Toxic Talk: How Online Incivility Can Undermine Perceptions of Media

Scholars are increasingly concerned with the potential for uncivil discourse to enhance political polarization in society (Mutz, 2006; Stryker, 2011). Political elites and partisans boost levels of incivility in news media (Muddiman, 2013), and people perceive higher levels of incivility in politics when individuals rather than issues are attacked (Stryker, Conway, & Danielson, 2014). The concerns over incivility extend to the online information environment, where nasty comments can harm healthy back-and-forth dialogue central to democracy that often happens in spaces such as newspapers (Coe, Kenski, & Rains, 2014; Meltzer, 2015). Early research shows that incivility in online comments can be a polarizing factor in how people perceive issues in media, particularly for individuals who hold stronger opinions before seeing the comments (Anderson, Brossard, Scheufele, Xenos, & Ladwig, 2014). In particular, incivility has been found to affect people’s perceptions of the content covered in news articles, a phenomenon dubbed the “nasty effect” (Anderson et al., 2014). It can also increase the perception that individuals in society hold polarized attitudes (Hwang, Kim, & Huh, 2014). Yet, we still do not fully understand how incivility in comments affects perceptions of the news stories themselves or the effect it can have on lines of political polarity in society. Here, we explore the effects of uncivil comments proximate to a blog post containing hard news content on individuals’ perceptions of media. We examine these effects across two controversial scientific issues—nuclear energy and nanotechnology—on perceptions of bias of the news blog post. We also examine how political ideologies are related to how people interpret incivility in online comments as it relates to perceptions of bias. In doing so, this study contributes to understanding of the polarizing nature of incivility. While this is a common concern in the literature, few have empirically tested whether online incivility heightens existing divisions in society. We find that perceptions of bias are greatest after seeing online incivility among politically conservative ideologues, providing evidence for the polarity of incivility in online comments. Predicting Perception of Bias in the Blog Post (N = 2,224) *p ≤ .05; **p ≤ .01; ***p ≤ .001. Predicting Perception of Bias in the Blog Post (N = 2,224) *p ≤ .05; **p ≤ .01; ***p ≤ .001. Certain aspects of online communication may make it particularly susceptible to the proliferation of uncivil discussion. The absence of nonverbal communication cues and physical isolation on the Internet may encourage more uncivil conversation online compared with a face-to-face environment (Dutton, 1996; Hill & Hughes, 1999; Papacharissi, 2002). Such behavior has been defined as “hostile and aggressive interactions via text-based computer-mediated communication” (O'Sullivan & Flanagin, 2003, p. 69). Incivility is prevalent in online commenting. Analysis of one major daily newspaper found incivility in >20% of its comments, with name-calling being the most frequent form of incivility (Coe et al., 2014). Another analysis of mainstream media and blog sites separated out individual types of uncivil behavior, and found >25% of comments contained either insulting language or mockery, the most common types of incivility in their study (Lee & Pang, 2014). Research on incivility in various online discussions has demonstrated several effects, including erosion of open-mindedness about issues (Borah, 2013), expectations that public discussion can have multiple points of view (Hwang et al., 2014), and satisfaction with online discussions (Gervais, 2015). It can also enhance political engagement (Borah, 2013; Brooks & Geer, 2007) and perceptions of credibility in a blog (Borah, 2014; Thorson, Vraga, & Ekdale, 2010). Previous research provides evidence that incivility may encourage biased perceptions. Uncivil messages are seen as unfair (Brooks & Geer, 2007), yet a balanced news story is perceived to be more credible if it is embedded within an uncivil blog post than a civil one (Thorson et al., 2010). People who read uncivil online discussions are more likely to perceive the discussants as dominant and less credible compared with those who read civil discussions (Ng & Detenber, 2005). Finally, a content analysis of uncivil comments around news media stories has shown that uncivil messages are not based on evidence (Coe et al., 2014). Building on what Anderson and colleagues (2014) call a “nasty effect,” where perceptions of issues covered in news were negatively affected by exposure to uncivil comments, we expect to find that individuals exposed to incivility are likely to have more negative perceptions of the news article that precedes the comments. We pose the following: H1: Exposure to uncivil comments following a news blog post is positively related to perceptions that the blog post is biased. Scholarship on the effects of incivility in online discussions has begun to explore potentially polarizing effects of online discussions. For instance, Anderson and colleagues (2014) showed that incivility in online comments can encourage negative perceptions about the issue featured in the news content among individuals who are more religious or who have less support for the issue before seeing the uncivil comments. This indicates those with moderate views are less likely to form negative attitudes when exposed to uncivil discussions. Moreover, exposure to uncivil comments leads audiences to perceive that there is polarization of attitudes among the general public (Hwang et al., 2014), and uncivil language in a blog post produces more attitude certainty and less open-mindedness (Borah, 2014), outcomes that further indicate incivility could be pushing people toward more polarization. So far, research on incivility in online discussions has not examined how a core component of polarization—political ideology—influences interpretation of media stories connected to uncivil comments. Yet, political ideology is a core value that plays into how people make decisions about information seeking (Stroud, 2008; Yeo, Xenos, Brossard, & Scheufele, 2015) and consumption (Iyengar & Hahn, 2009), as well as interpretation of that information (Ho, Brossard, & Scheufele, 2008; Yeo et al., 2014). Here, we examine this phenomenon in the context of perceptions of media bias. There is some literature that suggests that we will find a moderating effect of incivility on perceptions of bias, given that conservative media are more likely to contain “outrage” language than liberal media are (Sobieraj & Berry, 2011). In addition, evidence has shown that people with stronger views are more likely to respond to incivility negatively (Anderson et al., 2014). Furthermore, people are more likely to perceive polarity among attitudes following exposure to online uncivil discussions even if their attitudes are not more extreme (Hwang et al., 2014). However, given the relative paucity of empirical evidence that addresses this topic, we ask the following: RQ1: Is the effect of incivility in comments following a news blog post on perceptions of bias of the blog post moderated by political ideology? This study was conducted using an online experiment. Participants were given pretest measures designed to tap concepts such as media use habits, scientific knowledge, and trust in media. Then, they read one of two neutral news blog posts on either nanotechnology or nuclear energy from the Vancouver Sun newspaper (see Supplementary Material for a screenshot). This newspaper was chosen because participants were less likely to have preconceived ideas about the source, and the goal of this study was to test the effect of the comments and not the source. The blog posts were each three paragraphs related to water toxicity and contained equivalent information regarding the risks and benefits of the issues. Risks and benefits were mentioned throughout the comments an equivalent number of times. The comments were identical except when specific words had to be changed to reference the scientific issue. Within each issue, participants were given one of eight manipulations, each showing a different version of the comments section of the news blog post. Each post contained six comments in total, with two sentences in each comment. Participants viewed three comments about the risks of the technology and three about its benefits. In other words, the manipulations include an equivalent number of comments in favor of and against the technology. This study is part of a broader experiment examining other communicative elements in addition to civility and incivility, including agreement and disagreement, and reason and emotion. Each sentence in the comments section represented a different concept analyzed in this study. Therefore, each user’s comment represented two concepts, and civility or incivility was repeated in four total sentences throughout the comment section (see Supplementary Material). All usernames were vague and neutral and did not indicate any self-identifying features, such as gender or race.1 The manipulations were presented as a screenshot that included blue hyperlinks on two neutral phrases, social networking site share buttons, and an advertisement and links to other blog posts on the right-hand side of the page. Each comment contained buttons to allow readers to report abuse and vote on the comment. No comments had ratings. After reading the news blog post, participants were asked a number of posttest questions, such as attitudes, perceptions of bias in the blog post, and risk and benefit perceptions. This study used KnowledgePanel, a probability-based web panel designed to be representative of the United States. The survey was fielded from July 9 to 23, 2010, and took approximately 28 min to complete. GfK Knowledge Networks initially selects households using random digit dialing sampling and address-based sampling. Households are provided a laptop computer and Internet access, if needed. We used a poststratification weight provided by GfK Knowledge Networks to ensure the sample is representative of the U.S. population. GfK Knowledge Networks sent two email reminders to nonresponders. The completion rate was 54.2% (N = 2,338). The dependent variable is perception of bias in the blog post. Respondents were first prompted to “think only about the blog post and not its comments” and to rate their agreement with the statement, “This blog post is biased” (M = 4.70, SD = 2.29; 10-point scale). Previous studies have used similar measures of media bias (Flanagin & Metzger, 2000). Political ideology was created by averaging two items (social ideology and fiscal ideology) measured on 6-point scales, ranging from Very liberal (1) to Very conservative (6) (Pearson’s r = .75, p ≤ .001, M = 3.61, SD = 1.24). Attention to news in newspapers (Cronbach’s alpha = .92, M = 2.61, SD = 0.97), television (Cronbach’s alpha = .93, M = 2.72, SD = 0.96), and online (Cronbach’s alpha = .95, M = 2.11, SD = 1.04) is an average of five items: (i) international affairs, (ii) national government and politics, (iii) science and technology, (iv) scientific studies in new areas of research, and (v) social or ethical implications of emerging technologies (“None” [1] to “A lot” [5]). Trust in news media asked participants how much they trust news media to tell the truth about the risks and the benefits of technologies and their applications (10-point scale; M = 5.41, SD = 2.21). Trust in online-only sources is an index of two items on trust in blogs and Wikipedia to tell the truth about the risks and the benefits of technologies and their applications (Pearson’s r = .43, p ≤ .001, M = 4.51, SD = 1.66). Experimental conditions were included in the model using dichotomously coded variables. Issue identified whether respondents were assigned to the nanotechnology or nuclear energy condition (0 = “Nuclear energy,” 1 = “Nanotechnology”). Tone of comments indicated whether respondents read civil or uncivil comments (civility coded high). We analyzed these data using a hierarchical ordinary least squares (OLS) regression model. Independent variables were entered in blocks in order of their relative explanatory power: (1) individual characteristics, (2) media use, (3) trust in information sources, and (4) experimental conditions. Block 5 included the interaction, which was created by multiplying standardized versions of the variables to prevent collinearity between the interaction term and its components (Cohen, Cohen, West, & Aiken, 2003). Moderation was also tested using a conditional process modeling program, PROCESS, which uses a OLS path analytical framework to test for both direct and indirect effects (Hayes, 2013). In this case, Model 1 was used to examine the moderating effect of political ideology. The interaction, which was significant, was also probed for pattern of effect using the Johnson–Neyman technique (Hayes & Matthes, 2009). We conducted a manipulation check to ensure participants exposed to the civil condition perceived the comments as more civil than those in the uncivil condition. Participants in the civil condition (M = 6.38, SD = 2.29) were more likely to agree with the statement, “The blog post and its comments were civil” (10-point scale) than participants in the uncivil condition (M = 4.24, SD = 2.20), t(2,294) = −22.79, p ≤ .001 (r = .01067). Those in the uncivil condition (M = 5.74, SD = 2.64) were more likely to agree with the statement, “The blog post and its comments were impolite” (10-point scale) than those in the civil condition (M = 3.06, SD = 1.77), t(2,282) = 28.55, p ≤ .001 (r = .05).2 Our first hypothesis predicted incivility in online comments would increase perceptions of bias of the post. The data support our hypothesis. There is a negative relationship between civility in the comments and perceptions of bias (β = −0.10, p ≤ .001) (see Table 1). Those who saw uncivil comments perceived more bias in the news blog post compared with those who saw civil comments. We asked a research question about the interactive effect of ideology and exposure to incivility in comments on perceptions of bias in the news blog post. We found a significant interactive effect, indicating that ideology acts as a moderator of incivility in comments to affect perceptions of bias among respondents (β = −0.05, p ≤ .05). Using the Johnson–Neyman technique (Hayes & Matthes, 2009), we found exposure to uncivil comments affected perceptions of bias for 77.7% of respondents, that is, those who scored 2.84 (p ≤ .05) or higher on the political ideology scale. To visualize the interaction, we recoded political ideology into a nominal variable with three categories. The continuous variable was divided into tertiles, which approximated the points on the continuous scale corresponding to liberals, moderates, and conservatives. Political ideology did not affect perceptions of bias among those who encountered civil comments. However, those exposed to uncivil comments perceived significantly different bias in the news blog post, depending on political ideology. On one hand, moderates and liberals perceived similar levels of bias in the post regardless of the tone of the comments following the information. On the other hand, conservatives exposed to incivility perceived significantly more bias than their counterparts who saw civil comments (Figure 1). Interactive effect of ideology and tone of comments on perception of bias in the blog post. Error bars represent 95% confidence intervals. Note: Scale on y-axis is partially displayed The goal of this study was to explore and examine the polarizing effects of incivility. We find that uncivil user comments increase perceived bias in a news blog post to which they are connected. This indicates the “nasty effect” extends beyond effects of uncivil comments on issues covered in the media to perceptions of the media. Our findings also show that these effects are strongest for those who hold strong conservative ideological viewpoints. Thus, our study points to the increased polarization that can occur along existing divisions in society when incivility is present in online comments. There are some limitations of this study. First, our dependent variable was measured using a single item. However, random error in this measure would weaken the relationships we found. Had we been able to use multi-item measures, it is likely we would have found stronger relationships between our independent and dependent variables. In addition, other studies have used single-item-dependent variables to measure bias (Eveland & Shah, 2003; Lee, 2012). Future research should examine effects of incivility for multiple items regarding the news source. It is likely they will find even stronger relationships among ideology, incivility, and detrimental perceptions of media credibility. Second, while our model demonstrates statistical significance, the magnitude of these results is relatively small. Nonetheless, we repeated our results using a stripped down model containing only the most important predictors—trust, ideology, civility, and the interaction term—and found similar statistically significant coefficients for our predictors. Third, the large number of participants in our sample could present a type I error. Power analysis using G*Power (Faul, Erdfelder, Lang, & Buchner, 2007) indicates a sample size of 816 would be adequate to obtain the results in this model. Thus, it is likely that our model is overpowered. However, the stimulus used in the study is weak. The main manipulation appears after the main news media effects studies on an manipulation the of the stimulus we exposed the participants only to the Thus, we are our results are these aspects of our Finally, our survey was fielded in 2010, and our study should be with data to understand effects of incivility. People have more in over the with more than three of in a survey they have read the comments some (Stroud, & It is that individuals are more to online incivility than they were six This to more for incivility, which may the effects of incivility. increased of incivility may its The main effect of exposure to uncivil comments on perceptions of bias has important implications for news because of the of incivility in Research provides evidence that of comments, user a such as and user on comments encourage more civil discussion in 2015). In addition, interaction with more civility in discussions (Stroud, & 2015). such as these news are increasingly important the negative our study that uncivil comments can have on perceptions of news We also found ideology to moderate the effect of incivility on perception of bias in the news blog post. and moderates were less affected by incivility, while conservatives were most The concept of provides one for Political and are and have found a relationship between and political among & & and among political elites The of is an of & and can be viewed as and It may be that the around a central of is more for some than In other words, those with stronger views may be more susceptible to the effects of incivility compared with those with and conservatives to be the most affected by incivility. It may be that incivility is perceived as about a of who on a scale more susceptible to the effects of incivility. This is a of for which should the relationship between and political ideology, and the tone of comments. In addition, existing research suggests that perceptions of bias to be stronger among individuals who toward the of the political compared with those who & 2014; Lee, 2005). Our study suggests incivility can these existing divisions and encourage more polarity among those with moderate ideological views and those with stronger conservative viewpoints. Our findings are in the context of incivility, which is a concern about Internet discussion among news and information While we this tone to the concerns have over the nature of discussion on the research should explore whether similar in the context of or comments. This would extend the of our findings to the comment context more The online media environment a for the of content and news media, and our study one of this of uncivil comments. Our findings indicate that social context for how people perceive media sources, and is a potentially polarizing of communication in society. This is based on by a from the to the and in and the and or in this are those of the and do not the views of the Anderson of is an in the of and research include public and public engagement around issues of science and emerging technology, particularly in the online communication a Research in the for Yeo is an in the of the of research include science and risk public of emerging and information and is a and of the of of and a of the of the and for and research on the between media, and the and on the understanding of public in the context of controversial has and the The the and the in the of and is of the for in research with public on emerging technologies and the political effects of is an in the of and of the for Research the of has research in political science public and