國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5. Discussion and conclusion
Considering that social media is a widespread source of news worldwide, and political polarization has been on the rise, this study built on hostile media effect (HME or hostile media perception, HMP) literature to understand how issue involvement affects bias perception in a social media context. The research also sought to understand how partisanship affects media trust, news credibility, how journalistic roles affect bias perceptions, and what social media content segments contribute to them. This investigation is significant as it has a unique focus on social media news content, and a holistic understanding of various media perceptions amongst partisans, of journalistic roles, and of the meaning of the journalistic text in a social media context.
5.1 Hostile media effect under social media
Perhaps the most important finding of this study is that the HME occurs under a strictly social media context. Participants were shown a Facebook news post about a divisive issue and then asked to assess how biased it was in favor and against the issue in question. Drawing from HME literature, we expected that partisan participants (those with a strong opinion for or against the issue in question) will perceive a piece of relatively objective reporting as biased.
We expected that this effect would occur no matter which side of the issue the partisan is on.
(H1a, H1b) Both these expectations were proven supported. In this study, nationality was not a significant predictor of partisanship on the given issue – this underlines the validity of researching media perceptions using issue involvement as the measurement of partisanship, since this value-based type of involvement is a better predictor of media perceptions than outcome-relevant involvement would be: in this case, it did not make a difference whether the participant was a citizen of any of the two countries involved in the U.S.-Iran military conflict.
Existing literature on the HME also says that pro-partisans will perceive a truly biased piece of reporting as more biased in favor of the anti-partisans, than anti-partisans will. This phenomenon should also occur on both sides of the issue, meaning that anti-partisans will perceive a pro-partisan piece of content as more biased than those who are pro-partisans.
‧
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧
anti-partisan news story as significantly more biased in favor of the anti-partisan side than did anti-partisans. H2a showed no significant results, as in, anti-attack participants did not show significant signs of the relative hostile media effect. This result might indicate that there was an overlap between partisan groups in this study and partisanship in the traditional sense of political leaning. It may have been the case that pro-partisans in this study, meaning those supporting the military actions, fell on the conservative or even Republican political spectrum, while those who were anti-partisans, and opposed the military actions, fell on the liberal, Democrat spectrum. Literature shows that there is a relationship with political partisanship and bias perception, Watts et al. (1999) found that the public is more likely to perceive liberal bias in the media, not because of the actual quality of the content, but because of the narrative of a liberally biased media became prominent during political campaigns in the 1990s. A decade later Feldman (2011) running an issue involvement based study on the relative hostile media effect, also found that selective perception was more significant amongst the partisan group in her study that aligned with conservative political values.
Arguably more research would need to be done to fully understand why H2a was not supported. However, it is still an important finding that the hostile media effect occurs even in a strictly social media environment, effectively meaning that partisan participants did not have to read the actual article behind a Facebook post to perceive it as biased against their own point of view.
The relevance of this finding comes from the nature of SNS networks to cater to personalized needs, providing self-, peer-, and algorithm-curated flows of information, and resulting in extremely selective exposure. (Thorson & Wells, 2016) This, combined with the general vicious nature of discussion that is prevalent in social media environments (Perloff, 2018) may lead to extreme opinion polarization. HME may play an important role in this, as partisans may infer public opinion from both slanted and objective social media news content that they perceive as biased, (Gunther, 1998) and as we have discussed, this may lead to outcomes such as corrective action, that is not always within the realm of democratic thinking. (Barnidge & Rojas, 2014) As there were no interaction effects observed regarding partisanship and research conditions, the significant main effects of partisanship on the hostile media perception are important.
‧
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5.2 Facebook’s credibility amongst partisans
The findings suggest that partisanship is negatively correlated to news credibility, as asked in RQ1. In this study social media, such as Facebook, Twitter, Youtube, or Reddit was the second most popular main source for news (29.8%). Perceived credibility of Facebook as a news source was significantly lower amongst partisans than moderates.
The significance of the findings on Facebook’s perceived news credibility is twofold – first, it is a direct investigation of this specific SNS platform’s credibility as a news source, and second, it suggests that partisanship is negatively correlated with Facebook’s credibility as a news source.
Facebook is still the most widely adapted SNS and a significant source for news, therefore its credibility is worthy of attention. Source credibility of SNS platforms is not widely researched, and results are hard to compare between platforms due to their layout and functional differences. Therefore research about the credibility of news tweets on Twitter (Schmierbach & Oeldorf-Hirsch, 2012) is not necessarily indicative of Facebook’s credibility of a news source. Some studies have shown that social media use worsens heterogeneity of discussion networks (Kim et al., 2013), while others (e.g. Bakshy et al., 2015; Kim, 2011) argue that Facebook, or in general, SNS use exposes users to diverse views, and yet other research says that this exposure might lead to further polarization (Bolano, 2018) – but in any of these cases it is an important finding that those with stronger opinions in general contribute lower credibility to Facebook as a source than moderates do. Furthermore, as social media was the most preferred news source for the youngest demographic in this study, the fact that Facebook’s perceived credibility as a news source is low might continue to grow in importance as time passes. As there were no interaction effects observed regarding partisanship and research condition, the significant main effects of partisanship on the perceived credibility of Facebook as a news source are important.
5.3 Social news content segments and perceived bias
Overall, it is found that proprietor content is the main source of users’ bias perception (78.7%
and 77.8% under the two conditions). This is significant for news practitioners, as at the
‧
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧
moment of posting news content on Facebook, proprietor content, unlike aggregate user representation numbers or user generated content, is directly under the control of the news organization. Walther and Jang’s (2012) segmentation of social content defines three types of content segment in social media: proprietor content, aggregate user representation, and user generated content. In this study we did not look into UGC, but amongst proprietor content and aggregate user representation numbers, participants said proprietor content was a major cause of their bias perception.
An interesting finding is that, while the number of checked segments under the two conditions are almost identical, their distribution is different. In the case of the advocacy journalist role, almost all participants identified the journalist’s text as a source of their perceived bias, but less votes were casted on different segments of the post; while under the objective-normative condition, the most often chosen segment was the description of the article, which was identical under both conditions. Under the advocacy journalist role condition, 34.8% of participants said that the text written by the advocate journalist contributed to their bias perception, while only 15.2% said so under the objective condition.
This makes sense, as the advocacy condition was designed to express bias. Under the objective condition the biggest contributor to bias perception was the description of the article, which was adapted without change from an actual New York Times article, while under the advocacy role condition this segment only received 14.3% of the votes. A possible explanation of this difference would be that participants compared the segments to each other to weigh how much each segment contributed to their perception of bias. Although the number of checked segments is almost identical, we know that the perceived bias was not equal under the two conditions, in fact, the advocacy condition caused significantly more bias perception.
This finding reveals the difficulty of directly assessing bias perception in news audiences – when asked to find the source of bias, participants might look for bias even where they previously perceived less of it. Further research may focus on whether participants truly process the social news content by comparing segments to each other – if so, that might be an important implication for news practitioners and for the news production process. In news production, these content segments (the title, the description, the thumbnail image, and the social media text) are in most cases not produced by the same member of the
‧
between each other, then news content production must be more mindful of this phenomenon in order to achieve the desired media perception.
5.4 Journalistic roles
As predicted in H4, bias perception under the advocacy role was significantly higher than under the objective-normative role. The novelty of this finding is that the perception of journalist roles performed in a social media context concurs with research about perception of journalistic roles under mass media logic. However, this results does not resolve the notion that in practice objective, dry content might contradict the logic of SNS platforms, where more engagement often leads to more exposure. (Oeldorf-Hirsch, 2018) On the other hand, social media news content creation is a complicated ecosystem, in which not all journalists are adapting SNS platforms (determining factors being age, type of work, and professional attitude, see Hedman & Djerf-Pierre, 2013), and not all news outlets have guidelines for journalists’ social media participation, which makes news workers uncertain as of how to behave professionally in an SNS environment.
As objective-normative journalism was designed exactly with the purpose of raising news credibility to better accommodate advertising and generate more revenue, (Charles, 2019) it might be an important implication of this finding that this type of social media content might drive less bias perception and in turn possibly raise platform credibility. As social media trust is generally low, (Edelman, 2019) and lower perceived credibility of Facebook was observed in this current research, SNS platforms must be aware that these negative user perceptions might result in less attention paid to these platforms (Gaziano, 1988), and in consequence, gaining less revenue through social media advertising. As SNS platforms do not control their published content directly, particularly in the case of news and journalism, Facebook has made various efforts to fact-check, screen for misinformation, and promote quality journalism and trusted sources. (Facebook, 2019)
However, these findings underpin the value of objective-normative journalism in a social media context, as in, this journalistic role resulted in significantly lower perceived bias by both partisan and moderate participants of this study. That being said, this research on its own is not sufficient to perfectly model the experience of coming across a news article on
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
journalistic text in a traditional media context includes perception of the presenter, source of the news (Tsfati & Cohen, 2012), while in this study we aimed to design a generic, anonymous fictional journalist. In a real-life scenario, it is arguably more likely that audiences will seek out or come across social media news content by journalists they are interested in, and might even agree with. Yet, it is important to note that, although this study did not focus on this particular bias, anti-partisans (those who are against the military strike) perceived anti-partisan slanted news content (M = 2.26, SD = .94) as significantly more biased against the attack, than how they perceived bias in objective content. ( M = 3.11,SD = .88) ( t(93.26) = 4.63, p < .001) This suggests that bias is recognized by partisans even when they receive content that shares their opinion.
Altogether, this study underlines that objective-normative journalism results in significantly lower perceived bias than advocacy journalism is, a potentially important consideration for news practitioners who produce news content for social media platforms. A practical implication for those who partake in social media news content production would be to develop codes of conduct for staff members regarding the tone and factuality of the social news content, while being mindful of the long-term professional goals and marketing targets of the given news outlet. Globally there exist a multitude of different journalistic mindsets – Hanitzsch and Vos (2018) specified 18 different roles that are valid under different cultural, contextual, political settings. In this framework of globally diverse concepts of journalism it might not be of utmost importance for each and every news outlet to have an objective-normative performance, or to convey unopinionated information to their audience.
It has to be taken into consideration that this study focused on an international news outlet’s international audience – on a culturally diverse news audience that cannot be characterized by one political or contextual setting. Implications therefore apply to news content producers who are primarily targeting this diverse demographic, which, in the age of the international news flow of social media, might be a growing chunk of the existing outlets.
For these outlets, desired levels of bias perception must be discussed on an editorial level and social media guidelines might help journalists to navigate this new media environment.
‧
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
5.5 Generalized media trust
The findings suggest a modern international news audience further proves what previous research argues about media trust, that is often dependent on variables such as partisanship and demographics. (Ladd, 2013; Lee, 2010; Edelman, 2019; Mourão, Thorson, Chen, &
Tham, 2018) In this study participants were asked to think about their primary news source and answer items measuring generalized media trust. Partisans had significantly lower levels of generalized media trust compared to moderates. Overall, journalistic assessment and selectivity of facts were the areas that respondents trusted their main news source less with, while trust in the selectivity of topics and accuracy of depictions was higher. Asking all participants about their primary news source, the majority said they get their news from news sites (44.8%) and social media (29.8%). Age was also an important factor, with the youngest age group preferring social media as a news source, the majority preferring news sites, and the oldest age group preferring traditional media. As there were no interaction effects observed regarding partisanship and research condition, the significant main effects of partisanship on generalized media trust are important.
5.6 New media environment
The relevance of all the above findings is underlined by two major factors built into in this study: a strictly social media context, and an international news media context.
By ensuring that participants received a piece of non-interactive (not clickable) social media news content, this research built an environment in which participants’ initial perceptions of social media news content can be investigated. This means that impression formation, namely, hostility perception was exclusively based on the amount of information presented in the stimulus material – relatively little compared to previous hostile media perception studies, where participants received longer pieces of traditional news media content, e.g. a full article or television report. The implications of the observed hostile media perception in this new media ecosystem might go above and beyond the implications of HMP under strictly mass media logic. In mass media logic the role of journalists, editorial boards, outlets is gatekeeping and filtering information, and therefore bias perception affects the
‧
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
‧
audience’s relationship with them. However, in the case of network media logic where the roles of content creator and content consumer are blurring into each other, it is not clear what relationship these bias perceptions end up playing.
This underlines the strength of this study, a unique combination of an international news audience as the sample and a purely social media context. This combination contributes to better understand the observed media perceptions in the new media ecosystem, where international news is globally available in real time, forming international clusters of issue-partisans and echo-chambers that exist beyond borders, while connecting international online news consumers based on their personal values. With the aforementioned recent trend of rising news engagement, whereas social media users are more likely to be actively sharing news within their network, (Edelman, 2019) but also considering the fact that these sharing processes occur under the self- and peer-curated, selective sharing environment of network media logic, especially prominent amongst those with high levels of involvement (Thorson &
Wells, 2016; Shin & Thorson, 2017), the implications of these findings are many.
One has to consider how the globally most widely adapted social media platform, Facebook, has an especially low credibility as a source for news. Most of all, partisanship in this international news flow under social media might play a more extensively impactful role than ever before. In this low credible setting research observes that hostile media perceptions occur in a similar way than in a traditional media context, but with the significant difference that the value-based issue involvement doesn’t only unify a localized audience, instead it creates international groups of issue-partisans and internationally involved clusters of opinion. This also implies that all outcomes of the observed media perceptions might also be less localized – therefore, for example, this internationally occuring hostile media effect might result in corrective action across borders. These media perceptions might play a similar key role in the international news flow as they did in localized mass media information flows.
On the positive side, these high-involvement international information flows might result in a more diverse news consumption than before globally speaking, where reporting on issues tied to certain values can attract a wider readership than merely the local news markets, therefore international news outlets and opinion leaders are incentivized to share them, thus engaging the international attention. On the negative side, this might result in changes in news consumption preferences, specifically, a turn towards non-mainstream sources. Audience
‧
‧
disengaged, on the contrary, engaged, but selectively exposed and selectively sharing. In behavioral outcomes, high-involvement international audiences may engage in corrective action and activism (Feldman et al., 2017) that arch over borders, and are discussed within these value-based online opinion clusters.
5.7 Limitations and future research
Important findings aside, this research has limitations that need to be addressed. As Hypothesis 2a was unsupported, further investigation would be necessary to determine whether anti-attack respondents would perceive a pro-attack slanted post as more biased towards the pro-attack side, than would pro-attack respondents. For this, it would possibly not be sufficient to present participants with either objective or slanted stimulus material, but the
Important findings aside, this research has limitations that need to be addressed. As Hypothesis 2a was unsupported, further investigation would be necessary to determine whether anti-attack respondents would perceive a pro-attack slanted post as more biased towards the pro-attack side, than would pro-attack respondents. For this, it would possibly not be sufficient to present participants with either objective or slanted stimulus material, but the