda in framing the Iraq war: A pre-liminary review. Public Relations Review, 29(3), 243-255.

Holbert, R. L., Garrett, R. K., & Gleason, L. S. (2010). A new era of minimal effects? A response to Bennett and Iyengar. Journal of Communication, 60(1), 15-34. Humphreys, M., & Weinstein, J. M. (2008). Who fights? The determinants of

partici-pation in civil war. American Journal of Political Science, 52(2), 436-455. Jost, J. T., Krochik, M., Gaucher, D., & Hennes, E. P. (2009). Can a psychological

the-ory of ideological differences explain contextual variability in the contents of political attitudes?. Psychological Inquiry, 20(2-3), 183-188.

Kaarbo, J., & Kenealy, D. (2016). No, prime minister: Explaining the House of Com-mons’ vote on intervention in Syria. European Security, 25(1), 28-48.

Kim, Y., & Chen, H. (2015). Discussion Network Heterogeneity Matters : Examining a Moderated Mediation Model of Social Media Use and Civic Engagement, 9, 2344–2365.

Kull, S., Ramsay, C., & Lewis, E. (2003). Misperceptions, the media, and the Iraq war. Political Science Quarterly, 118(4), 569-598.

Kunda, Z. (1990). The case for motivated reasoning. Psychological bulletin, 108(3), 480.

Lynch, M., Freelon, D., & Aday, S. (2014). Syria’s socially mediated civil war. United States Institute of Peace, 91(1), 1-35.

Lu, Y., Heatherly, K. A., & Lee, J. K. (2016). Cross-cutting exposure on social net-working sites: The effects of SNS discussion disagreement on political partici-pation. Computers in Human Behavior, 59, 74–81. https://doi.org/10.1016/ j.chb.2016.01.030

Luther, C. A., & Miller, M. M. (2005). Framing of the 2003 US-Iraq war demonstra-tions: An analysis of news and partisan texts. Journalism & Mass Communica-tion Quarterly, 82(1), 78-96.

Mcclurg, S. D. (2006). Networks on Political Participation. Political Science, 50(3), 737–754.

McLeod, J. M., Scheufele, D. A., Moy, P., Horowitz, E. M., Holbert, R. L., Zhang, W., ... & Zubric, J. (1999). Understanding deliberation: The effects of discus-sion networks on participation in a public forum. Communication Research, 26(6), 743-774.

Nir, L. (2011). Disagreement and opposition in social networks: Does disagreement discourage turnout?. Political Studies, 59(3), 674-692.

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

Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In Communication and persuasion (pp. 1-24). Springer. New York.

Powers, S., & O’Loughlin, B. (2015). The Syrian data glut: Rethinking the role of in-formation in conflict. Media, War & Conflict, 8(2), 172-180.

Scheufele, D. A., Nisbet, M. C., Brossard, D., & Nisbet, E. C. (2004). Social structure and citizenship: Examining the impacts of social setting, network

heterogene-ity, and informational variables on political participation. Political Communi-cation,, 21(3), 315-338.

Schuck, A. R. T., Vliegenthart, R., & De Vreese, C. H. (2014). Who’s Afraid of Con-flict? the Mobilizing Effect of Conflict Framing in Campaign News. British Journal of Political Science, 46(1), 177–194. https://doi.org/10.1017/ S0007123413000525

Schwalbe, C. B., Keith, S., & Silcock, B. W. (2004). Embedded images: The framing of the first visuals of the US war in Iraq. Southwestern Mass Communication Journal, 19(2), 61-72.

Shi, Y. (2016). Cross-cutting Messages and Voter Turnout: Evidence from a Same-Sex Marriage Amendment. Political Communication, 33(3), 433–459. https:// doi.org/10.1080/10584609.2015.1076091

Sides, J., & Citrin, J. (2007). European opinion about immigration: The role of identi-ties, interests and information. British journal of political science, 37(3), 477-504.

Strömbäck, J. (2005). In search of a standard: Four models of democracy and their normative implications for journalism. Journalism Studies, 6(3), 331–345. https://doi.org/10.1080/14616700500131950

Torcal, M., & Maldonado, G. (2014). Revisiting the dark side of political deliberation: The effects of media and political discussion on political interest. Public Opin-ion Quarterly, 78(3), 679–706. https://doi.org/10.1093/poq/nfu035

Vaccari, C. (2013). From echo chamber to persuasive device? Rethinking the role of the Internet in campaigns. New Media and Society, 15(1), 109–127. https:// doi.org/10.1177/1461444812457336

APPENDIX 1: TABLES AND FIGURES Figure 1

Conceptual Model For Hypothesis 1

Figure 2

Conceptual Model For Hypothesis 2a

Figure 3

Conceptual Model For Hypothesis 2b


Prior attitude strength

Involvement

Counter-att info Heuristic

process-ing Involvement

Figure 4

Conceptual Model For Hypotheses 3a and 3b


 Involvement Heuristic process-ing Systematic pro-cessing

Table 1

ANOVA summary: Grouping variable sympathy (pro-government, neutral, pro-opposi-tion), dependent variables listed below (credibility and expectedness of stimulus).

df F p Credibility of stimulus 1a 2 1.351 .278 Expectedness of stimulus 1a 2 3.200 .059 Credibility of stimulus 1b 2 .497 .614 Expectedness of stimulus 1b 2 .372 .693 Credibility of stimulus 2a 2 3.883 .035 Expectedness of stimulus 2a 2 3.827 .036 Credibility of stimulus 2b 2 2.218 .129 Expectedness of stimulus 2b 2 .124 .884 Credibility of stimulus 3a 2 2.533 .100 Expectedness of stimulus 3a 2 .339 .716 Credibility of stimulus 3b 2 .181 .836 Expectedness of stimulus 3b 2 .131 .878

Table 2

Descriptives for main variables

Table 3

Chi Square tests for randomization checks on the equa

M SD

Attitude amplification (-8 = high attitude change; 0 = no attitude change; 8 = high attitude

amplification)

-.26 2.34

Involvement

(1 = not involved at all; 6 = very involved) 4.23 1.20 Heuristic processing (HP) index (0 = no evidence of HP; 3 = high evidence of HP) 1.07 .70 Systematic processing (SP) index (0 = no evidence of SP; 5.50 = high evidence of SP) 2.41 1.49 df N chi squared p Gender 1 100 .00 .964 Nationality 1 101 .08 .770 Educational attainment 1 100 3.32 .344

l distribution of gender, nationality and educational attainment over the two experi-mental conditions


Figure 5

H1: Graph showing the interaction between prior attitude strength and exposure to counter-attitudinal information on attitude amplification

Figure 6

H2a: Graph showing the interaction between involvement and exposure to counter-attitudinal information on activation of systematic processing

Figure 7

H2a: Graph showing the interaction between involvement and exposure to counter-attitudinal information on activation of heuristic processing

Figure 8

H3a & H3b: Results of the moderated mediation model of exposure to counter-attitu-dinal information, involvement, processing route and attitude change

APPENDIX 2: STIMULUS SAMPLE

APPENDIX 3: QUESTIONNAIRE

Heidi Main Experiment

Start of Block: Factsheet

Q1 Thank you for agreeing to take part in this research for my political communication thesis at Graduate School of Communication, a part of the University of Amsterdam.

The study will take about 10 minutes. In this study, I will first ask you a few question about your thoughts towards a certain political topic. Then I will show you a news item followed by a few questions to inquire about your opinion on the news item.

As this research is being carried out under the responsibility of the Amsterdam School of Communication Research (ASCoR), University of Amsterdam, we can guarantee that:

1) Your anonymity will be safeguarded, and that your personal information will not be passed on to third parties other than researchers under any conditions, unless you first give your clear permission for this.

2) You can refuse to participate in the research or cut short your participation without having to give a reason for doing so. You also have up to 7 days after participating to withdraw your permission to allow your answers or data to be used in the research.

3) Participating in the research will not entail you to any appreciable risk or discomfort, and you will not be exposed to any explicitly offensive material.

4) No later than five months after the conclusion of the research, we will be able to provide you with a research report that explains the general results of the research.

For more information about the research and the invitation to participate, you are welcome to contact me at heidi.hindam@student.uva.nl at any time.

Should you have any complaints or comments about the course of the research and the proce-dures it involves as a consequence of your participation in this research, you can contact the designated member of the Ethics Committee representing ASCoR, at the following address: ASCoR Secretariat, Ethics Committee, University of Amsterdam, Postbus 15793, 1001 NG Amsterdam; 020‐525 3680; ascor‐secr‐fmg@uva.nl.

Any complaints or comments will be treated in the strictest confidence.

I hope that I have provided you with sufficient information. I would like to take this opportu-nity to thank you in advance for your assistance with this research, which I greatly

appreciate.

Kind regards, Heidi Hindam

End of Block: Factsheet

Start of Block: Informed consent

Q2 I hereby declare that I have been informed in a clear manner about the nature and method of the research, as described in the email invitation for this study.

I agree, fully and voluntarily, to participate in this research study. With this, I retain the right to withdraw my consent, without having to give a reason for doing so. I am aware that I may halt my participation in the experiment at any time.

If my research results are used in scientific publications or are made public in another way, this will be done such a way that my anonymity is completely safeguarded. My personal data will not be passed on to third parties without my express permission, except for scientific re-search purposes.

If I wish to receive more information about the research, either now or in future, I can contact Heidi Hindam at heidi.hindam@student.uva.nl. Should I have any complaints about this re-search, I can contact the designated member of the Ethics Committee representing the AS-CoR, at the following address: ASCoR secretariat, Ethics Committee, University of Amster-dam, Postbus 15793, 1001 NG Amsterdam; 020-525 3680; ascor‐secr-fmg@uva.nl.

I understand the text presented above, and I agree to participate in the research study. (1) I do not agree to participate in the research study. (2)

Skip To: End of Survey If I hereby declare that I have been informed in a clear manner about the nature and method of the r... = I do not agree to participate in the research study.

End of Block: Informed consent

Start of Block: Intro to knowledge questions

Q3 In the following section, you will be asked a few questions to assess your knowledge and opinion about the Syrian conflict.

End of Block: Intro to knowledge questions Start of Block: Involvement

Q4 Which one of these names is the name of the current Syrian president? (please avoid 'googling', it's okay if you do not know for sure)

Muammar Gaddafi (1) Hosni Mubarak (2) Bashar al-Assad (3)

Zine el Abidine Ben Ali (4) I don't know (5)

Q5 Please indicate the extent to which you agree or disagree with the following statements

End of Block: Involvement Start of Block: Media trust

Q6 Please indicate the extent to which you agree or disagree with the following statements

End of Block: Media trust Start of Block: Attitudes_before

Q7 In the following sections, the parties opposing the Syrian government in the conflict are referred to as both "opposition/rebels" in order not to pick sides by including only one of the possible labels.

Q8 How much sympathy do you have for the following parties in this conflict?


Q9 In the Syrian conflict, who do you sympathise with more?

Q10 To what extent do you think that the involvement of Western countries in Syria was jus-tified/necessary?
 
 Completely justified (1) Justified (2) Somewhat justified (3) Undecided (4) Somewhat unjustified (5) Unjustified (6) Completely unjustified (7) I don't have an opinion (8)

Q11 How strongly do you support or oppose the following courses of action?

End of Block: Attitudes_before Start of Block: intro to stimulus

Q12 Now you will be shown an excerpt from a news report that you may or may not have already come across published online in December 2016.

End of Block: intro to stimulus Start of Block: Stimulus_a Q13 Q14 Timing First Click (1) Last Click (2) Page Submit (3) Click Count (4)

End of Block: Stimulus_a

Start of Block: Information seeking

Q15 Would you like to read the rest of the article? Yes, I would like to read the rest of the article now (1) Yes, send it to my email (2)

No, thank you (3)

End of Block: Information seeking Start of Block: Manipulation check

Q17 To what extent do you agree with the following statements?

End of Block: Manipulation check Start of Block: Processing

Q18 Now, you will be asked 4 questions about the news item you just saw. If you do not re-member the answer, please avoid googling it and just type "do not rere-member". Thank you.

Q19 Please write below the name of the organisation who conducted the investigation about the chemical attack (if you remember it, if not, just write "do not remember")

________________________________________________________________

Q20 Please write below the name of the Syrian city in which the chemical attack took place (if you remember it, if not, just write "do not remember")

________________________________________________________________

Q21 Please write below the name of gas used in the chemical attack (if you remember it, if not, just write "do not remember")

________________________________________________________________

Q22 Please write below the name (full name or abbreviation from the logo) of the news outlet the article came from (if you remember it, if not, just write "do not remember")

________________________________________________________________ End of Block: Processing

Q23 In light of the news item you just saw, please answer the following questions one last time.

Q24 To what extent do you agree with the following statements?

Q25 How much sympathy do you have for the following parties in this conflict?


Q26 In the Syrian conflict, who do you sympathise with more?

Q27 To what extent do you think that the involvement of Western countries in Syria was jus-tified/necessary?
 
 Completely justified (1) Justified (2) Somewhat justified (3) Undecided (4) Somewhat unjustified (5) Unjustified (6) Completely unjustified (7) I don't have an opinion (8)