Abstract
This study examines a way to promote civility in online political discussions through modeling discursive cues. An online experiment (N = 321) was conducted to investigate the impact of civil and uncivil discursive cues on participants’ mode of discussion. Results show that participants who were exposed to civil cues were more likely to engage in civil discourse themselves, stay on-topic, and offer additional perspectives in their comments. We also found that metacommunication (i.e., talking about the tone of discussion) engendered more metacommunication. This study illustrates the impact of modeling discursive cues and illuminates the possibility that participants in online discussion can improve their discursive environment.
At one time, the Internet was hailed as a space that could transform democracy, allowing more and different voices into political discussion. The anonymous nature of Internet discussion could strip conversation of extraneous cues, especially those related to status, and allow participants to focus on ideas and solutions. Unfortunately, that early optimism has been tempered by the realities of online discussions, which often include comments that are rude, demeaning, and less-than-democratic (Blom, Carpenter, Bowe, & Lange, 2014; Hlavach & Freivogel, 2011). In particular, the comments sections of news articles are widely acknowledged as home to some of the worst discussion on the Internet, prompting users to share articles on social media with the caveat, “Don’t read the comments!” (Moosa, 2014; Williams, 2015).
In recent years, news outlets have tried a variety of strategies to combat incivility in their comment sections. Huffington Post (2015) requires commenters to sign in with their Facebook accounts, a common strategy. Likewise, YouTube at one time required commenters to sign in with a Google+ account (Ferenstein, 2012). The New York Times (2015) pre-moderates comments and many sites hire full-time moderators to read, respond, and even ban users (Ingram, 2014; Sax, 2011). The results of these measures have been mixed; after all, real names do not necessarily make conversations more civil (Ferenstein, 2012), commenters can sign in with fake profiles, and quality moderation is expensive (Ingram, 2014; Sax, 2011). Many outlets, including heavy-hitters such as Popular Science, Reuters, Chicago Sun-Times, and CNN, eliminated or significantly reduced the number of stories with comments enabled in response to these issues (Gross, 2014; LaBarre, 2013).
Scholars, recognizing the still-unrealized democratic potential of these online spaces, have analyzed the problem of Internet incivility, and specifically incivility in news comments, in a variety of ways. Some solutions (e.g., Stroud, Scacco, Muddiman, & Curry, 2015) have been met with success, but it is clear that there is still much to learn. In this study, we investigate the influences of civil and uncivil discursive cues on comments posted about a news article, with an eye toward practical ways that online commenters can improve their own discursive spaces. More specifically, by considering the relationship between role modeling behaviors and observer responses, we offer a framework for encouraging civil discourse online.
Defining Civility
Civility is of interdisciplinary interest, with scholars from political science, communication, sociology, and other fields expressing dismay about what some call the “coarsening” of our political discourse (Kurtz, 2013). Scholars have reiterated the importance of civil discourse to an effectively functioning democracy (e.g., Annenberg Public Policy Center, 2011; Boyd, 2006; Papacharissi, 2004), but a clear definition of civility has been elusive (Herbst, 2010; Maisel, 2012). One area where the literature does agree is that politeness or mutual respect is a necessary, and for some, sufficient part of any definition of civility (Mutz, 2006; Mutz & Reeves, 2005; Ng & Detenber, 2005; Sobieraj & Berry, 2011). However, there are also concerns that overemphasis on politeness might inhibit the free flow of ideas in political conversation, resulting in what Papacharissi (2004) calls “a discourse that is so polite and restrained that it is barely human” (p. 266).
Sobieraj and Berry (2011) present a definitional middle ground, acknowledging that politeness facilitates democratic discussion, but not at the expense of conflict or disagreement. In their view, civility is “characterized by speakers who present themselves as reasonable and courteous, treating even those with whom they disagree as though they and their ideas are worthy of respect” (p. 20, emphasis added). Such discourse is polite, but also addresses conflict in such a way that the threat to another’s face is lessened (Williams & Humphrey, 2007). Civility, then, goes beyond mere politeness to also offer an explicit affirmation of the value of the person (and ideas) with whom the speaker disagrees. Similarly, incivility can be defined as lacking in respect for others and their ideas. Coe, Kenski, and Rains (2014) describe incivility as, “features of discussion that convey an unnecessarily disrespectful tone toward the discussion forum, its participants, or its topics” (p. 660). An additional key part of this definition, as noted by Coe et al., is that incivility is viewed as unnecessary. This could manifest in comments that involve name-calling, lying, aspersion, vulgarity, or the use of pejoratives (Coe et al., 2014). In drawing on past examples of how incivility has been conceptualized, Coe et al. suggest that the key distinction lies within the relevance of information shared, stating, “uncivil comments do not add anything of substance to the discussion” (p. 660).
Incivility in Online Discussions
Citizens engage in substantial political conversation online, not just in politically oriented spaces, but also in nonpolitical discussion groups (Stromer-Galley, 2003; Wojcieszak & Mutz, 2009). Unfortunately, incivility has been a common feature of online political discourse since the earliest days of the Internet. Benson (1996) examined Usenet discussions, which were online groups focused on a common topic and, not surprisingly, found the political groups particularly vitriolic. As the Internet has evolved, so have the spaces in which political discussion occurs, and many studies in recent years have focused on incivility in news comments. Eisinger’s (2011) study of several news outlets found significant amounts of incivility in comments on a variety of political topics. Coe et al. (2014) reported that one in five comments on a local newspaper’s website were uncivil. Coffey and Woolworth (2004) analyzed comments on a local newspaper’s online discussion forum and found numerous instances of angry, extreme discourse that were not present in public meetings about the same subject, suggesting that norms related to civility may be different in an online setting.
Many scholars argue that the anonymous nature of online comments, as well as a lack of face-to-face social reprimands from other posters, contribute to the presence of uncivil discourse (Anderson, Brossard, Scheufele, Xenos, & Ladwig, 2014; Coffey & Woolworth, 2004; Eisinger, 2011; Ng & Detenber, 2005; Papacharissi, 2004; Rowe, 2015). Indeed, past work has found that requiring those who want to comment to connect their real name to the post has some notable impact on the quantity and quality of posts (Fredheim, Moore, & Naughton, 2015; Santana, 2014). In particular, Fredheim et al. (2015) indicate that the variance in posters declines greatly (i.e., readers simply refrain from posting if they have to authenticate their identity) and fewer offensive words and all caps (that would likely denote shouting) were observed. That said, incivility is not entirely removed by the presence of real names; Santana (2014) highlights this in noting that roughly a third of non-anonymous posts still contained uncivil remarks in their analysis of newspaper comment sections.
Likewise, Citron (2011) noted that real names can foster the potential for harassment if posts turn personal, pointing out that users are often targeted on the basis of race and/or gender. Citron and Norton (2011) offer exemplars of the types of graphic and threatening posts that targeted groups regularly endure online. Similarly, Sobieraj (2017) identified distinct patterns of online misogyny that seek to discredit, shame, and intimidate women online, which drastically limit women’s participation in digital spaces. Gray’s (2012) study of online gaming communities examined the particular types of harassment that women of color receive online (racism, sexism, heterosexism, and discrimination based on citizenship and linguistic abilities), and describes her own contested efforts to raise consciousness online. Harassment impedes the ability of targeted groups to fully participate in civic life online because it suggests that not all commenters are equal or welcome. It also promotes prejudiced views of others and their ideas, which Citron and Norton (2011) describe as a “defect in the marketplace of ideas” (p. 1451).
Incivility on these sites can have meaningful, negative impact on readers. The lack of respect for others created by uncivil comments can impede the ability of everyday people to engage in productive dialogue about issues. Uncivil discussion can also induce feelings of anger and aversion (Chen & Lu, 2017; Coe et al., 2014; Gervais, 2015) and lead to the perception that uncivil discussants are more aggressive and less credible than civil discussants (Ng & Detenber, 2005). Research from Han and Brazeal (2015) found that incivility dampened the willingness of others to participate in the online conversation at all, much like negative advertising in political campaigns can discourage voting (Ansolabehere, Iyengar, & Simon, 1999). Other research, however, has found that the heightened emotions caused by incivility can actually encourage others to jump into the conversation with uncivil comments of their own. Sometimes these comments are in response to feeling attacked for a position they hold (Chen, 2017; Chen & Lu, 2017; Wang & Silva, 2018). Other times the comment serves as an effort to join attackers in “piling on” someone who is already a target of uncivil comments (Gervais, 2015). It is important to note that research suggests that the quality of these uncivil conversations is often much lower than civil conversations (Coe et al., 2014).
On a larger scale, incivility can also undermine democracy as a whole. One well-known study by Mutz and Reeves (2005) found that viewing incivility resulted in a loss of trust in politicians, all of Congress, and the entire political system. They warned that low levels of trust threaten both the stability of political institutions and the ability of those institutions to function properly. Uncivil conversations can also lead to opinion polarization, particularly when partisan in nature (Anderson et al., 2014; Suhay, Bello-Pardo, & Maurer, 2018). Such polarization is both attitudinal, which involves a hardening of attitudes toward those with opposing viewpoints, and social, with express preferences for family and neighbors to be politically like-minded. This has the potential to limit the range of viewpoints to which people are exposed. Additionally, the media, which play a critical role in holding leaders accountable, can suffer a loss of credibility when uncivil comments are present online (Anderson, Yeo, Brossard, Scheufele, & Xenos, 2016; Waddell, 2017). A lack of trust in government, unwillingness to associate with people who do not share one’s views, and increased skepticism of the media have the potential to do damage to the way our democracy functions. Given the prevalence and the potential impact of uncivil online discussions, exploring potential remedies for incivility merits scholarly attention.
Modeling Civility in News Comments
One framework that can be applied to better understand the potential for positive influence online is modeling. Bandura’s (1977, 2001) social cognitive theory elevates the role of modeling in human learning, explaining that we learn most behaviors through the observation of models. While Bandura’s early studies concerned children and learned aggression, he quickly saw the potential of the theory to be extended to other types of situations and behaviors (Bandura, 2011). While many of our models will be the people around us, models also appear in various forms of media including models that appear live, symbolically, electronically, or in print (Bandura, 2002; Schunk, 2012). Bandura positioned his theory in contrast to behaviorist schools of thought prevalent at the time, arguing that observational learning was a much more efficient method than the trial-and-error approaches favored by behaviorist scholars (Bandura, 2011).
As Bahn (2001) explained, social learning involves the “vicarious acquisition of knowledge” (p. 111), a process Bandura calls vicarious verification. First, the learner gives attention to and observes several models. Then, the knowledge gained from these observations is cognitively processed and internalized, so that the behavior can then be performed. Although reinforcement is helpful, it is not required for learning (Bandura, 1977). This approach, according to Bandura, allows a person to learn more quickly and avoid experiencing the negative consequences of incorrect behaviors for themselves (Schunk, 2012).
Studies that focus specifically on civility in news comments generally support the idea that the level of civility can be influenced by modeling. Han and Brazeal (2015) examined the impact of role models on group members’ subsequent comments and found that participants in civil conversations tended to be more civil and, in fact, mimic the language used by civil role models. Recent work from Molina and Jennings (2017) also found evidence for modeling civility through comments on Facebook news articles. In their experimental study, Sukumaran, Vezich, McHugh, and Nass (2011) investigated if situational norms set by previous commenters were modeled in the comments left by participants. The study revealed that those who were exposed to more thoughtful comments were more likely to provide thoughtful comments in response than those who were exposed to unthoughtful comments.
Given the previous literature on social cognitive theory, modeling, and news comments, we explore the possibility that civility in discussion can influence the mode of news comments positively. Specifically, we predict that in an online discussion environment where civility is modeled, people are more likely to provide civil comments:
H1a. When exposed to civil comments, participants are more likely to post comments that are civil.
Research also demonstrates the link between civility and relevance to the issue at hand. Relevance, or focusing on the issue at hand, is so integral to the notion of civility that Coe et al. (2014) identify it as one of the defining factors in determining what is civil. Despite its importance, few studies of news comments have examined whether comments are relevant. Ng and Detenber (2005) point out that flaming, an extreme form of incivility, shifts focus from the topic at hand to the impoliteness of the commenter. This diversion of attention could result in comments that are not pertinent to the issue at hand. Conversely, without such diversion, civility can promote message elaboration and issue-relevant comments. Molina and Jennings (2017) reported that civility on Facebook encouraged participants to engage in message elaboration, which resulted in issue-relevant thoughts. Stroud et al.’s (2015) study on a reporter’s involvement in online comments also suggests that, in general, the presence of a moderator to promote civility may lead to more genuine questions and relevant comments. Together, these two studies would support the idea that incivility diverts focus from the topic, while civility can encourage the potential to engage in relevant discussion on the topic through message elaboration. With these studies in mind, we anticipate that the mode of civil discussion leads to relevant comments:
H1b. When exposed to civil comments, participants are more likely to post comments that are relevant to the issue at hand.
Similarly, scholars suggest that civility opens space for a variety of ideas to be expressed. Arnett (2001, p. 328) states, “civility brings persons together, not as an empty gesture to maintain the status quo, but to permit diverse voices to enter the public conversation.” Herbst (2014) also argues, “civility is being open to exploration rather than just presenting information” (p. 9). This openness to diverse opinions and the face-saving nature of civil discourse may encourage discussants to feel more open and willing to share their ideas with strangers without fear of retribution. In fact, scholars have found that civil comments generally increase willingness to participate (Han & Brazeal, 2015; Molina & Jennings, 2017), which can open up space for new ideas. In addition, as mentioned earlier, civility can lead to greater elaboration of the issues at hand (Molina & Jennings, 2017). This, in turn, could promote diverse perspectives. This aligns well with the study on modeling “thoughtful” comments online, which found participants responding in kind (Sukumaran et al., 2011). Following these ideas, then, civil discussion should encourage different ideas from participants. Therefore, we anticipate that, under the mode of civil discussion, people are more likely to offer additional perspective to the discussion:
H1c. When exposed to civil comments, participants are more likely to post comments that offer additional perspective on the issue.
Counter(modeling): The Role of Metacommunication
A final outcome related to civility, and more specifically, the lack thereof, is the emergence of metacommunication regarding the tone of conversation. While research has generally supported the potential to model civility, the same has not been found in the case of incivility. Instead, users opt to speak out against the incivility. For example, Phillips and Smith’s (2004) work on everyday incivility found that social sanctions against the uncivil were regular occurrences. Lanamäki and Päivärinta’s (2009) study of metacommunication in online communities found that ad hoc conversations, such as those in a newspaper comment section, contained both topical and metacommunicative content. Among the patterns of metacommunicative content they identified was content that specifically addressed communicative practices of group members and their impact on the community.
Through discursive interaction, the appropriateness of certain communicative practices can be established, questioned, and reformed. This happens in news comments as well. Ruiz et al. (2011), for instance, pointed out that it is not at all unusual to encounter news commenters telling other users to tone down their comments. In their exploratory study, Han and Brazeal (2015) found that those who were exposed to uncivil online comments were no more likely than those who were exposed to civil comments to make uncivil comments. Rather, they engaged in metacommunication by “comment[ing] on the overall tone of the discussion” (p. 23). While the existing research is somewhat divided on whether uncivil comments lead to (Gervais, 2015; Kim & Herring, 2018) or do not lead to (Han and Brazeal, 2015; Rösner, Winter, & Kräme, 2016) uncivil comments, it does suggest that metacommunication arises in the presence of incivility (Ruiz et al., 2011; Santana, 2014). Thus, we test the following hypothesis that the uncivil mode of discussion will lead to metacommunication:
H2. When exposed to uncivil comments, participants are more likely to engage in metacommunication.
Another important aspect of metacommunication is the response such comments elicit from others in the community; that is, the potential for the presence of metacommunication to improve the civility of the discussion. Ostensibly, the purpose of metacommunication is to change the tone of the conversation, making it more civil. However, few studies have examined whether this is what actually occurs. Recent work from Molina and Jennings (2017) shows promise in noting that, “cues that scold incivility can encourage individuals to engage in a Facebook conversation by fostering more elaboration when processing arguments” (p. 16). While their study was specific to Facebook, likely the potential to model metacommunicative behavior and promote civility extends to other platforms as well. In order to better understand how metacommunication impacts the tone of a conversation, we present the following research question:
RQ1. Does the presence of metacommunication about the tone of comments promote civil comments?
Method
To examine these hypotheses and research question, this study utilized an online experiment using the web-based survey software Qualtrics. A posttest only between-subjects experimental design was implemented with random assignment to one of three conditions: the civil condition, the uncivil condition, and the metacommunication condition.
Procedure
Each participant completed a questionnaire regarding media habits and then was randomly assigned to one of three conditions: the civil condition, the uncivil condition, and the metacommunication condition. In all three conditions, participants read two newspaper articles—one distractor article on the topic of fast food consumption and another on the issue of gun control. After reading the article on gun control, participants were randomly assigned to read comments that were either civil, uncivil, or uncivil with metacommunication (detailed in Experimental Manipulation, below). Participants were then asked to type in their own comments immediately below the comments they read. After typing their comments, participants were directed to complete a series of questions pertaining to their attitudes and demographic characteristics.
Participants
Participants were recruited through the participant pool of the Communication Studies department at a large Midwestern university and provided with extra-credit upon completion of the study. A total of 321 students completed the experiment (age, M = 19.7, SD = 3.0). Of the participants, 53% was female; 18% identified themselves as strong Republican, 25.2% as not so strong Republican, 34.0% as Independent, 12.8% as not so strong Democrat, 5.9% as strong Democrat, and 4.4% as other. In terms of political ideology, 6.2% of participants identified themselves as very conservative, 26.8% as conservative, 33.6% as middle-of-the road, 14.0% as liberal, 4.0% as very liberal, and 14.6% as “haven’t thought much about it.” In addition to demographic information, participants were also asked about gun ownership. In all, 62% of participants reported that either they or their family owned a gun, 33.6% responded neither they nor their family owned a gun, and 3.7% reported “don’t know.” Chi-square tests for sex, party identification, political ideology, and gun ownership indicated no significant differences between the experimental conditions.
Experimental Manipulation
To create a realistic comments section, an actual news article from The Chicago Tribune (chicagotribune.com) was used alongside comments culled from various news sites. To maintain consistency between conditions, only a small portion of the comment sections were manipulated (see the Appendix). In constructing the model of “civil” comments, the researchers relied on Sobieraj and Berry’s (2011) definition of civil discourse. Based on their definition, comments in the civil condition were created in such a way that disagreements were expressed respectfully. In addition, a fellow discussant was addressed by name (e.g., “I respectfully disagree, Pat.”) to add politeness and lessen the perception of threat to face (see Williams & Humphrey, 2007).
In creating the model of “uncivil” comments, the researchers inserted language markers that are considered “uncivil” (see Gervais, 2011) within each comment. Comments in the uncivil condition included insults and dismissive language. Disagreements were addressed using rude comments and name-calling (e.g., “You idiot”) and political labels were also inserted to add hostility to the discussion (e.g., “Liberal bullshit”).
For the metacommunication condition, the researchers inserted the following comment to the uncivil condition:
Why do people have to be so ugly? I know this is a heated debate, but there is no need for name-calling or obscene language. If people would refrain from being nasty to each other, we can have more intelligent and coherent conversation.
The manipulation check confirmed that the civil condition and the uncivil conditions (with and without metacommunication) were perceived as significantly different in terms of the level of civility (see Figure 1).
Manipulation checks for civil versus uncivil experimental conditions. The civil discourse version was consistently perceived as more respectful (F(2, 274) = 97.05, p < .001), calm (F(2, 275) = 74.46, p < .001), and friendly (F(2, 287) = 61.17, p < .001), while the uncivil version was perceived as more rude (F(2, 307) = 92.19, p < .001) and emotional (F(2, 315) = 26.26, p < .001) by participants. Post hoc pairwise comparisons show significant differences between the civil and the uncivil condition and the civil and the uncivil with meta-communication condition, but not between the uncivil and the uncivil with meta-communication conditions.
Measures
To examine the impact of modeling civility on news comments, the researchers analyzed comments provided by each participant. Each response was coded according to the following variables.
Level of Civility
This variable examined the general tone of participants’ comments in three levels—uncivil, polite, and civil. Comments that included an element of incivility (i.e., obscene language, vulgarity, insulting language, name-calling, ideologically extreme language, and dismissive language) were coded as “uncivil” = 0 (e.g., “ChrisB sounds like an idiot,” “Your argument is ridiculous”), comments that were devoid of incivility were coded as “polite” = 1 (e.g., “I don’t think taking guns away from law-abiding citizens will help these issues”), and comments that explicitly acknowledged different viewpoints in a respectful manner were coded as “civil” = 2 (e.g., “We should not be restricted from having guns, but as is the case there is some truth to the opposing argument. People should be educated and be made to understand what a responsibility it is to have a gun”). If comments contained elements of both civility and incivility (e.g., “For a psycho-liberal, you’ve actually made a good point”), they were coded as “uncivil” due to the presence of incivility.
Relevance
This variable indicates if comments were relevant to the issue at hand. Comments that were relevant to the topic of gun control were coded as “1,” other comments were coded as “0.” Comments that included only singular term responses (e.g., “wow”) were also coded as not relevant.
Additional Perspectives
This variable examined whether participants offered a new idea on the issue that was not discussed previously. This variable was added to capture the scope of the ideas provided by the participants. Comments that offered an additional perspective were coded as “1” (e.g., “Maybe the government could require classes to take in college so people can be educated on the matter”), all others were coded as “0.”
Metacommunication
This variable investigated whether participants commented on the overall tone of the discussion. Comments that addressed the tone of the discussion were coded as “1” (e.g., “Stop being so nasty,” “There is no need to name call,” “Yelling at each other won’t do anything”), all others were coded as “0.”
After an initial training session, a researcher involved in this study and two graduate students who were blind to the purpose of the study coded comments individually. Coding disagreements were resolved through in-depth discussions. Krippendorff’s alpha indicates acceptable levels of intercoder reliability for all variables (level of civility = .76, relevance = .79, additional perspective = .71, metacommunication = .80).
Results
A total of 321 comments were collected and analyzed. The average length of comments was 43.7 words (the civil discourse condition = 48.0 words, the uncivil discourse condition = 43.8 words, the metacommunication condition = 40.2 words).
The hypothesis that the mode of civil discussion would lead to civil comments (H1a) was supported. A chi-square test of independence indicated that there was a significant difference in the level of civility between the civil and uncivil conditions, χ2 (2, N = 204) = 8.29, p = .008, V = .202. Post hoc analyses using adjusted standardized residuals revealed that the largest discrepancy was found in the civil comments (z = 2.9). As Table 1 shows, 23.3% of participants in the civil condition wrote comments that were considered civil, acknowledging views different from their own in a respectful manner, while just 8.8% of those in the uncivil condition did.
Percentages of Use in Open-Ended Responses, by Experimental Condition.
Not only did the participants in the civil condition follow the mode of civility, they were also more likely to stay on-topic (H1b). The results showed that while over 20% of participants in the uncivil conditions made comments that were irrelevant to the issue of gun control, less than 9% of participants in the civil condition did. A chi-square test indicated that the difference between the civil and the uncivil conditions was significant, χ2 (1, N = 204) = 4.97, p = .013, V = .156.
Support was likewise found for H1c; participants in the civil condition were more likely to provide additional perspectives to the discussion compared to the uncivil condition. Our data showed that 43.3% of participants in the civil condition offered a new perspective that had not been discussed earlier, while 25.4% of those in the uncivil condition did, χ2 (1, N = 204) = 7.25, p = .004, V = .188. For instance, one participant in the civil condition broached the idea of “apathy” on the part of non-gun owners and further elaborated on the complexity of the issue by linking it to the use of guns in sports: I agree partially with each side. There are loopholes that I believe should be closed but I also feel that criminals by nature won’t be deterred by a majority of these laws. Also there is an issue of apathy on the side on non-gun owners as the freedom in question is not one they exercise, so they are by and large unopposed to most any suggested gun law…It is however a complicated subject and can’t be simplified to “self defense, nothing more and nothing less” as there are also those who are sportsmen, (such as hunters and exhibition shooters), and also collectors, especially of the historical variety. There should be limits as to what the average citizen may keep in his arsenal, but I think both sides need to more sensitive to the opinions of the opposition to achieve a more balanced compromise.
These findings indicate that participants in the civil condition followed the mode of civility, were less likely to go off-topic, and were more likely to offer a fresh perspective to the discussion.
Regarding metacommunication (H2), our hypothesis that the uncivil discourse would lead to metacommunication was supported. Results indicated that those who were in the uncivil condition were almost five times more likely to engage in metacommunication (expressing dissatisfaction with incivility in the discussions) compared to those in the civil condition (10.5% vs 2.2%). A chi-square test indicates that the difference was significant, χ2 (1, N = 204) = 5.42, p = .01, V = .163.
The following textual examples illustrate that those who were in the uncivil condition lamented incivility in their comments. One participant in the uncivil condition wrote, “Enough everybody! Name calling and obscene language won’t solve this problem.” Some also called for working together to solve the problem rather than calling names stating, “How about instead of calling each other names and arguing with each other, let’s focus on the matter on hand . . . Let’s work together to stop more of these incidents from happening.”
Finally, we found that the presence of metacommunication did not lead to civil comments (RQ1). Results show that those who read uncivil comments with metacommunication were no more likely to provide civil comments than those who read uncivil comments without metacommunication (6.8% vs 8.8%), χ2 (2, N = 231) = .896, p = .639, V = .062. What was interesting, however, is that participants who were exposed to metacommunication were more likely to comment on the tone of incivility themselves. We found that 18.8% of the participants in the metacommunication condition used metacommunication in their comments compared to 10.5% of those in the uncivil condition, and the difference was marginally significant, χ2 (1, N = 231) = 3.15, p = .076, V = .117.
Discussion
Increasingly, news media are seeking to find ways to encourage civil discourse online; in some cases, this is through moderation (Ingram, 2014), attaching one’s real name to posts (Ferenstein, 2012), or removing the potential to post altogether from articles (Gross, 2014). However, forcing users to disclose identity (i.e., the real name approach) only does so much to decrease incivility online (Santana, 2014) and could potentially lead to harassment (Citron, 2011). Likewise, eliminating the potential for discussion is not ideal for newspapers who often seek to engage with their audience (Gross, 2014; LaBarre, 2013). Given the interactive potential of online newspapers (Rowe, 2015), this study set out to examine how modeling can create more civil discourse in the comment sections of an online newspaper website.
We found that within the context of civil discussion, people would model civility in their comments (H1a), were more likely to focus on the topic at hand (H1b), and brought additional perspectives to the table (H1c). These findings suggest that civil discourse can indeed be modeled and create more civil and robust online discussion. On the other hand, we found that incivility did not evoke a significant amount of uncivil responses. Instead of modeling the uncivil discourse, participants opted to engage in metacommunication, voicing their frustrations and calling for a more civil conversation (H2). This is in line with Rösner et al.’s (2016) study on modeling, which suggested that uncivil comments increased aggressive feelings, but not the potential to model incivility by respondents. Recent research on modeling civility in comments on a news article on Facebook also found similar results wherein the civil condition was modeled but the uncivil condition was not (Molina & Jennings, 2017).
Interestingly, Molina and Jennings (2017) suggest that the lack of a hot-button issue (their study used a newspaper article about GMOs) could explain for the lack of modeling. However, as this study shows, even when accounting for a hot-button topic (gun control), no such modeling occurred. Santana’s (2014) research on incivility within the comments on news articles may shed light on this to some extent, wherein the author notes that in studying a hot-button topic (immigration) civility was often discussed among participants, with users opting to instead metacommunicate on the topic. Examples offered by Santana highlight this phenomenon, with one commenter noting, “sorry, joe. we will not come to an agreement on this issue. I’ll bow out before it comes to insults” (p. 28), and another writing, “I simply cannot understand the intolerance and hatred you folks exhibit here” (p. 28). In this regard, some users see where the conversation is headed and opt to avoid further (or potential) incivility. More research on what causes some individuals to reach this conclusion versus others, however, is still needed; it may be that metacommunication supersedes the focus of the initial article, which can detract from the benefits of civil deliberation on the topic (Annenberg Public Policy Center, 2011; Boyd, 2006; Papacharissi, 2004).
Participants taking the initiative to redirect the discussion to a more civil, meaningful one by commenting on the tone of posts while not reflecting modeling within the uncivil condition, does highlight the potential effects of platforms, something that future work might explore. For example, this may be participants’ attempt to correct the current direction of discussion, taking on the role of moderator. In an online environment where most discussions happen without moderators, this type of behavior by participants themselves is helpful and much needed. The role of moderator is supported by our results in showing that the metacommunication condition saw almost twice as much metacommunication produced compared to the uncivil condition. Notably, Feenberg, Xin, and Glass (2014) posit that, “moderators play an important role in initiating and sustaining metacommunication . . . metacommunication is particularly important as a means for re-establishing a threatened communication link by calling attention to problems in the communication process” (p. 15). Considering the significance of metacommunication and the absence of moderators in most online discussions, it is encouraging that participants engaged in modeling. When participants saw one person voicing his/her disapproval of the tone of the conversation, they modeled that behavior and joined the call for more civil discourse.
This seems to indicate that, instead of following a presumed norm of discussion, participants opted to model the exemplary behavior of other posters on the forum in trying to set a proper tone for discussion. It is worth noting, however, that despite another discussant’s call for civil discourse, participants in the metacommunication condition were no more civil in their comments than those who were in the uncivil condition. This may be attributed to the fact that participants in the metacommunication condition were focused on the discourse itself, trying to set a proper tone for the conversation, rather than providing their opinions on the issue in a civil manner. It is also possible that our manipulation may have contributed to the result. Although participants were not the targets of the reprimand, the comment used in our manipulation (“Why do people have to be so ugly?”) may have been perceived as hostile or negative. Given the importance of metacommunication and its potential to create a more civil discussion, future work should explore the impact of metacommunication on subsequent discussions and examine the effects of different types of metacommunication.
Overall, this study offers several implications. First, civil participants may have more control over the tone and the scope of online discussions than they think. Engaging in the civil behaviors they want others to use and creating the norm of civil discourse could potentially have positive impacts on the civility of their political conversations and multiplicity of the ideas expressed in the discussion. Additionally, there may be value in participants challenging uncivil behaviors. Our study found that if one participant was willing to speak out about the perceived incivility of the conversation, others would follow the lead and attempt to set a proper tone for the conversation. Papacharissi (2004) found that small outbreaks of incivility and impoliteness were quelled by other participants, and that apologies for incivility often followed. While metacommunication may not lead to a sudden change in civility, the presence of it, particularly if it comes from several participants, could encourage some posters to set a tone for other participants. Future research should examine the effects of metacommunication on the rest of the conversation, as this could be a useful strategy for moderators.
Limitations and Future Directions
There are several limitations to this study. First, as with many other studies, we used college students as our sample. Therefore, the results are limited regarding generalizability. That said, given the general interest in mobilizing youth around deliberation and discussion through technology (see Peacock & Leavitt, 2016), there is value in focusing on this particular sample. Ideally, future studies would examine the impact of modeling using more representative samples of the public to expand these results beyond youth.
Second, we examined news comments on a single topic, gun control. Some recent studies indicate that the quality of comments could vary depending on the topic of the post (see Coe et al., 2014; Halpern & Gibbs, 2013; Stroud et al., 2015). As discussed earlier, however, despite being a hot-button issue, the results of this study were like those of recent work considering comments on articles about GMOs (Molina & Jennings, 2017). Therefore, further research is needed to investigate under which circumstances discussants are more likely to adhere to modeling.
Another consideration is the absence of a control condition within the experiment; it is possible that the lack of incivility, rather than the presence of civility, could be what is being modeled by participants. Future research would benefit from including this third condition to better understand the relationship between civility and incivility within discussion board posts.
Finally, like many other experimental studies, the issue of experimental realism should be considered. While our experiment featured many qualities that mirror real online news comment sections, given that it was conducted using survey software, it is possible that some may not have perceived the discussion as “real.” It should also be noted that participants were asked to comment on a topic that they may not have been interested in or might otherwise choose not to comment. Consequently, our findings should be interpreted with caution. Future studies should examine how modeling works in real online conversations. Given the importance of civil and open discussions for the proper functioning of democracy and the potential of online political discussions, we should continue exploring a better way to create and sustain civil and robust online discourse.
Footnotes
Appendix
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.