Partisan Selective Exposure to Fake News Content

Selective exposure to political news in social media in Indonesia is escalating along with the increasing polarization of Indonesian people. This research aims to investigate: 1) differences in selective exposure to fake news content among incumbent and opposition supporters; 2) the association between critical thinking ability and partisans’ selective exposure. Repeated measures design was employed as the experiment design. Respondents were student activists of extra-campus organizations with particular political ideologies, who have pro-incumbent or pro-opposition preference. Seventy-one respondents were recruited, consisting of 34 incumbent (Jokowi) supporters and 37 opposition (Prabowo) supporters. Data was analyzed using independent t-test, a paired sample t-test, and correlational analysis. Results show that the opposition side was more inclined to demonstrate selective exposure by believing in fake news about their political enemy, compared to the incumbent supporters. This is shown by their tendency to believe and spread discrediting news about their political opponents rather than doing so for news which discredits their side. No association between critical thinking and partisan selective exposure was found. The implication of these results is that the critical point in debiasing is not necessarily predicated merely on analytical thinking ability but might also rest on one’s ability to think open-mindedly. Selective Exposure Partisan pada Konten-konten Berita Palsu


Introduction
Research on selective exposure to news coverage shows a relatively consistent finding that individuals tend to only select and validate news which confirms their In the political years ahead of the presidential election in Indonesia, selective exposure of political news was indicated by the increasing spread of news which demonstrated polarization between supporters of the presidential candidates. Since the 2014 election, people have been divided into two heavily polarized groups of supporters. Mietzner (2015) states, "no election since the end of authoritarianism had presented such stark alternatives as Prabowo and Jokowi." The two candidates represent a battle between two populists, where Jokowi is seen as the technocratic populist, and Prabowo is regarded as the ultra-populist. Contention between the two has been ongoing since the 2014 presidential election, and was extended up until the 2017 Jakarta gubernatorial election, in which Ahok, the then-governor of Jakarta, was seen as a representation of support for Jokowi, and Anies Baswedan, the current governor of Jakarta, was seen as support for Prabowo (Lim, 2017).
As a consequence of the stark polarization, post-truth political practices were born, characterized by an all-out support from volunteers, "buzzers," and celebrities, as well as vulgar support from mainstream mass media (Lim, 2017). Instead, of turning mainstream media or social media into a platform to gather actual information, die-hard supporters use the media as a place to express "freedom to hate" by legitimating their own group and excluding others (Lim, 2017). The practices of fake news are carried out by delegitimating the other group based on fake news, and the production and dissemination of fake news has become a political business network, which was uncovered by the Indonesian police (e.g., the Saracen case) (Chan, 2017).
As in social polarization, partisans on each side believe that any negative news concerning their group is an offense against them. Partisans tend to only trust information which supports things they already believe in. They are inclined to ignore truth beyond their own current belief system. A study by Knobloch-Westerwick and Kleinman (2012) showed that those who were affiliated with supporters of a party and wanted to win in an election tended to show support for media that confirmed their previously held beliefs. Barnidge et al. (2017) demonstrated that partisans who were affiliated with political activities were more likely to assume bias in general mass media if it broadcast news incompatible with their previously-held political beliefs, but regarded self-selected media or preferred media which was seen as representing their views, as not biased. Ditto et al. (2019) in their metanalysis noted that there is no difference in bias among individual ideology preferences when it comes to accepting political information which supports their stance. It implies that there is no political asymmetry between liberal and conservative groups, which distinguishes them in terms of selective exposure. Both groups are prone to bias. Nickerson (1998) argued that confirmation bias in selective exposure occurs in complex and ambiguous situations, as political situations often are. This might be due to failure to notice counterevidence or counterarguments of the pre-existing belief (Mercier & Sperber, 2011). In such situations, individuals only want to find supporting opinions and evidence for their believed claim and ignore negative arguments except to the extent that they have prepared a refutation of the negative argument (Mercier & Sperber, 2011).
A finding by Mahoney (1977), which was later highlighted by Fischhoff and Beyth-Marom (1983) and Nickerson (1998), reveals that scientists tend to remain uncritical even toward fictional research which supports dominant hypotheses in their discipline, as compared to research findings which go against dominant hypotheses. Shin and Thorson (2017) noted that, even during factchecking, partisans selectively choose news that supports their candidate of choice and which discredits the opponent. In the political context, it is analogous with the behavior of partisans who react differently toward fake news concerning political candidates that they support, and candidates on the opposite side. Fake news which defames the rival would be viewed favorably, while news discrediting candidates on their own side would be deemed as not credible.
In relation to the selective exposure phenomenon, one popular and intuitive statement holds that analytical and critical thinking can improve discernment of truth in the media and reduce bias. Pennycook and Rand (2018) said that, regardless of the group they belong to, individuals with the ability to think analytically will be more capable of distinguishing hoaxes from real news.
In critical thinking, individuals are asked to analyze assumption and bias, to avoid oversimplification, to take the interpretation of others into account, and to tolerate ambiguity (Wade, 1995). Beyer (1995) also highlighted that critical thinkers are those who are skeptical, openminded, value fair-mindedness, appreciate evidence and reasoning, and are capable of taking account of different points of view. Critical thinking ability is assumed to improve awareness, so that one could avoid emotional (Thagard, 2011) and irrational (Johnson & Blair, 2006) thinking, by deliberating and arguing in a correct manner.
However, a contradiction to this intuitive view exists, in which another study's finding shows that bias is not affected by critical thinking. Mercier and Sperber (2011) explain that reliance on rationality can instead trap an individual in a bias blind spot. Individuals will look for arguments using their rationality to justify their belief and behavior.
This research aims to study selective exposure to fake news content among supporters of Jokowi as the incumbent and of Prabowo as the opposition. In addition, it will also investigate the role of critical thinking in reducing selective exposure occurring among those partisans. Partisans' tendency to trace the source of fake news information will also be a part of the observation in this study. Fake news content of each candidate was intentionally chosen due to the voters' proclivity for getting stuck in ultimate attribution error and correspondence bias, when they focus on and exaggerate mistakes of the opposing side (Hogg & Vaughan, 2017). Also, people are more likely to be sensitive to negative information than positive information (Fiske, 1980;Hogg & Vaughan, 2017) This research posits several hypotheses and research questions, as follows: H1. There are response differences among partisans where supporters of both candidates will tend: (a) to believe and to spread fake news content concerning the opposite side; (b) not to believe or spread fake news content concerning the side that they support. H2. Critical thinking is correlated with the belief in fake news and tendency to spread fake news. RQ1. Will partisans have the tendency to trace the source of the fake news?

Methods
Instruments. Selective exposure is seen as the tendency to favor and validate news which supports existing beliefs, and in this case, is the participants' tendency to believe in negative news which discredits the opponent and not to believe negative news that dishonors the candidate they support. Additionally, the degree of proclivity for spreading news which supports their political preference will also be measured, in the sense that if participants' political preference says that the opposing candidate is bad, they will be more likely to spread the negative news. Participants were asked to choose one of two options: whether to support Jokowi (incumbent president) or Prabowo (opposition leader).
Operationally, participants were given six news links, comprising three links of fake news defaming the opposition side (i.e., Prabowo), and another three links of news discrediting the incumbent (i.e., Jokowi). They were then asked to respond to each link. The questions presented for each link were: 1) Is this news reliable? Give your rating from 1 to 9 (1 = very unreliable and 9 = very reliable); 2) Please give your rating from 1 to 9, on how important is it for this news to be spread and known to the public? (1 = very unimportant for the public to know about it and 9 = very important for the public to know about it). Our news item can be found in Supplementary Materials (SM).
Critical thinking ability was measured using an instrument formulated by Fajrianthi, Hendriyani, and Septarini (2016). This test measures the actual ability of general critical thinking, consisting of 37 items which comprise inference, recognition of assumption, deduction, interpretation, and evaluation aspects. Any correct answer to each item gives a score of 1 and any false answer gives a 0 score. If participants answer all questions correctly, a maximum score of 37 will be attained. Estimation of the internal reliability of this test using the KR-20 formula resulted in a coefficient of 0.57.
Data on fake news source tracing was obtained through an observation checklist in which a research assistant recorded if participants tracked the source of information by browsing on the link of each item of fake news. If they accessed the available link of any fake news, a score of 1 would be given, and if they did not, a score of 0 was given. The maximum score that participants could possibly attain if they accessed the link to the news discrediting the incumbent was 3, and likewise with the criteria applied browsing on links of the news concerning the opposition party. If participants did not access the links at all, a 0 score would be given.

Design and Participants.
Repeated measures design was employed as the research design, in which each participant, regardless of their affiliation and political preference, would receive the same treatment related to fake news. They would be exposed to fake news content concerning the candidate they supported and fake news content about the opposing party.
The variables being studied in this research were: selective exposure (dependent variable), manifested in two dimensions, namely belief and the tendency to spread fake news content to the public; presidential candidate preference, where participants indicated their support for the incumbent (Jokowi) or the opposition (Prabowo) (independent variable); and critical thinking ability (independent variable). Demographic data included were age, gender, political party preference, affiliation with student organization, and time spent online per day.
Recruitment of participants was conducted through snowballing. Inclusion criteria are politically active university students, preferring either Jokowi as the incumbent, or Prabowo as the opposition candidate for July 2019 ½Vol. 23 ½ No. 1 the presidential candidate, and evidence of joining extracampus student organizations with a political ideology. The names of the extra-campus organizations are classified. 25.4% of participants joined a puritan Islamic extra-campus organization, 32.4% joined a moderate Islamic organization, 40.8% were members of a nationalist Islamic organization, and the rest, 1.4%, joined a nationalist organization.
Data was collected between October 1st and November 15th of 2017, from 78 respondents who were members of extra-campus organizations in one public university in Surabaya. Eight respondents were excluded from analysis because the omitted to respond to some crucial items. Out of 71 analyzed respondents, 62% were males and the other 38 % were females. The age of respondents ranged from 18 to 23 years old (M = 20.49, SD = 1.06). All respondents were Muslim and students in bachelor programs in a public university in Surabaya. Thirty-four (47.9%) respondents were Jokowi supporters, while the remaining 37 (52.1%) respondents, were Prabowo supporters. 45.1% of respondents spent more than four hours a day online, 23.9% spent 3-4 hours a day online, 18.3% spent 2-3 hours a day, and the remainder spent fewer than two hours online per day.
All respondents identified with Islam. 63.4% reported religious nationalism as their preferred ideology, 14.1% reported being nationalistic, 8.5% wanted a nation based on religious law, 7% claimed to be a democratic socialist, and 1% chose not to say. 23.9% claimed that their interest was accommodated by PKS (the Prosperous Justice Party), 15.5% by PKB (the National Awakening Party), 8.5% by PDIP (the Indonesian Democratic Party of Struggle), 8.5% by PAN (the National Mandate Party), and the other 11.2% was spread among other parties (e.g., PPP, Gerindra, PBB, Nasdem, Partai Demoktrat, and others), while the rest, 32.4%, preferred not to say.
The majority of the incumbent's supporters did not access the three links to fake news about the opposition side (88.2%), nor did they access all links to fake news, discrediting the incumbent that they supported (88.2%). Only four out of 34 supporters of the incumbent accessed the provided links. The majority of the opposition supporters demonstrated similar behavior, where only three out of 37 participants accessed the links to fake news of the incumbent side, meaning 91.9% of participants did not access the links at all. Only two people of the opposition side accessed the links to fake news about their side, meaning the other 94.6%of them did not access the link to fake news concerning their own side.
Based on a hypothetical norm, critical thinking ability was classified into three categories, namely: low (x ≤ 12.34), moderate (12.34 < x ≤ 24.66), and high (x > 24.66). Twenty-eight respondents were identified as having a high level of critical thinking ability, while 43 of them had a moderate level of critical thinking ability. Based on the result of an independent sample t-test, both the incumbent side (M = 24.08; SD = 3.44) and the opposition side (M = 22.86; SD = 3.18) are on the same level of critical thinking ability (t (69) = 1.54, p = 0.127, Cohen's d = 0.36).
Research Procedure. The experiment was conducted individually, in which one participant would fill out the questionnaire in the presence of a research assistant at an agreed time. Participation was voluntary. A number of key contact persons affiliated with extra-campus student organizations were contacted and informed about the research. They were also offered the chance to join the research, by recommending some of their organization members. The experiment was conducted based on an agreement between research assistants and participants. Data collection was paper based. This research comprised three steps which were carried out consecutively at one time.
Step 1. Before starting the experiment, participants were informed about the procedure of the research and filled out a participation consent form.
Step 2. Partisans filled out demographic data and a questionnaire about presidential preference and were then asked to take the critical thinking ability test. This step took about 20 to 30 minutes.
Step 3. In relation to selective exposure, partisans were asked to respond to some news items, as follows: 1) Partisans read six headlines containing fake news about the incumbent and the opposition side, along with a preview of the news and the respective links to access the full articles of each headline. Provision of the links is expected to be a trigger for participants to trace the source of the news by browsing on the provided link. The research assistant would then fill out a checklist on whether participants had accessed the link of each news items using their smartphone; 2) Participants rated the degree to which they found each news item trustworthy from 1 to 9 (1 = very unreliable and 9 = very reliable); 3) They responded to an instruction saying "please rate from 1 to 9 how important it is for this news to be spread and known by the public? (1 = very unimportant for the public to know about it and 9 = very important for the public to know about it)." This step took from two to five minutes.

Results
To test Hypothesis 1, an independent sample t-test and paired sample t-test were conducted. The objective of these two analyses is to find any between-group and within-group difference in responding to fake news July 2019 ½Vol. 23 ½ No. 1 content. Meanwhile, to test the second hypothesis, Pearson's correlational analysis was carried out. Table 1 shows the statistical result of independent sample t-test, comparing the incumbent side and the opposition side in terms of the type of fake news. Meanwhile, Table 2 depicts the tendency that each side had in responding to fake news, regarding the side they supported and the opposing side. Based on the independent sample t-test (Table 1), incumbent supporters were found, more so than the opposition, to show trust in fake news discrediting the opposition side (t (69) = 3.32; p < 0.01, d = 0.78). Also, opposition supporters, as compared to the incumbent group, were more likely to believe fake news that dishonors the incumbent (t (69) = −2.31; p < 0.05, d = 0.54) (see Figure 1). However, based on the result of paired sample t-tests (Table 2), the incumbent supporters were found to be less likely to demonstrate meaningful response change in their trust toward fake news both concerning the opposition or the incumbent (t (33) = 0.29, p = 0.77, d = 0.044). Unlike their counterparts, the opposition supporters showed a difference in responding to fake news, in which they tended to demonstrate trust in fake news regarding their opponent (i.e., the incumbent), compared with fake news about their own side (t (36) = −5.62, p < 0.01, d = 0.929) (see Figure 1). Similar patterns are also found in partisans' tendency to spread fake news content. Based on an independent sample t-test (Table 1), the result, prima facie, shows that the tendency to spread anti-opposition fake news was higher in incumbent supporters than the opposition counterpart (t (69) = 2.35; p < 0.05, d = 0.56) (see Figure 2). Proclivity to spread anti-incumbent fake news was similarly demonstrated more by the opposition supporters, as compared to the incumbent group (t (69) = −2.10; p < 0.05, d = 0.50). Looking into more detail with paired sample t-tests (Table 2), the incumbent supporter group tended to show no significant response difference pertaining to the tendency to spread all types of fake news (t (33) = 0.12, p = 0.90, d = 0.022). In contrast, the opposition supporters showed a stark difference in responding to different types of fake news, where they were more inclined to spread fake news that discredited the rival (i.e., the incumbent), compared to the fake news of their own side (opposition) (t (36) = −4.68, p < 0.01, d = 0.774) (see Figure 2).

Figure 2. Group Comparison to Spread Each Fake News Story
Based on the data in Table 2 above, the incumbent side tended to respond similarly to all fake news (to both those items regarding the opposition and the incumbent), as compared to the opposition group. The opposition showed a contrasting response toward fake news, based on their presidential preference (see Figure 3). They tended to trust fake news concerning their political opponent more than they did fake news which discredited the candidate they supported. It implies that H1a and H1b are partially accepted because both have only been confirmed in the opposition group, and not on the incumbent group.
To test Hypothesis 2, a Pearson's correlational analysis was conducted (see Table 3). The result shows no significant correlation between critical thinking and trust in fake news among the incumbent supporters, either in terms of anti-opposition fake news (r anti opposition = −0.21, p > 0.05) or of anti-incumbent ones (r anti-incumbent = −0.089, p > 0.05). A similar pattern was also demonstrated by the opposition group, among whom critical thinking was not found to correlate significantly with trust in fake news (r anti opposition = 0.199, p > 0.05; r anti-incumbent = −0.246, p > 0.05). Similarly, the case with tendency to spread fake news, was not markedly correlated with critical thinking, both in the incumbent supporters (r anti opposition = −0.220, p > 0.05; r anti opposition = −0.070, p > 0.05) and the opposition counterpart (r anti opposition = 0.283, p > 0.05; r anti-incumbent = −0.278, p > 0.05). These show that H2 is not confirmed.
Consistency in tracing information was found between the incumbent and opposition groups (see Table 4). In terms of anti-opposition fake news, both groups of supporters did not trace the sources of information (t (69) = 0.58; p < 0.05, d = 1.41). Neither the opposition group nor the incumbent counterpart traced the sources of information pertaining to anti-incumbent fake news (t (69) = 0.57; p < 0.05, d = 1.39).

Discussion
Results show that hypotheses regarding a selective exposure tendency on each side are partially confirmed. Difference in trust (H1a) and tendency to spread (H1b) each type of fake news content is accounted for more by stark differences among the opposition support, compared to the incumbent supporters. The opposition group had a markedly higher tendency to trust and spread news content which discredits their political opponents, than to do so for the fake news content that dishonors the side they support. On the other hand, this was not observed in the incumbent group, where pro-incumbent partisans tended to be neutral and demonstrated no difference in responding to fake news of both their favored candidate and the rival. In contrast with the political symmetry as reported by Ditto et al. (2019), this research instead supports the political asymmetry perspective, whereby the opposition supporters are more likely to demonstrate selective exposure by trusting fake news about their political enemy compared to the incumbent supporter group.
Trust in fake news that discredits the incumbent rather than the opposition might also be investigated by considering the characteristics of partisans. Pro-opposition participants dominated by alliances with conservative ideology (Mietzner, 2015) are suspected to have influenced their tendency to trust fake news content. Research by Guess, Nyhan, and Reifler (2018) showed that individuals who tended to be conservative and pro-President Trump in the United States were more likely to visit fake news websites. Faris, Roberts, Etling, Bourassa, Zuckerman, and Benkler (2018) also found that disinformation news driven by political clickbait had become a prominent issue and had been exaggerated more in conservative media than in progressive media. Bronstein, Pennycook, Bear, Rand, and Cannon (2018) reported that religious fundamentalists and dogmatic followers are more likely to trust fake news and are less sensitive to the need to distinguish fake news from real news (i.e., media truth discernment). Dogmatic and ideologically extreme (as opposed to moderate) people tend to share simplistic, certainty-oriented cognitive styles (Sterling, Jost, & Pennycook, 2016) and are less open-minded (Bronstein et al., 2018). A study by Yilmaz and Saribay (2017) indicates that conservatism (whether social, economic, or general) supporters tend to show low trait reflectiveness.
Research by Amanullah and Dwisusilo (2018) can also be used to explain political asymmetry in fake news content. Considering the number of followers and content in the Facebook group of Jokowi (incumbent) and Prabowo (opposition) supporters, opposition supporters are more prone to the echo-chamber effect than their counterparts. This indicates that opposition supporters are more vulnerable to the post-truth phenomenon compared to incumbent supporters (Amanullah & Dwisusilo, 2018). The most frequent content found in the opposition-supporting Facebook group (i.e., Prabowo for NKRI) are negative comments, mockery, and hate speech directed toward the incumbent (Amanullah & Dwisusilo, 2018). Such content reached 91 items (33%). It is different from the incumbentsupporting Facebook group (i.e., Loyalis Jokowi-Ahok group), where the majority of content produced by the supporters is propaganda/campaign/defense on Jokowi/PDI-P/supporters/their families, programs, and policies, reaching 105 counts of total content (41%). Negative comments, mockery, hate speech, and negative campaigning toward Prabowo and his supporters take up only 21% of the total posts. This research found no notable correlation between general critical thinking ability and the tendency to trust and to spread fake news content in both groups. It is also supported by the finding that indicates a relatively similar level of critical thinking ability among the two sides, but they have different proclivities for selective exposure to fake news content. It seems to conflict with the intuitive approach which says that critical thinking is closely related to debiasing. Kenyon and Beaulac (2014) said that critical thinking ability is inadequately reliable to ensure debiasing when an individual makes a decision or judgment. Further, according to the argumentative theory, reasoning ability often causes individuals to be more biased if they have already had a particular standpoint beforehand (Mercier & Sperber, 2011). Instead, individuals are capable of bolstering their opinion using the reasoning ability they have, by which they will seek out justification of their opinion, rather than minimize their bias (Mercier & Sperber, 2011).
One important thing to keep in mind is that the instrument constructed by Fajrianthi et al. (2016), which was used in this research, is designed to measure the degree to which an individual has the analytical ability to draw conclusions, recognize assumptions, think deductively, interpret information, and evaluate the strength and relevance of arguments. The critical point of debiasing is not merely how individuals can recall their ability to think analytically in drawing a conclusion, but also on their perspective-taking ability. An experiment by Todd et al. (2011; demonstrates that perspective-taking can change implicit bias by increasing psychological interconnectedness between the self and target perspective-taking. Therefore, it takes not only the ability to think analytically, but also the ability to detach from one's own perspective and see things from a different perspective. This was termed as critical open-mindedness by Lambie (2014). Future researchers are expected to investigate open-minded thinking ability in terms of its role in reducing bias toward media content. July 2019 ½Vol. 23 ½ No. 1 In relation to the second research question, it was found that regardless of their political allegiances, almost all respondents did not browse the provided links of the fake news. This could be due to a number of possibilities: 1) respondents have already had an intuitive and/or a rational tendency pertaining to which content they feel they can trust and which they cannot, without browsing the source of the news; 2) Paper-based survey might not make it easy for respondents to trace the news source, far less to do fact-checking, and using their own smartphone to do so might be seen as time-consuming. Research by Guess, Nyhan, and Reifler (2018) actually shows that only 14% of internet user partisans in the United States do fact-checking. However, the findings of this experiment are not strong enough to be interpreted more generally due to the limitation in the research design. Future research should consider the ease for respondents of tracing news sources and factchecking.

Conclusion
This research shows that the opposition group is more inclined to conduct selective exposure than the incumbent group. Selective exposure is manifested in the opposition side's tendency to trust news that discredits their political enemy and demonstrates the contrary response to news which defames the candidate they support. Also salient is the opposition group's higher tendency to share fake news concerning their political opponent than to those items regarding their favored candidate.
There are some limitations and potential criticisms of this study. The first pertains to the sample, both the small sample size and the sample recruitment through the snowballing method. Both objections make comparisons between the two groups displaying a lack of generalizability if a national-scale conclusion is to be made. However, participant recruitment from one population unit is regarded as adequate in describing the inferential relation between critical thinking and selective exposure tendency.
The second potential criticism is related to the usage of the critical thinking test. The critical thinking measuring instrument constructed by an Indonesian researcher is initially expected to minimize cultural bias, so it can truly reveal the critical-analytical thinking ability of Indonesian partisans. However, in this research, the test created by Fajrianthi et al. (2016) is found to have low internal consistency. Further improvement on this test's validity is expected to confirm its predictive power.
The third objection pertains to the experiment design to observe cognitive bias tendency when one traces the source of information. The required manipulation is deemed not practical enough to actually reveal the target behavior. The initial assumption of this experiment is to find out how pronounced the bias tendency is, whether an individual is inclined to do fact-checking and browsing the source of information in order to reduce cognitive dissonance, especially when they get negative news concerning the candidate they support. Future research should really take into account the employment of practical design, so that it can reveal the bias tendency in fact-checking, as well as investigate intuitive and rational tendency when one does it.

Declaration of Interest
The author has never been a partisan, a member, and a consultant of any political party and organization. The research was funded by the faculty of Psychology, Airlangga University in 2017. The authors declare that there is no conflict of interest.