失信的社群媒體 : 探討臉書新聞內容的敵意媒體感知 - 政大學術集成

全文

(1)國立政治大學國際傳播英語碩士學位學程 International Master’s Program in International Communication Studies College of Communication National Chengchi University. 碩士論文 Master’s Thesis. 學學. ‧ 國 ‧ 國. 治政政治大大立立失信的社群媒體 : 探討臉書新聞內容的敵意媒體感知. s iitt yy. er. n i on Naatti o N. Facebook news content. ‧ ‧. In social media we distrust: Investigating hostile media perception of. aal l Student: Helga Gabriella 研究生：莫荷嘉 i v Csarnó nn CChh U U eennggcchhii. 指導教授：韓義興. Advisor: Professor Yi-Hsing Han. 中華民國 109 年 7 月 July 2020. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(2) 失信的社群媒體 : 探討臉書新聞內容的敵意媒體感知 In social media we distrust: Investigating hostile media perception of Facebook news content. 研究生：莫荷嘉 Student: Helga Gabriella Csarnó 指導教授：韓義興. Advisor: Professor Yi-Hsing Han. 國立政治大學. 治政國際傳播英語碩士學位學程政治立立. 碩士論文. 大大. ‧ ‧. ‧ 國 ‧ 國. 學學 s iitt yy. n i on Naatti o N. A Thesis Submitted to International Master’s Program in International. er. Communication Studies National Chengchi University In Partial Fulfillment of. aa. v. l l CCFor the degree of Master the Requirement nni of Arts. hheennggcchhii UU. 中華民國 109年 7月 July 2020. II. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(3) Acknowledgement I would like to express my deepest gratitude to my advisor, Dr. Paul Yi-Hsing Han. Without his insightful, patient guidance and encouragement to deepen my understanding of communication science this research would not have been possible. I extend my sincere appreciation to the committee members, Dr. Elaine Jhih-Syuan Lin and Dr. Sandra Shih-Hsien Hsu for their constructive feedback and continued support. I would also like to thank the faculty of NCCU and the IMICS department who inspired and encouraged me to pursue ambitious academic goals. I owe the opportunity to proceed with my graduate education to the Ministry of. 治政治大honors, not having lived in the 政 to support me with the Taiwan Scholarship. Without these 大立立 country of Taiwan, my life would be radically different and less colorful today.. Education of Taiwan for awarding me the Huayu Scholarship five years ago, and continuing. ‧ 國 ‧ 國. 學學. Finally, I would like to thank my family in Gödöllő, Budapest, and Luxembourg. Thank you, friends, roommates, and classmates for keeping me company these two years – I. ‧ ‧. hope to see you again, safe and content. Thanks to the Geese of Beverly Road for helping me. s iitt yy. n i on Naatti o N. out with survey responses.. I’m especially grateful to Damar Hoogland and Dési Ádám Dusi for their wise. er. insights on my research, and to Bobály Sára, whose friendship was my home away from. aal l v i n n CChh eennggcchhii UU. home. I offer this thesis to Csarnó László Tamás, who loved to see me write.. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(4) Abstract The majority of internet users today find their news on social media (Gil de Zúñiga et al., 2017), however, media trust, and especially trust in social media is low. (Edelman, 2019) In growing political polarization the effects of perceived media hostility are also gaining more importance. In this research readers of international news were surveyed in an online experiment to assess how issue involvement on the 2020 military conflict between the United States and Iran correlates with general trust in the media and with the credibility of the largest social media network, Facebook, as a news source. The current research investigated whether the hostile media effect still occurs in a purely social media context and results showed that. 治政治大 lines as they do in a traditional 政 perceive news content on Facebook as hostile along the same 大立立 media context. This study also made an attempt to understand how content creation is related. partisans (those with a strong supporting or opposing opinion on the military conflict). ‧ 國 ‧ 國. 學學. to bias perception, namely, what segments of social news content and which mainstream journalistic role contribute to perceived bias. This study fills the literature gap of the hostile. ‧ ‧. media effect in a social media context. Findings may also have implications for the news. s iitt yy. er. n i on Naatti o N. industry as to how journalist roles and content segments influence audience perceptions.. Keywords: social media, media trust, media credibility, hostile media effect, journalist roles, United States, Iran. aal l v i n n CChh eennggcchhii UU. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(5) Table of Contents 1. Introduction 1.1 Research background 1.2 Research purpose 1.3 Contributions. 1 1 3 4. 2. Literature Review 2.1 The hostile media effect 2.2 News media trust 2.3 News credibility 2.4 Journalistic roles 2.5 Social media news content. 6 6 12 15 17 19. 3. Method 3.1 Research framework 3.1.1 Hypotheses and research questions 3.2 Survey design 3.3 Sample and data collection 3.4 Stimulus material 3.5 Measures and reliability 3.5.1 Partisanship 3.5.2 Hostile Media Perception 3.5.3 Generalized trust in the news media 3.5.4 News credibility 3.5.6 Social media news content segments. 20 20 20 21 22 23 25 25 27 28 29 30. 學學 er. n i on Naatti o N. s iitt yy. ‧ ‧. ‧ 國 ‧ 國. 治政政治大大立立. aal l v i n n CChh eennggcchhii UU. 4. Data analysis and results 4.1 Survey participants 4.2 Media consumption habits 4.4 Generalized media trust 4.5 News credibility 4.6 Partisanship 4.7 Testing the Hostile Media Effect 4.8 Testing general media trust in partisans 4.9 Testing the impact of journalistic roles 4.10 Partisanship and Facebook’s news credibility 4.11 Social news content segments and perceived bias 4.12 Comparative analysis. 31 31 32 36 37 38 38 40 40 40 41 43. 5. Discussion and conclusion. 46. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(6) 5.1 Hostile media effect under social media 5.2 Facebook’s credibility amongst partisans 5.3 Social news content segments and perceived bias 5.4 Journalistic roles 5.5 Generalized media trust 5.6 New media environment 5.7 Limitations and future research. 46 48 48 50 52 52 54. 6. References. 56. 7. Appendix 7.1 Stimulus material 7.2 Questionnaire. 73 73 7 4. 治政政治大大立立 ‧ ‧. ‧ 國 ‧ 國. 學學 er. s iitt yy. n i on Naatti o N. aal l v i n n CChh eennggcchhii UU. VI. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(7) List of Figures and Tables Table 1. Sample’s partisanship Figure 1. Frequency of issue involvement measures Table 2. Descriptive statistics Table 3. Main news source Table 4. Main news source per age groups. Crosstabulation Table 5. News consumption habits Table 6. Sample’s social media use habits Table 7. Sample’s social media platform use Table 8. Objective-normative condition. Segments contributing to bias perception Table 9. Objective-normative condition. Segments contributing to bias perception Table 10. Advocacy condition. Segments contributing to bias perception Table 11. Advocacy condition, Content segments contributing to bias perception. 治政政治大大立立. 26 27 31 32 33 34 35 36 41 42 42 43. er. s iitt yy. ‧ ‧. ‧ 國 ‧ 國. 學學. n i on Naatti o N. aal l v i n n CChh eennggcchhii UU. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(8) 1. Introduction 1.1 Research background From the mid-2000s to the early 2010s, journalistic practices started to shift to fit the new attention economy (Nixon, 2017), resulting in the rise of perceived misleading headlines, sensationalism, and the birth of clickbait and listicle journalism. Online discussions were increasingly based around polarized clusters of opinion, forming echo-chambers (Thorson & Wells, 2016) and resulting in a heavily divided environment of public discourse that is prevalent up to day. A major part (42%) of the American public has assessed national news. 治政政治大 biased and left-leaning news 立 media, while those holding 大 a populist worldview tend to have 立 low trust in mainstream media and a preference for commercial and tabloid media. (Fawzi,. media as too liberal, (Gallup, 2019), and populistic politics also profited from the rhetoric of. ‧ 國 ‧ 國. 學學. 2019) As societies came to realize the political and social meddling power of misinformation. ‧ ‧. in the mid- to late-2010s, higher than ever levels of media distrust set in.. Although we are past the all-time low of media trust levels of 2015–16, media remains. s iitt yy. n i on Naatti o N. the least trusted public institution in 2019, with trust in online news, and particularly in social. er. media at an extreme low across countries. (Edelman, 2019) According to recent white papers such as the 2019 Edelman Trust Barometer or the 2019 Reuters Institute Digital Report. aal l v i n n CChh i UUinformed public says that they trust eenng58% gcchofhithe the media, pointing to quasi stagnation. While (Newman et al., 2019), the past year marked either a modest rise or slight decrease of trust in. the news media (up 4 percentage points from 2018), only 47% of the general population agrees. In the meantime, despite widely available media literacy education and various other efforts from governments, social media platforms, and other organizations, 73% of the general population remains worried about weaponized misinformation. The late 2000s saw a rise of content offered for free that posed a challenge to traditional and online media, applauded by advocates for public and citizen journalism. As smartphone penetration increased back in the time when 4G technology first became widely accessible, social media platforms, like Facebook, and micro-blogging platforms, such as Twitter, grew to play a prominent role in people’s daily lives – the number of minutes the globally speaking average person spends on social media rose from 90 minutes in 2012 to. 1. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(9) 144 in 2019, a 60 percent increase over 7 years. (Clement, 2020) These platforms offered a space for computer-mediated communication and instant connectedness free of charge. Nightly network television news programs in the US had less than half the daily viewership in 2019 (27%) compared to 1995 (62%), while in 2019 49% of respondents answered that they use the internet every day to get their news. (Gallup, 2019) The lines between news production and consumption started to blur (e.g. Bruns, 2008, who talks about prosumption, when in the value chain, consumers become active in production as well) the concept of owned media (i.e., media owned and controlled by a given company or institution) changed drastically, relying less on branded websites and content and more on social media presence. Political communication started relying on and thus be formed by social media as well. (e.g. Klinger & Svensson, 2015) Skipping the middlemen – journalists and editors – meant that the. 治政政治大大 2015)立rose to its full power, the norm shifted to extremely 立. importance of gatekeepers has diminished, and as so-called social or network media logic (Klinger & Svensson,. ‧ 國 ‧ 國. 學學. individualized information feeds and a constant flow of algorithm- or self-curated information. Users relied less on traditional filters of information, such as media workers, and. ‧ ‧. more on the curation of their own social circles, such as friends, social contacts, and also algorithms. (Thorson & Wells, 2016). n i on Naatti o N. s iitt yy. However, as Monica Williams writes in the editor’s note of the Quill, (Williams,. er. 2018) the industry should not abandon hope – there are recent examples of forward leaps in positive perception of the news media, such as the case of South Korea, where news media’s. aal l v i n n CChh i UUGeun-hye. Although social media eennminister ggcchhiPark imprisonment of compromised former prime. rigorous coverage on government corruption fueled mass protests and resulted in the. logic is still absolutely prominent, with changing political and social environments such as the one exemplified in South Korea, significant changes in news consumption preferences are to be predicted. As Costera Meijer (2007) found, citizens’ expectations towards journalists change in times of need: young Dutch people reported enjoying infotainment media content even though were aware of the content’s low informational value, but still expected high-quality, “boring” hard news coverage to be readily available should events of great importance unfold. Recent surveys also describe changing patterns of news consumption: a “massive rise in news engagement” (Edelman, 2019, p. 17), where the rate of people disengaged from the news has fallen, and the ratio of those who actively share news is higher than ever. With. 2. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(10) tendencies to distrust social media, this might be cause for worry, however, the Reuters Institute (Newman et al., 2019) has also reported that concern about the quality of information caused people to seek out ‘reputable’ news sources (40% in the US), while 24% says they stopped using sources with questionable reputation altogether. The notion and impact of sharing news might also be changing, as the Reuters Institute observed a rise in using messaging apps such as WhatsApp as a general tendency. They also describe a phenomenon where users are more likely to be in large group chats with people they do not personally know. Users are now more likely to be sharing news in these private environments than to post them publicly. Private and public Facebook groups are also prevalent, however, more so in certain countries than others. In such a rapidly changing media landscape and in the light of rising polarization of. 治政政治大 involvement influences different 立立media perceptions, such大as bias perception (the hostile. opinion (see e.g. Schaeffer, 2020), this current research aims to look into how issue. ‧ 國 ‧ 國. 學學. media effect) and media credibility. Since today’s news consumption is so closely intertwined with social media use (Gil de Zúñiga et al., 2017), the research investigates how. ‧ ‧. those with strong opinions on controversial issues perceive the credibility of a specific SNS platform (in this case, Facebook). An attempt is made to look into the relationship between. n i on Naatti o N. s iitt yy. specifics of news content creation (such as, the role the journalist takes on, and various. er. segments of the social media news content) and the perception of bias by the audience. Looking into these media effects, perceptions, and content characteristics might offer insight. aal l v i n n CChh eennggcchh–ii anUUimportant area of research in an especially in case of high-involvement individuals. into how social media contexts impact the audience’s relationship with news content,. extremely personalized, self- and peer-curated information environment that is prone to selective sharing, echo chambers and might enhance opinion polarization. (Thorson & Wells, 2016; Oeldorf-Hirsch, 2018; Shin & Thorson, 2017). 1.2 Research purpose This research aims to investigate how news content shared on the largest SNS, Facebook, is perceived by the audience, and how partisanship (defined in this study as a strong supporting or opposing opinion on the 2020 U.S.-Iran military conflict) affects this perception. It looks into participants’ perceived credibility of Facebook as a news source, as well as their general. 3. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(11) trust towards the media. Also, to better understand what, if anything, about social media news content induces bias, participants are asked about social media news content segments, and their bias perception is tested on two different journalistic roles. To do so, an online experiment is conducted to investigate whether there is a difference between bias perception in case participants see a news post written by a journalist who employs the objective-normative role, versus when they employ the advocacy role. The research assesses participants’ opinion on a divisive issue, the 2020 military conflict between the United States of America and Iran. In January 2020 President Donald J. Trump ordered the assassination of Iranian commander Qassim Suleimani, resulting in a surge in tension between the two countries. HME suggests that audiences with a strong opinion either supporting or opposing a controversial issue such as this military conflict. 治政政治大大 own point of view. According立 to this media effect theory this hostility perception will occur 立. (called partisans in this study) will perceive fairly objective reporting as biased against their. ‧ 國 ‧ 國. 學學. on both sides of the issue. Therefore in this study, based on their answers, participants will be categorized as either pro-partisans, anti-partisans, or moderates (those with a less strong or. ‧ ‧. neutral opinion). The broader problem this research focuses on and aims to solve is how high involvement and a purely social media context impact different news perceptions, with an. n i on Naatti o N. 1.3 Contributions. er. outlets.. s iitt yy. additional focus on aspects of social media news content that are under direct control of news. aal l v i n n CChh eennggcchhii UU. This proposed research would offer contribution to hostile media perception literature in a strictly social media context, specifically for news content on Facebook, a research gap in published research at the time of writing this thesis. Effectively, this research can address whether the hostile media effect occurs even before Facebook users click to read a full article. As a theoretical contribution, this investigation would answer exactly which segments of a social media post contribute to bias perception in an SNS setting. Findings may help scholars and media practitioners to better estimate and understand audience perceptions in this relatively new platform of communication. This may also contribute to a better understanding of what exactly the journalistic text entails in a social media setting – the journalistic text is an elusive concept that varies in its content from one communication. 4. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(12) channel to another. In a social media context the lines of creator and user are blurred, therefore the journalistic text itself may also be harder to define. Hopes are, this study may clarify this question to some extent. There are also only a few published research articles in medium credibility literature on Facebook as a source of news media, only on Twitter. This research contributes to literature not only by measuring credibility of the medium, but also segmenting it to four dimensions using a scale that has been applied widely to traditional and new media, thus making it possible to compare the platform’s credibility to other types of media channels. Findings of the study may offer clarity to media practitioners and journalism studies scholars as to whether advocacy journalism in a social media context induces bias. In the current atmosphere of social media and citizen journalism, editorial boards and journalists. 治政政治大 in-house social media guidelines, 立立something that currently大only a few online news media. might find practical use of this information, and may apply it in developing their own. ‧ 國 ‧ 國. 學學. outlets practice.. The literature review explores theoretical underpinnings of the hostile media effect,. ‧ ‧. and reviews published research on the possible explanations of the underlying mechanisms as well as potential consequences and outcomes of this media perception phenomenon. Then. n i on Naatti o N. s iitt yy. generalized media trust is reviewed, news credibility, and their relationships with high. er. involvement. Journalist roles under a social media context as well as the segmentation and characterization of social media news content is also to be found in this chapter. The method. aal l v i n n CChh ee theggcoperalization chhii UU and reliability of the research sample, the process of data collection andnn section describes the research design including hypotheses and questions, the survey, the. variables. The results section continues with descriptive and analytic exploration of the research variables and provides testing of the research hypotheses and questions. In the discussion chapter relationship of the findings is explored regarding the reviewed literature and then in the conclusion and limitations section the major findings of this study will be explained and explored, and potential future research paths laid out.. 5. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(13) 2. Literature Review To develop hypotheses and research questions, the current study looked into literature on the hostile media effect, news trust, news credibility, and journalistic roles. Within all topics the effect of a social media context was considered. Attitudes towards the media are one of the most important pillars of communication research because as their outcome political and social interactions and behavior might be impacted. Tsfati and Cohen (2012) argue that media trust, the hostile media effect, and media influence perception as media perceptions are related empirically and in their consequences, and even correlated as in they enhance each other. They argue that one of the key factors in. 治政政治大大立立involvement drives perception Therefore they theorize that issue of media hostility, the belief. the correlatedness of these media perceptions is ego-involvement – the tendency to be more. defensive against media coverage of issues that are relevant to one’s personal values.. ‧ 國 ‧ 國. 學學. that media is influential over moderates, and it lowers general media trust. In this current research two of these media perceptions are connected (media trust and the hostile media. ‧ ‧. effect) to issue involvement, and investigate the relationship between a social media platform’s credibility and issue involvement. An attempt to understand how these media. s iitt yy. n i on Naatti o N. perceptions are affected by variables that news practitioners are more in control of, such as. er. journalistic roles and different segments of the social news content.. aal l v i n n 2.1 The hostile media effect CC h h eennggcchhii UU. In a groundbreaking research article, Vallone et al. (1985) coined the term for the hostile media phenomenon (HMP, or hostile media effect, HME). Using an unbiased piece of reporting that aired on television covering the 1983 Beirut massacre, they observed that both pro-Israeli and pro-Arab partisans not only perceived the objective reports as biased, but they also perceived them to be biased against their own stance, hence the use of the word hostile. They called this “biased perception of media bias” (Vallone et al., 1985, p. 578). Perloff, in a meta-analysis overarching 30 years of hostile media effect literature, defines the hostile media effect as “the tendency for individuals with a strong pre-existing attitude on an issue to perceive that ostensibly neutral, even-handed media coverage of the topic is biased against their side and in favor of their antagonists’ point of view.” (Perloff, 2018, p. 707). 6. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(14) This effect has been widely studied in communication science since then, but has also peaked the attention of public opinion and social science researchers. (Perloff, 2018) Since Vallone et al. used a controversial issue with high relevance in their experiment, many others followed by repeating the experiment using different political and social issues and defining partisans as those who have high levels of issue involvement on the issue in question. Other issues have also proven to cause this effect such as genetically modified food (Gunther & Schmitt, 2004; Gunther & Liebhart, 2006), the 1997 United Parcel Service strike (Christen et al., 2002), the abolishment of the National Security Law in South Korea (Choi et al., 2009), the use of primates in lab testing (Gunther et al., 2001) and sports coverage (Arpan & Raney, 2003) just to mention a few examples. Research methods have overwhelmingly been laboratory experiments or survey research according to Perloff. (2018) Hostile media effect,. 治政政治大大立立. leaning on Hansen and Kim’s (2001) meta-analysis, accounts for a medium effect size (r = .296).. ‧ 國 ‧ 國. 學學. As for the objective content that triggers the HME, first hostile media studies made use of content that was unbiased to test the effect. The vast majority of these studies used the. ‧ ‧. terms of neutral, evenhanded, or balanced content (Perloff, 2018), with less regard to the generally accepted notion that unbiased reporting is not necessarily possible, or doubts raised. n i on Naatti o N. s iitt yy. by post-modernism whether an objectively valid reality exists. In their exploratory study,. er. Vallone et al., 1985 defined their concept of unbiased reporting as reporting that in the eyes of an impartial reader would seem objective and unbiased. However, recognizing that in. aal l v i n n CChh ii UU eennggcch become a more common phenomenon, Gunther ethal. (2001) designed a study with less. practice objective reporting might not always be the case, since partisan leaning media has. objective material. They coined the term for relative hostile media effect, wherein “supporters and opponents of an issue perceive bias in a consistent direction (i.e., leaning toward one side), but each group perceives coverage as significantly more unfavorable to its own position relative to the other group.” (Feldman, 2017, p. 1) A key moderator for HMP is involvement, however, the definition of involvement (just as the definition of the hostile media effect) is not coherent all throughout published research, also, there is no consensus in the necessity of issue-involvement or ego-involvement for HME to occur, as some studies, such as Arpan and Raney (2003) explicated this aspect as a prior attitude. Hansen and Kim (2011) however concluded that involvement is a moderating variable in the occurrence of HME. Different types of involvement, such as value-relevant or. 7. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(15) affective involvement were investigated by Choi et al. (2009) and Matthes (2013). Choi and colleagues built on meta-analysis by Johnson and Eagly (1990), and differentiated three types of involvement: value-relevant (other times called as ego-serving), outcome-relevant, and impression-relevant involvement, and out of these three, focused on value-relevant and outcome-relevant involvement. Their results showed that value-relevant involvement (involvement that is rooted in one’s important personal values) was a critically significant predictor of HMP, while outcome-relevant involvement (involvement in an issue that is relevant to one’s current goals and outcomes) was not significant as predictor. Other than involvement, they also found media skepticism to be the strongest single predictor in their model. Matthes (2013) set out to investigate whether the hostile media phenomenon is more affected by cognitive or emotional (affective) involvement, and found that the latter was. 治政政治大大 audience was found significant no matter the reason of the type of involvement. (Tsfati & 立立. overwhelmingly supported by survey data. Furthermore, the level of involvement of the. ‧ 國 ‧ 國. 學學. Cohen, 2012) Hansen and Kim (2011) found that the extent of the hostile media effect is positively correlated to the level of issue-involvement: the more extreme the perceiver’s. ‧ ‧. views were, the more they perceived content to be hostile, or at least not supportive enough towards the issue in question – finding however significant hostile media perception under. n i on Naatti o N. s iitt yy. low involvement as well.. er. Another important moderator of the hostile media phenomenon is perceived reach or presumed media influence. Perceived media reach is relevant to the partisan audience’s. aal l v i n n CChh ii UU eeSchmitt nggcchh (Gunther et al., 2009) In fact, Gunther and n (2004) have shown that the hostile media. assessment of media hostility as in they perceive more impactful content as more biased.. effect is gone or even reversed when the perceived reach is low: they used the case of a student essay as minimum reach, and contrasted it with high perceived reach, in case of a newspaper article. Participants did not experience the hostile media effect to a significant degree under the low perceived reach condition even though the content was identical. This impact of perceived reach has been repeatedly reproduced in research. However, Perloff (2018) in his meta-analysis notes that perceived reach is complicated to manipulate, for example, in Gunther and Schmitt’s 2004 study, the student essay was said to have been written by a college student, while the newspaper article was said to have been produced by a journalist. Therefore, it is not clear if the lack of hostile media perception in the low reach condition was truly due to perception of lower reach, or was it because of different. 8. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(16) perceptions of the source. The fear that a hostile and biased piece of content will influence others is understood to happen due to the third-person effect, (Davison, 1983) a cognitive bias that leads readers to think that others are more likely to be influenced by biased coverage than they themselves are. The in-group effect also plays a role in the third-person bias, as in the bigger the social distance is between the perceiver and the audience they imagine reading the article, the stronger the third-person effect is. (Tsfati & Cohen, 2012) This leads us to the third key moderating factor of the HME, social identity. Social identity theory (Tajfel & Turner, 1979) describes that social groups and belonging are important aspects of the self, and involves stereotyping out-groups and striving to see our group, the in-group in a positive light. Through the third-person effect, this social theory is applicable to the hostile media phenomenon, whereby ego-involvement can render a story on. 治政政治大 (Hartmann & Tanis, 2013) Looking where the controversial reporting 立立 specifically into cases大. a controversial issue highly relevant to social identity, and therefore cause the HME to occur.. ‧ 國 ‧ 國. 學學. covered low-status groups or groups with which participants strongly identified, and therefore. were more motivated to preserve a positive image of it, Hartmann an Tanis (2013) describe. ‧ ‧. higher sensitivity to how they perceived the media. In Tsfati (2017) perceived media bias by Israeli Arab citizens led to perceived media stigmatization, and a relationship with increased. n i on Naatti o N. s iitt yy. minority alienation was established. Similar effects were observed in strong political group. er. identification (Duck et al., 1998) where social in-group thinking was the theoretical framework of hostile perception of a political campaign; or religious groups. (Ariyanto et al.,. aal l v i n n CChh ii UU between the audience and the ee angsocial gcchhdistance perceptions of the source come to affect, n. 2007) The message source may also trigger the third-person effect: as in-group or out-group. source is established, which affects hostility perception. This may cause for bias perception to be lower when the source is understood to be an in-group member than in case of an out-group member source (Bolsen et al., 2014, Reid, 2012) The underlying information processing mechanisms of the HME have been of scholarly attention from the very beginning. Vallone et al. (1985) speculated selective perception and selective evaluation may both explain the audience’s bias, but since their data supported both processes, the question remained unexplained. Schmitt et al. (2004) proposed three information processing mechanisms that might explain the HME: selective recall (partisans disproportionately better remembering information they disagree with), selective categorization (the tendency for partisans to categorize more of the news content as. 9. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(17) unfavorable than moderates would), and different standards (thinking that counterarguments to our viewpoint are invalid or irrelevant). Running an experiment with the divisive issue of genetically modified food in the focus, they concluded that “selective categorization appears to be the one viable explanation for the hostile media effect,” (Schmitt et al., 2004, p. 636) while selective recall has not shown to have a significant relationship with the HME. This means that partisans have a tendency to perceive and categorize more of the media content as unfavorable than moderates would. However, there is no support that selective memory would mediate hostile media perceptions. In a separate study, Giner-Sorolla and Chaiken (1994) hypothesized that pre-existing negative attitude towards the media in general have a mediating effect on HME, calling it a cause for the hostile media judgments. Outcomes and consequences of the hostile media effect for partisans include corrective. 治政政治大 influence of the perceived biased 立立 media coverage makes大partisans voice their opinions in. action (Barnidge & Rojas, 2014) – this describes the phenomenon where presumed strong. ‧ 國 ‧ 國. 學學. public spaces as if to correct the wrong coverage’s effects. In other cases, related to the aforementioned effects of social identity and intergroup perceptions, minorities may. ‧ ‧. self-perceive as having less political power than others, resulting in lower trust in the media and in the institution of democracy. (Tsfati & Cohen, 2005) Hostile media effect is also. n i on Naatti o N. s iitt yy. theorized to be related to public opinion inference, a concept coined in Gunther (1993).. er. Persuasive press inference is a phenomenon whereby people gather their perception of public opinion from what is being communicated by mass media. This happens through a. aal l v i n n CChh ee toggbeccimpactful, hhii UU will influence others’ opinions. media content – and that media, perceivednn. generalization that the news content that they perceived as biased is similar to the rest of. This holds as a way of public opinion inference even though idea of media power has been disproven, showing that biased media coverage does not affect moderates’ opinion, merely reinforces the opinion of partisans. (Tsfati & Cohen, 2012) Partisans in particular are described to hold the belief that biased media content may affect others’ opinions, however, findings on this are contradictory. Christensen et al. (2002) observed that while moderates do in part infer public opinion from their subjective understanding of the news content, partisans however, when triggered for hostile media effect, are also likely to project their own opinion on the public. In this scenario, partisans did not perceive the entire public opinion as hostile, only the specific piece of news content. (Christen et al., 2002; Perloff, 2018) notes that. 10. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(18) comparing the strength of projection and of persuasive press inference on perceptions of of public opinion, the former is a more important factor. Although the majority of HME literature was produced under mass-media logic and with traditional media content (Hansen & Kim, 2011), with the growing weight of social media and its implications on information abundance (Klinger & Svensson, 2015) and on attitude towards news consumption (Gil de Zúñiga et al., 2017) makes it an important and relevant area for hostile media perception research. Published research on HME under social media environments show these platforms of communication, that are furthering selective exposure as compared to traditional mass media (Thorson & Wells, 2016), are perhaps deepening political and ideological polarization. (Perloff, 2018) For one, partisans tend to selectively share news even when it’s factual: Shin and Thorson (2017) found that partisans. 治政政治大 sharing of news with a heightening influences discussion and opinion 立立sense of media hostility大. selectively share fact-checking messages that support their own candidate, and selective. ‧ 國 ‧ 國. 學學. polarization. Perception of media hostility is also worsened by politicians’ narratives on unfair media coverage and fake news – even following politicians on social media may lead. ‧ ‧. to a heightened sense of media hostility. (Weeks et al, 2019; Perloff, 2018) also notes that the tone of discussion on social media tends to be more vicious than previously observed in. n i on Naatti o N. er. to look into.. s iitt yy. traditional media settings, and argues that relative hostile media effect might be a key aspect. Since most key factors – involvement, perceived reach, social identity, the third-person. aal l v i n n CChh eennggcch hii UU namely, interpersonal and mass and in fact the two main sources of public opinion inference,. effect – are not specifically more limited under social media than in other media channels;. media messages (Glynn & Noelle-Neumann, 1986) come together on SNS; in this research it is hypothesized that the hostile media effect is likely to occur in a similar way and have similar consequences under a social media context like in a traditional media one. While the total reach of a particular social media post is only available to page managers and not to the public eye, SNS users may deduce potential reach by looking at aggregate user representations, e.g. the number of likes and reactions, the number of comments, and the number of shares on a post. Therefore it is hypothesized that the hostile media effect occurs in a purely social media context as well in the same way it does in online and traditional news media.. 11. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(19) Hypothesis 1a: Pro-attack participants reading a balanced news story on the 2020 U.S.-Iran military conflict will perceive the content to be more hostile toward their personal position than moderates, those who are neither pro-attack or anti-attack. Hypothesis 1b: Anti-attack participants reading a balanced news story on the 2020 U.S.-Iran military conflict will perceive the content to be more hostile toward their personal position than moderates, those who are neither pro-attack or anti-attack. Research on the effect of partisanship on hostility perception showed that audiences are significantly more critical towards information that counters their beliefs, and less so against confirming information. (e.g. Gunther & Schmitt, 2004). 治政政治大大 position than pro-attack a slanted news story to be more 立立biased in favor of the pro-attack participants.. Hypothesis 2a: Anti-attack participants will perceive slanted information presented as. ‧ 國 ‧ 國. 學學. Hypothesis 2b: Pro-attack participants will perceive slanted information presented as a. participants.. ‧ ‧. slanted news story to be more biased in favor of the anti-attack position than anti-attack. s iitt yy. n i on Naatti o N. er. 2.2 News media trust. aal l v i n n CChh U ennggcchh–iiin U media – is at an all time low. Trust – ore skepticism the institution of media has been in Trust in the media – defined as a generalized presence or lack of trust in the institution of. the center of public attention for the second half of the 2010s, and has been a widely researched topic in the fields of communication, political science, cognitive science and psychology. Trust in online news media ranks lower than in traditional media, although traditional media’s perceived trustworthiness has also dropped significantly over the last five years. (Ipsos, 2019) Currently the least trusted branch of media is social media, with only 44% of the general population globally finding it trustworthy. (Edelman, 2019) If one subscribes to the idea that surveillance of power by an independent body of journalists is essential to democracy, the current lack of trust is cause for worry – studies have shown that low media trust in consumers is an indicator of their preference for non-mainstream news sources (Fletcher & Park, 2017), which in turn tends to further. 12. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(20) decrease trust in the media. (Tsfati & Cohen, 2012). Low media trust effects are both systemic and individual – as one needs to spend time and effort coping with mistrusted information, media trust can be interpreted as a form of social capital. Generally speaking, news media evaluations are rooted in one’s expectation of and satisfaction with the political, social, and entertainment functions of media. (Fawzi, 2019) Van der Wurff and Schoenbach’s study (2014) shows that what the public and what experts and media workers consider important public functions of the press generally collide. However, the public and journalists disagree in how much significance should be assigned to certain journalistic practices. (Tsfati et al., 2006) How individuals evaluate news media also depends on the audience’s own naive media theories, developed based on everyday interactions. In terms of the relationship between political alignment and media trust and. 治政政治大大trust and satisfaction. All three those with a populistic worldview 立立tend to have lower media. satisfaction, Fawzi (2019) conducted a representative study in Germany and confirmed that. ‧ 國 ‧ 國. 學學. dimensions of populism used in this study, namely, anti-elite populism, anti-outgroup. populism and people centrism, contradicted normative expectations towards the media,. ‧ ‧. higher trust levels were only observed for tabloid-like and commercial media. Roots of trust in the media have been conceptualized in a multitude of ways in both. n i on Naatti o N. s iitt yy. psychology and communication research. Kohring and Matthes (2007) break it down to. er. willingness to trust on the side of the audience, and trustworthiness on the side of the message communicated. They also found four dimensions of trust in the news: the selection. aal l v i n n CChh UU eenngg(2012), cchhii trust Tsfati and Cohen in the institution of media is. of topics, the selection of facts, the accuracy of descriptions, and the assessment of the journalist. Drawing from. dependent on two major clusters of factors: audience characteristics, such as psychographics, political predispositions, and involvement; and the journalistic text (piece of content) itself, meaning the impressions on the presenter or journalist, specific characteristics of the given medium, and the quality of journalism. Recent digital media reports paint a picture of societies divided amongst class, gender and political lines when it comes to trust in the media. Low media trust is prevalent amongst those with a strong political opinion: Lee (2010) found that it is in fact political ideology and partisanship that are determining factors for believing that news media is politically biased and therefore not to be trusted. However, not all partisans are created equal, in the United States specifically there is a 36% gap between the media trust of Republican voters versus. 13. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(21) Democrats, in fact, Mourão et al. (2018) found that support for Donald Trump was the single strongest predictor of news distrust in their study. Ladd and Podkul (2019) argue that intentionally eroding mistrust in the media has now become a tool of electoral strategy. Generally speaking, trust inequality (trust levels analyzed on the lines of social class and education) is at an all time high between the informed public and the general public. (Edelman, 2019) Based on the above literature, we hypothesize that when reading about an issue that is divisive amongst political lines, readers with stronger political predispositions will report lower levels of generalized media trust. Hypothesis 3: Participants who have a stronger opinion on the 2020 US-Iran conflict will show lower levels of media trust when they see news content on Facebook, than. 治政政治大大立立. participants who do not have a stronger opinion.. ‧ ‧. ‧ 國 ‧ 國. 學學 er. s iitt yy. n i on Naatti o N. aal l v i n n CChh eennggcchhii UU. 14. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(22) 2.3 News credibility Media credibility is an important indicator of successful communication – previous studies have shown that if medium credibility is low, people are less likely to pay attention to it (Gaziano, 1988), while Wanta and Hu (1994) argue that the success of mass media communication altogether (may the goal be to inform, to educate, to influence, or to pursue the audience) depends on credibility. The concept of media credibility is a complex, multidimensional one that scholars began to study in the 1950s. There are two major branches of credibility research – one focuses on source credibility (e.g. Hovland & Weiss, 1951), analyzing the features of the. 治政治大政 focuses on how different channels of communication, e.g.大 newspapers, television, radio, etc. 立立 influence perceived credibility. A third approach is to investigate what audience. communicator, while the other branch, medium credibility (e.g. Gaziano & McGrath, 1986). ‧ 國 ‧ 國. 學學. characteristics influence the variable of perceived credibility, with findings on most major demographic factors: age, gender, race, income, and education. The first line of source. ‧ ‧. credibility research (Hovland and Weiss, 1951) identified the source credibility model, with. s iitt yy. n i on Naatti o N. two key dimensions (expertise and trustworthiness), while later Whitehead (1968) found competency and objectivity to be more essential, and Berlo et al. (1969) identified three. er. dimensions: safety, qualifications and dynamism – generally speaking, there is no general. aal l v i n n CC h Perceived credibility of news inh ae social i UU is important because according ennggmedia cchhicontext. consensus (Golan, 2010) as to what dimensions of source credibility should be differentiated.. to a 2018 Pew Research Center survey, almost two thirds (68%) of Americans used social media to get their news at least occasionally. (Shearer & Matsa, 2018) It is a unique medium in terms of audience perception – usually when a particular medium becomes widely popular, its perceived credibility will go up, (Flanagin & Metzger, 2000) but, in the case of news on social media, many are skeptical about the information they are getting. (57% of respondents in Shearer & Matsa, 2018) This suggests a general low perceived credibility of news in a social media context, and, because of its widespread use, it might affect a large part of the population. For online news media, research initially measured the perceived credibility of online news to be between that of newspapers and television (a rating of 2.06 on a 0-4 Likert scale as written by Kiousis, 2001).. 15. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(23) It is unclear how the unique traits and segments of social media platforms influence source credibility. Because of the diversity of SNS platforms, it is not possible to talk about general medium credibility of social media, thus Schmierbach and Oeldorf-Hirsch (2012) focused on comparing news credibility in the context of a specific SNS platform with the credibility of online news in itself: in the case of Twitter, they found that the perceived credibility of a New York Times story shared as a tweet was lower than that of the original article itself. Still about Twitter Westerman et al. (2014) found that more recent tweets were positively correlated with all factors of cognitive elaboration, and elaboration was in turn positively correlated with source credibility. Involvement still seems to play an important role in perceived credibility. Arpan and Raney (2003) did an experimental study on sports fans and found that the source of the sports. 治政政治大 found the same article to be significantly towards the rival town’s team if it 立立 more supportive大. news affected perception of credibility. Participants who supported their hometown team. ‧ 國 ‧ 國. 學學. was printed in a neutral town’s or the rival town’s newspaper, in contrast to their hometown’s. newspaper. As for the effect of involvement in an online news setting, Choi et al. (2006). ‧ ‧. asked partisan participants to evaluate the credibility of news coverage on the Iraq War, and found that anti-partisans (those who opposed the war) perceived the Internet as a news source. n i on Naatti o N. s iitt yy. to be less aligned with a pro-government position than traditional media, and simultaneously. er. found the Internet as a news source more credible than did neutral respondents. Therefore there seems to be an enhancing effect of involvement of issue involvement on news. aal l v i n n CChh eennggcchhii UU issues and communication platforms.. credibility, but it is unclear in what specific way this effect unfolds in the case of different. Considering this information this current research proposes a research question to investigate the relationship between partisanship and perceived credibility of Facebook as a source for news information. Research Question 1: What is the relationship between participants’ strength of opinion about the 2020 U.S.-Iran military conflict and their perceived credibility of Facebook as a news source?. 16. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(24) 2.4 Journalistic roles Journalistic values and practices, which Deuze (2005) sums up as an occupational ideology, reflect the fact that journalists see their work as a profession, rather than a job, and find their service to the public important – in fact, a significant factor in their occupational satisfaction derives from carrying out the public service role of journalism. (Hedman & Djerf-Pierre, 2013) There are definitely more than one journalistic roles that media workers take on simultaneously, such as the interpretive, the disseminator, the adversarial, or the mobilizer roles, with the first two being the most prevalent, (Van der Wurff & Schoenbach, 2014) and the latter two frowned upon under the objective-neutral media model of the United States.. 治政治大政 These roles and values hold up under social media大 as well, journalists, pressured to 立立 keep up with Twitter and citizen journalists use social media to perform their journalistic (Tsfati et al., 2006). ‧ 國 ‧ 國. 學學. roles. (Hedman & Djerf-Pierre, 2013; Lasorsa et al., 2012). However, in social media environments, lines of professional and personal life have become blurry for media workers.. ‧ ‧. Journalists use social media to varying degrees, for either, both, or neither private and. s iitt yy. n i on Naatti o N. professional purposes – important factors in their choices are their age, their outlook on their profession, and skepticism towards the platform. Many of them reported concerns about their. er. objectivity versus opinionatedness in social media spaces. (Lasorsa et al., 2012). aal l v i n n CChh UU the tendency for journalists to continue values and practices to etoenapply hhii occupational nggcctheir Deuze (2005) talks about the process of normalization (originally of the blogosphere):. new and emerging platforms. Larsorsa et al. (2012) argue that normalization continues well into social media. This platform might have first seemed like a “new tool for carrying out the traditional tasks” (Hedman & Djerf-Pierre, 2013, p. 9), but social media also seems to be greatly influencing the way journalists perform their roles, and the importance they give to different journalistic values and practices. The new environment is characterized by increased transparency of journalistic processes, and Singer (2005) concludes that social media is merely a new platform for journalists to display their professional identity and values. However, Hedman and Djerf-Pierre (2013) argue that this performative environment leads to a need for news media to continuously justify itself in front of the public. The perception of the role of a journalist in the community has also been influenced, namely, it has become. 17. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(25) more important in the new media news environment, as news consumers now have a closer, more interactive relationship with journalists. (Nah & Chung, 2012) Regarding new journalistic practices that emerged under online news production and later under social media logic, Karlsson (2010) talks about new features to make the journalistic practice transparent and thus increase credibility, such as: time stamps, highlighted corrections, external links, chat platforms, links to blogs, a comment section, and email addresses to contact members of the editorial team. These professional tools create a direct line of communication, which Hardee (2016) describes as a step in the natural evolution of journalism towards its goal to strengthen and deepen connectivity to audiences. Although a multitude of alternative models of journalism exist globally (e.g. Hanitzsch & Vos, 2018), the most prevalent ones in American journalism are the objective-neutral. 治政政治大大 may take one side of an issue, (Schudson, 2001) and the advocacy 立立 model, in which journalists. model in which journalists refrain to factual information and aim to balance their sources,. ‧ 國 ‧ 國. 學學. often advocating for social change. (Janowitz, 1975) Strictly speaking, advocacy is the difference in communication between public relations messages and journalism. However,. ‧ ‧. many argue that the reality of the two models is less black and white (see e.g. Charles, 2019; Fisher, 2016). Fisher (2016) writes that today advocacy journalism exists on a continuum. n i on Naatti o N. er. well as personal factors on behalf of the journalist.. s iitt yy. from subtle to overt, determined by macro, organizational, production and source factors, as. Although historically partisanship of newspapers was once accepted in the United. aal l v i n n CChh UU ee credibility. ggcchhii (Charles, better accommodate advertising and to raisenn 2019) Some still argue that States also, since the commercialization of legacy papers objectivity became the norm to. advocacy journalism is more transparent of the journalist’s bias, and therefore it is more trustworthy and credible than strictly objective journalism, traditionally objectivism is considered to be more credible, as it has been specifically designed for this desired effect. In this research the objective-normative landscape of journalism is accepted, and it is hypothesized that advocacy journalism under a social media environment will result in higher bias perception than the objective-normative role. Hypothesis 4: Advocacy journalism will result in higher perception of bias in a news post on Facebook than objective-neutral journalism.. 18. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(26) 2.5 Social media news content Social media platforms have inherently changed the flows of information, resulting in extremely selective exposure, self- and peer-curated information feeds, contributing according to many researchers to the formation of echo-chambers and the polarization of today’s political and ideological climate. (Thorson & Wells, 2016; Klinger & Svensson, 2015; Perloff, 2018) However, studies on the effects of social media news content on polarization is contradictory. Lin et al. (2011) in an effort to quantify and compare media bias in social media and in traditional media have found that in the global distribution of coverage of U.S. congress members, social media, due to network factors, is more prone to forming clusters of opinion with “skewed and uneven distribution of popularity.” (Lin et al., 2011, p.. 治政政治大大 of discussion networks (Kim 立 et立 al., 2013), and amplifies certain biases of partisan politics, 200) Other studies have shown similar results in that social media use worsens heterogeneity. ‧ 國 ‧ 國. 學學. such as gatekeeping bias, statement bias, and coverage bias. (Saez-Trumper et al., 2013). Meanwhile others (e.g. Bakshy et al., 2015; Kim, 2011) argue that Facebook, or in general,. ‧ ‧. SNS use exposes users to more diverse views than they would otherwise receive. Yet other research says that this exposure might lead to further polarization. (Bolano, 2018) As this. n i on Naatti o N. s iitt yy. study seeks to understand what parts of the social media content would contribute to. er. perceived hostility (RQ2), social media news content is operationalized along the lines of. aal l v i n n CChh content, aggregate categories of social content: proprietor user representation, and user eennggcchhii UU generated content. Proprietor content is understood to be the “main message”, such as the. Walther and Jang’s (2012) segmentation of social content. In this reading there are three. news article itself, including text or visual cues created by a principal author or organization. Aggregate user representations are twofold: deliberate representations, such as the number of likes on a Facebook post, and incidental representations, such as the view count under an Instagram video. Finally, user generated content is the reactions, e.g. comments of nonproprietary users – this research would omit this third segment of content, as they are not under the control of the news organization, and therefore less relevant to the purpose of the research question. Research Question 2: What segments of the news content shared as a social media post on Facebook contribute to participants’ bias perception?. 19. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(27) 3. Method 3.1 Research framework This research uses quantitative methodology and employs two techniques: an online survey and within that, a two conditions between-groups online experiment. Hypotheses were raised based on relevant literature and resources, and hypothesis validity was tested with statistical analysis. Research questions were raised where published research was contradictory or where there was a research gap, particularly where research in a social media context was missing.. 治政治大research questions: (how) does 政 issue involvement and media perceptions, and answers two 大立立 partisanship (issue involvement on the controversial issue of the 2020 U.S.-Iran military Survey data analysis tests four hypotheses regarding the relationship between high. ‧ 國 ‧ 國. 學學. conflict) affect participants’ perception of Facebook’s credibility as a news source; and which segments of social media news content contribute to perceived bias. The experiment. s iitt yy. n i on Naatti o N. or the advocate role) result in higher perceived hostility.. ‧ ‧. investigates which of the two most prevalent journalistic roles (the objective-normative role. er. 3.1.1 Hypotheses and research questions. aal l v i n n CChh eennggcchhii UU Hypothesis 1a: Pro-attack participants reading a balanced news story on the 2020. Hypotheses and research questions are as follows:. U.S.-Iran military conflict on Facebook will perceive the content to be more hostile toward their personal position than moderates, those who are neither pro-attack or anti-attack. Hypothesis 1b: Anti-attack participants reading a balanced news story on the 2020 U.S.-Iran military conflict on Facebook will perceive the content to be more hostile toward their personal position than moderates, those who are neither pro-attack or anti-attack. Hypothesis 2a: Anti-attack participants will perceive pro-attack information presented as a slanted news story to be more biased in favor of the pro-attack position than pro-attack participants.. 20. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(28) Hypothesis 2b: Pro-attack participants will perceive anti-attack information presented as a slanted news story to be more biased in favor of the anti-attack position than anti-attack participants. Hypothesis 3: Participants who have a stronger opinion on the 2020 U.S.-Iran military conflict will show lower levels of general media trust than participants who do not have a stronger opinion. Hypothesis 4: Advocacy journalism will result in higher perception of bias in a news post on Facebook than objective-neutral journalism. Research Question 1: What is the relationship between participants’ strength of. 治政治大政 as a news source? 大立立 Research Question 2: How will different segments of the news content shared as a opinion about the 2020 U.S.-Iran military conflict and their perceived credibility of Facebook. ‧ 國 ‧ 國. 學學. social media post on Facebook contribute to participants’ perception of bias?. ‧ ‧. 3.2 Survey design. s iitt yy. n i on Naatti o N. The online survey created on Qualtrics and shared on Facebook and Facebook Messenger. er. aims to better understand partisans’ media perceptions, namely how hostile, credible, and. aal l v i n n CChh find the complete survey in the appendix. eennggcchhii UU. trustworthy they saw media in general and the stimulus material they were presented. Please. The questionnaire titled Social Media News Content Perceptions consists of 7 parts: 1. Introduction, 2. Demographics, 3. Media consumption habits, 4. U.S.-Iran military conflict, 5. Generalized media trust, 6. News source credibility, 7. Hostile media perception, 8. and Debriefing.. 21. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(29) In 1. Introduction participants read an introduction of the research and then signed the consent form. In 2. Demographics they answered items on inquiring their age, gender, education level, nationality, and country of residence. 3. Media consumption habits questions were focusing on participants’ main news source, frequency and time spent with online media, international media, and social media. In 4. U.S.-Iran military conflict participants read a short description of the January 2020 events, they were asked about whether they are aware of the issue, whether they discussed the issue with others, how often they discuss international news stories, and they were asked about their opinion on the military conflict. In 5. Generalized media trust participants answered items on media coverage of the issue. In 6. News source credibility participants were asked about their impressions on their preferred news source and Facebook as a news source. In 7. Hostile media perception participants were. 治政政治大 where a fictional journalist shared 立立an objective news article大about the U.S. military conflict.. randomly shown one of two versions of the stimulus material – a fictional Facebook post,. ‧ 國 ‧ 國. 學學. The journalist’s commentary was either objective-normative or advocate. Having been shown the stimulus material, participants had to evaluate whether they found the post biased and to. ‧ ‧. what extent, and also, what segments of the news content led them to feel that way. In 8. Debriefing participants were informed about the research design and the manipulation of the. s iitt yy. n i on Naatti o N. stimulus material.. er. 3.3 Sample and data collection. aal l v i n n CC h Data was collected from internationalhreaders ofcc English eenngg hhii UUspeaking international news media through an online survey. The participants were adults (over the age of 18). The reason for selecting an international audience was that involvement on the issue selected for measuring partisanship (the U.S.-Iran military conflict of January 2020) is one of international interest. (Poushter, 2018) Participants were partially recruited through snowball sampling, where each subject is asked to recruit further subjects, (Babbie, 2016) partially through purposive sampling. Snowball sampling started by posting the survey link to the researcher’s personal Facebook account, and sent around in personal messages through Facebook Messenger. Purposive sampling meant sharing the survey on thematically specific public and private Facebook groups to ensure that a representative number of partisan respondents filled out the survey.. 22. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(30) Recruitment through Facebook made sure that survey participants were Facebook users, and therefore their insights on Facebook news content was potentially better informed. The survey questionnaire was designed and published on Qualtrics. Statistical analysis was conducted with SPSS v24.. 3.4 Stimulus material For this study, two images were created as stimulus material. The two variants represented the objective-normative journalist role and the advocacy journalist role to answer Research Question 2 about which of these roles will result in higher bias perception. These were also the content pieces that the hostile media effect was tested on, with the objective-normative. 治政治大政 Stimulus material was shown to survey participants at random 大assignment. 立立 The images were designed as screenshots of a Facebook post – a fictional news post. condition representing a balanced piece of content, and the advocacy role a slanted one.. ‧ 國 ‧ 國. 學學. by Reuters news agency titled “The U.S. Orders Strike Killing Top Iranian General Qassim Suleimani” with description “Suleimani was planning attacks on Americans across the. ‧ ‧. region, leading to an airstrike in Baghdad, the Pentagon statement said” and added text. s iitt yy. n i on Naatti o N. “Breaking News: The U.S. confirmed it was behind the strike that killed the powerful Iranian commander Qassim Suleimani, a major escalation in a conflict with Iran.” The new post also. er. came with a thumbnail image, and came with 21K likes, 7.1K comments, and 21K shares.. aal l v i n n CChh via a fictional journalist sharing this original eennpost. ggcchhii UU. This basic news post was unchanged under both conditions, as conditions were manipulated. The post was designed as if shared by a fictional journalist named John Journalist. To manipulate for the journalistic role condition, two variants were created: one where John Journalist, taking on the objective-normative role, shared his views in a factual and objective manner, and one where he took on the advocacy role, sharing both facts and opinions from an anti-attack viewpoint. The original news article post was selected after considering multiple controversial issues that would serve for testing the hostile media effect. After discussion, selecting on the 2020 U.S.-Iran military conflict served as a news story that had the potential to involve international audiences. For the selection of the original news post multiple major international news outlets’ first (breaking news) social media posts about this issue were. 23. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(31) considered both on Facebook and on Twitter. After consideration, one of the breaking news posts on the issue by the New York Times was selected. The original Facebook post linked to a New York Times article (Crowley et al., 2020), this interactivity was not provided during the experiment. Considering potential source effects, the New York Times was changed as the outlet is understood to have slight partisan bias. (Weinberg, 2010) Instead, Reuters, a news wire service was chosen in an effort to ensure minimum source effects. The original post was titled “Trump Orders Strike Killing Top Iranian General Qassim Suleimani” which could also have affected bias perception, so “Trump” was taken out wherever possible and the U.S. as a nation was presented as the actor in the news story to avoid unnecessary triggering. A Facebook post (versus a tweet from Twitter) was selected as internationally speaking the. 治政政治大 that the level of how knowledgeable 立立 participants are about大Facebook posts might influence platform is the most widely adapted. (Clement, 2020a) Choosing this platform also means. ‧ 國 ‧ 國. 學學. their perception, as may the fact that news posts on Facebook look uniform and don’t carry. individual news outlets’ visual or content layout traits. (Kalogeropoulos & Newman, 2017). ‧ ‧. The numbers of likes, comments, and shared (user aggregate representation numbers) were adapted from the original New York Times Facebook post. As the New York Times is a well. n i on Naatti o N. s iitt yy. established news outlet with a significant global audience, the numeric representation of the. er. reach of their breaking news post is potentially fit to meet one of the important factors in hostile media perception: high reach. HME literature suggests that partisans’ bias perception. aal l v i n n CChh UU eennggcontent cchhiiunder reach (in Gunther and Schmitt, 2004, the same the student essay condition was. disappears or even gets reversed if they don’t perceive the content to have a high enough. not perceived as hostile as under the newspaper article condition). However, how participants perceive the number of likes, comments and shares, whether they see these numbers as expressions of high reach, can not be guaranteed. To manipulate the journalist role condition, a fictional journalist, John Journalist was created, using a minimalistic vector graphic as profile picture featuring a white man. The journalist was designed as a man because according to Tsfati and Cohen (2012) gender and ethnicity plays a role in the perception of the presenter of the news, and therefore in the perception of the journalistic content. Using a vector graphic instead of a photo and using the pseudonym John Journalist was in an effort to avoid similarity with any real-life journalists or public figures, but it might have potentially impacted the perception of the post. For the. 24. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(32) manipulation itself two additional texts were created. Under the objective-normative condition, John Journalist shared the news post and wrote, “Iran’s top security and intelligence commander was killed early Friday in a drone strike at Baghdad International Airport that was authorized by President Trump” (word count: 24) These sentences were directly copied from the original New York Times online article (Crowley et al., 2020), and were deemed fit for the objective-normative condition as it featured sober and factual language expected in such journalist norms. (Schudson, 2001) As for the advocacy journalism condition, the text “America shocked and angry over the attack authorized by President Trump against Iran’s top security and intelligence commander” (word count: 18) was written, featuring strong, emotive language as well as factual information in an effort to model this school of journalism that relies on both factual reporting and on expressing. 治政政治大 Manipulation was carried 立立out by taking a screenshot大of the original New York Times. personal opinion or judgement of a situation. (Cáceres, 2019). ‧ 國 ‧ 國. 學學. Facebook post, and carefully re-designing it to fit all above considerations, while also keeping aesthetically consistent with Facebook’s layout as of February 2020. For this task we. ‧ ‧. used Photoshop CS6.. s iitt yy. n i on Naatti o N. 3.5 Measures and reliability. er. The survey included questions to operationalize partisanship, general media trust, news. aal l v i n n CChh eennggcchhii UU. credibility, and hostile media perception, and the segmentation of social media news content.. 3.5.1 Partisanship. Following hostile media perception literature (e.g. Vallone et al., 1985; Tsfati, 2007; Gunther & Schmitt, 2004), this study treats partisanship as involvement on a divisive issue. The study sample is international, thus asking directly about party loyalty or political alignment, concepts which may vary in meaning between cultures, might lead to inconsistent data. Measuring issue involvement instead makes sure that participants’ strength of opinion and their perception of the stimulus material are not irrelevant to each other. The issue in question needs to be divisive and of international relevance. The issue of the 2020 military conflict between the United States and Iran was selected. This event at the time of data collection was the last installment of a long history of conflict. 25. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(33) between the two countries, one that had been drawing international attention for decades and captured worldwide attention in January 2020. To assess partisanship (M = 3.81, SD = 1.86) a screening item adapted from Gunther and Schmitt (2004) asked participants to give their opinion to the 2020 US-Iranian military conflict on a 7-point Likert scale (1 – I strongly oppose the U.S. military actions. 4 – I am strictly neutral on this issue. 7 – I strongly support the U.S. military actions). Those whose opinion takes the value of 1, strongly oppose, 2, oppose, 6, support, 7, strongly support, are to be considered partisans. Those whose opinion falls between 3 and 5 were considered moderates. Table 1. Sample’s partisanship (N= 339). 59. I somewhat oppose the U.S. military actions.. 39. I am neutral on this issue.. 95. ‧ 國 ‧ 國. I oppose the U.S. military actions.. n i on Naatti o N. I somewhat support the U.S. military actions.. 36. I support the U.S. military actions.. 23. ‧ ‧. 42. (%). 學學. I strongly oppose the U.S. military actions.. Percentage. s iitt yy. Opinion. Frequency 治政治政 (n)大大立立. er. Variables. 12.4 17.4 11.5 28.0 10.6 6.8. aal l v 45 13.3 i n n CChh UU i i e e h h n n c g c g Running a frequency analysis shows that 170 respondents (50.1%) were moderates with. I strongly support the U.S. military actions.. 28% of respondents saying they are 4, neutral on the issue. There were 101 partisan respondents (29.8%) who opposed the military strike, and 68 (20.1%) who supported it.. 26. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.

(34) 治政政治大大立立 ‧ 國 ‧ 國. 學學. In this research two different variables were created to conduct analysis on partisanship: one differentiated 2 groups, moderates and partisans, while another. ‧ ‧. differentiated 3 groups, moderates, anti-partisans, and pro-partisans.. s iitt yy. n i on Naatti o N. 3.5.2 Hostile Media Perception. er. To measure their hostile media perception, participants are shown stimulus material and are. aal l v i n n CCranging U biased against, through 3 – heenngg were given on a 5-point Likert scale h from 1ii– U cchh strongly asked to answer two items (α = .813) adapted from Gunther and Schmitt (2004). Answers. strictly neutral, to 5 – strongly biased in favor.. 1. Would you say that the portrayal of the strike in this post is strictly neutral, or is it biased in favor of one side or the other? (M= 2.56, SD= 1.17) 2. Would you say that the journalist responsible for this post was strictly neutral, or were they biased in favor of one side or the other? (M= 2.30, SD= 1.09) These two items were combined into one variable, no reverse coding was necessary. The resulting variable (hostile media perception, M = 2.43, SD = 1.04) was used in testing perceived bias.. 27. DOI:10.6814/NCCU202000898 DOI:10.6814/NCCU202000898.