Who will intervene to save news comments? Deviance and social control in communities of news commenters

Abstract

Which bystanders will confront racist, misogynist, personal attacks in news comment sections? This article applies sociological concepts of deviance and social control to categorize efforts to moderate online news comments. Three dimensions of social control are theorized: affirming and sanctioning social control, formal and informal social control, and direct and indirect social control. Particular focus is on indirect informal social control (i.e. rating and reporting of news comments) in order to examine which users are likely to intervene to maintain social order. An analysis of secondary data from a survey of online news users found that demographics play an important role—younger, wealthier, White, males are most likely to report abusive comments. Trust in the news media and authoritarian personality traits also significantly predicted bystander intervention. Theoretical implications for the role of social control in enforcing social norms in news comment spaces and for professional comment moderation are discussed.

Keywords

Bystander intervention deviance digital journalism journalism studies news comments online news social control sociology

Deviant behaviors are threatening to tear apart communities of news commenters. Commenters are fleeing some communities, and other commenting spaces are being shuttered and razed because of deviant, abusive, misogynistic, racist, and antisocial posts (Finley, 2015; World Association of Newspapers and News Publishers, 2016). Which “bystanders” in these communities can be relied upon to intervene and try to save these communities from complete social disorganization?

Because news commenting is plagued by deviant social behaviors, this article argues for applying the sociological concept of social control. The concept of social control is not completely new to journalism studies (see, e.g. Warren Breed’s [1955] paper, “Social control in the newsroom”), but applying the concepts of social deviance and social control to understand how social norms are reinforced in participatory media spaces such as online comment sections is novel.

A degree of social conflict is foundational to social progress. By challenging entrenched stasis, conflict spurs the reevaluation of social norms and may result in social innovation (Coser, 1956; Simmel, 1955). Social conflict can also bring together a community responding to an outside threat. Furthermore, under some circumstances (e.g. to protect minority rights) behaviors that would otherwise be deviant (e.g. civil disobedience, resistance, armed intervention, etc.) can be considered noble to the extent that they reinforce pro-social values (Simmel, 1955). The goal of social control, then, is to not eliminate social conflict altogether. Rather, social control is defined as a series of interventions that encourage pro-social behaviors (i.e. affirmative social control) and discourage deviant behaviors (i.e. sanctioning social control). Deviant behaviors are not positive departures from existing norms (e.g. a robust public debate that considers and respects dissenting points of view); rather, such behaviors result in disorganization and a withdrawal of individuals from community (e.g. ad hominem personal attacks in news commenting spaces), ultimately to the detriment of society.

Social control is much more commonly used in studies of neighborhoods’ efforts to address criminal deviance (Bradford and Jackson, 2016; Ohmer and Beck, 2006; Sampson et al., 1997; Warner, 2007b, 2014). We suggest, though, that the concepts of deviance and social control can be useful in examining social dynamics of news comment moderation, which involves the symbolic participation and cooperation of authorities (i.e. formal social control exercised by professional moderators) and that of community members (i.e. informal social control exercised by news commenters).

The primary contribution of this article is to theorize social control in participatory commenting spaces along three dimensions (see Figure 1). The primary dimensions are formal (e.g. actions taken by moderators) and informal social control (e.g. actions taken by commenters themselves), and direct (e.g. deleting an abusive comment) versus indirect social control (e.g. requiring users to sign in using their Facebook profile). As previously discussed, social control can also be affirmative (e.g. users recommending or praising helpful comments) or sanctioning (e.g. users reporting abusive comments to moderators).

Figure 1.

Types of social control applied to traditional communities and communities of news commenters.

This article also presents an analysis of survey data that provides an exploratory look at which “residents” (i.e. commenters) of online news commenting communities will “do something” to intervene to address deviant social behaviors. Based on existing sociological literature that has examined residents’ willingness to address deviant behavior in traditional communities, we hypothesize that the propensity to engage in informal social control (e.g. rating and reporting comments) is a function of demographics, self-efficacy, trust in the media, and authoritarianism. Implications for theory and comment moderation are discussed.

Communities of news commenters and deviance

In 1998, the Rocky Mountain News began allowing people to comment under news stories, making it one of the first news sites to make space for online comments (Santana, 2011). In 2006 and 2007, The Washington Post and New York Times, respectively, set up commenting sections (Editor & Publisher, 2007), and the vast majority of news organizations online have done the same in the past 15 years, most by creating space at the bottom of articles to allow readers to express their thoughts and opinions (Santana, 2016). In the United States, for example, more than 90% of newspapers have adopted online comment sections as a way to attract and engage more online readers (Santana, 2014).

News comment forums constitute “communities” of commenters, albeit loosely knit ones. According to McMillan and Chavis (1986), the sense of community comprises four parts: membership, influence, integration and fulfillment of needs, and shared emotional connection. Commenters frequently refer to one another by screen names and discuss topics among themselves by replying to other users’ comments (Kalch and Naab, 2017). These activities build a sense of membership and shared emotional connection—or at least shared participation in the community—among news commenters. Membership in the community fulfills commenters’ needs to form opinions about news topics and engage in personal self-expression. News commenters will also sometimes provide additional sources for further reporting and assist in fact-checking (Robinson, 2015). These collaborative acts between news users and professional journalists, part of a broader set of activities given the label “participatory journalism” (Singer et al., 2011), may give members of the community of commenters a sense of influence. Indeed, when the New York Times conducted interviews with their “Super Commenters,” they found that identification with a community of commenters was a primary motivation for users’ commenting activities (Long, 2017).

Communities of news commenters, though, are frequently marred by social deviance. In announcing its decision to eliminate comment sections at the bottom of its articles, NPR shared an email from a user: “The comments have devolved into the Punch-and-Judy-fest of moronic, un-illuminating observations and petty insults I’ve seen on pretty much every other Internet site that allows comments” (Jensen, 2016). In addition to NPR, CNN, The Verge, Toronto Star, and Reuters are among major news organizations that have eliminated comment features at the end of their news articles (Finley, 2015; World Association of Newspapers and News Publishers, 2016).

As with other types of communities that struggle to control deviant acts (e.g. “the inner city”), perhaps communities of news commenters have been unfairly painted with a broad brush. Deviant commenting does not comprise the majority of social interactions in these communities. The percentage of deviant comments across different sites is estimated to be between 6% and 47% (Kalch and Naab, 2017). Many news comments have positive benefits, serving as a source of entertainment (Erjavec, 2014), allowing readers to formulate and express political opinions (Watson et al., 2017), and augmenting news events with additional first-person accounts (Robinson, 2015). Certainly, though, to the extent that a significant portion of news comments are abusive—and the problem is growing more severe in recent years (Moor et al., 2010)—deviant commenting must be better controlled to prevent individuals from withdrawing from communities of news commenters and to prevent those communities from devolving into complete social disorganization.

In a perfect world, perhaps news commenting would adhere to the standards of deliberation, with people discussing and exchanging views on important matters, providing reasonable arguments, and listening to others’ perspectives in a respectful manner (Price, 1992). The purpose of deliberative discussion is to increase knowledge, incite generation of different views, and facilitate decision-making (Cappella et al., 2002; Mutz, 2008). Thus, rationality and politeness are presumed by some to be requirements of online deliberation (Steenbergen et al., 2003; Stromer-Galley, 2007). It would be naïve, though, to assume that news commenters, many of whom still comment anonymously, would conform to ideal standards of deliberation. Imposing deliberative expectations on news commenting also glosses over the fact that even comments that fall short of the ideals of rationality, for example, may still have social, political, and journalistic value (Freelon, 2010). Some degree of respect for social norms, such as basic politeness and decency, however, has to be a basic, foundational expectation of communities of news commenters if they are going to maintain any semblance of social order.

Many scholars have focused on incivility in news commenting sections. Incivility has primarily been defined normatively. Papacharissi (2004), for example, defined uncivil comments as “a set of behaviors that threaten democracy, deny people their personal freedoms, and stereotype social groups” (p. 265). In some instances, particularly on mainstream news websites, uncivil comments are also likely to be deviant. Deviance, however, differs from incivility in that rather than being defined normatively, deviant commenting is defined socially.

Social deviance is defined as “acts, beliefs, and characteristics that violate major social norms and attract, or are likely to attract, condemnation, stigma, social isolation, censure, and/or punishment by relevant audiences” (Goode, 2015: 4). That is, deviant comments are defined by social interactions between comment authors and their audiences, including moderators and other users of a website. Deviant acts, or in the case of news comments, deviant comments, are those that draw condemnation and censure, in the instance of news comments, either by other users of the news site or by comment moderators. Again, in most instances of mainstream news websites, uncivil comments, personal attacks, “flaming” that is meant to belittle and silence others’ voices in commenting spaces, racists or sexist attacks, or other behavior that undermines democratic participation in online commenting spaces—would also draw condemnation as being deviant. Certainly, uncivil comments have been shown to draw strong reactions from audiences, for example, damaging newspapers’ reputations, discouraging engagement of other readers, and leading to polarization, negative intergroup emotions, less intellectual flexibility, and decreased open-mindedness (Anderson et al., 2014; Borah, 2014; Chen and Ng, 2017; Hwang et al., 2018).

It is possible that what is patently uncivil—for example, calling President Trump a vulgar name—may not necessarily be considered by a particular community—on a liberal political blog—as being deviant. But because this article addresses commenting on mainstream news organizations’ websites, incivility and deviance are practically interchangeable terms. That said, focusing on deviance is still useful in that it highlights the phenomenon—and solutions to the problem—as being inherently social.

Social control and comment moderation

Deviance and social control are central concepts of sociology, particularly in the subfield of criminology. Whereas deviance is those acts that are likely to draw social condemnation and censure, social control is those efforts of a community to enforce (or reinforce) conformity to the norms of society (Deflem, 2015). Deviance and social control are important community qualities. A lack of effective social control leads to social disorganization, marked by higher crime rates and delinquency, which in turn lead residents to withdraw from the community and become socially isolated, ultimately resulting in increased morbidity and mortality (Sampson et al., 2002; Woldoff, 2002). Morbidity and mortality—exhibited, in our case, by the criticisms that news commenting has faced and by those news organizations that have eliminated news comments altogether—can be seen as threatening communities of news commenters as a consequence of those communities’ failures to mobilize effective forms of social control.

A benefit of applying the concepts of deviance and social control to news commenting is that it labels abusive news comments for what they are, a controllable departure from community norms, rather than an inevitable feature of news comments per se. Social control theory also highlights that multiple, symbolic forms of social control coexist to help maintain social order. Social control has six types/levels: affirmative and sanctioning social control; imposed by informal and/or formal social actors (e.g. news commenters themselves vs professional moderators); either through direct or indirect actions (see Figure 1). The challenge of moderating news comments is not likely to be fixed by technological design (e.g. algorithms that automatically detect and delete abusive comments), or by journalists manually deleting comments that violate a site’s policies (a form of sanctioning, formal, direct social control). Rather, different forms of social control and different actors (i.e. professional journalists and news users) are often necessary, working in concert with one another, to maintain social order.

Formal social control is when those outside authorities (e.g. police) directly intervene in a community to enforce social norms (e.g. by arresting those who engage in deviant behaviors; Warner, 2007a). Likewise, professional moderators can turn off comments on controversial news articles, remove comments that do not adhere to community standards, or even ban individual commenters from the site. These actions are examples of direct social control, which involve an authority taking a specific action to stop a specific deviant act. The problem with this type of direct formal control, especially given the large number of news comments, is that it typically requires human moderators’ intervention and many news organizations do not have the financial and human capital for such resource-intensive interventions.¹ (Local police departments have similar resource constraints.)

Formal control can also be indirect. Consider, for example, elements of community design such as increased street lighting or beautified public spaces (Warner, 2007a). News organizations can also affect the design of news sections, for example, by requiring users to verify their identity by signing in to comment with a Facebook username (Curley, 2010). One study found that changes as minor as altering a button that invites users to “Respect” rather than “Like” can lessen partisan divisions that occur when users only positively respond to those comments that support one’s existing political beliefs (Stroud et al., 2017). However, despite these efforts of indirect social control, some individuals will inevitably continue to leave abusive comments despite these indirect efforts, requiring other forms of social control.

Most news organizations rely on formal and informal social control. For example, the efforts of professional journalists/moderators (sources of formal control) can be combined with user input, as users are encouraged to report abusive comments. When users report abusive comments, this is a form of sanctioning informal, indirect social control (Crawford and Gillespie, 2016; Goodman, 2013; Noci et al., 2012; Reich, 2011; Stroud et al., 2016). The reliance on social collaboration between those actors engaging in different forms of social control—in this case, between journalists and readers—is consistent with how social control operates in traditional communities (Bradford and Jackson, 2016; Ohmer and Beck, 2006; Sampson et al., 1997; Warner, 2007a, 2014). For example, if a police department does not have the cooperation of community members who help with informal social control—whether direct (e.g. confronting neighbors over their bad behavior) or indirect (e.g. calling the police to report trouble)—the police will struggle to enforce social order within a community. The same could be true for comment moderators.

Forms of informal social control—or what Naab, Kalch, and Meitz (2018) describe as “bystander intervention”—by news commenters include the ability to rate (e.g. thumbs up/thumbs down, or star ratings) or report news (e.g. flag abusive comments for comment moderators). Ratings are an example of direct informal social control. When a user rates a comment, they send a direct signal to the person who posted the comment, as well as to others in the community, about what types of comments meet the normative expectations of the community. Websites may also encourage commenters to report abusive comments to moderators, which would be a form of indirect, informal social control, since users are acting indirectly by reporting the issue to authorities.

While there is an extensive body of literature on deviant online comments (see, e.g. Coe et al., 2014; Ksiazek, 2015; Papacharissi, 2004; Prochazka et al., 2018; Santana, 2014), very few studies have focused on moderation of deviant comments, especially regarding examples of informal social control. Rather, previous studies have focused primarily on a formal social control perspective, as in the visibility of policies that express community norms (Kalch and Naab, 2017), the effect of different user verification and disclosure protocols (Cho and Kwon, 2015), or the role that the content of comments may have on experimental participants’ likelihood to report abusive comments (Kalch and Naab, 2017). This study deals primarily with forms of informal social control and a question that many sociologists have raised: What characteristics and factors encourage individuals to intervene to censure or attempt to stop deviant behavior in order to maintain informal social control in a community?

Predictors of informal social control in communities of news commenters

The traditional community effects literature primarily examines how neighborhood context affects types of social control and related outcomes (e.g. a decline in criminal and non-criminal forms of deviance, particularly among adolescents; Bradford and Jackson, 2016; Ohmer and Beck, 2006; Sampson et al., 1997; Warner, 2007b, 2014). The approach of this study necessarily differs from traditional community-sociological approaches, focusing instead on individual-level variables. An individual level of analysis is appropriate in an era of online, networked individualism, in which online news users are likely to have less involvement in and attachment to traditional forms of community (Rainie and Wellman, 2012). Thus, this study assumes that individual variables will be the primary predictors of whether online news users engage in indirect social control.

Some previous sociological studies of informal social control have taken a multilevel perspective, combining contextual and individual-level predictors of indirect social control behaviors. In these studies, individual-level variables that have been found to positively predict residents’ likelihood of engaging in informal social control include race/ethnicity (White), socioeconomic status (SES), age, and, depending on the nature of deviance being studied, gender (male) (Bradford and Jackson, 2016; Ohmer and Beck, 2006; Warner, 2007b, 2014). That is, White, male, wealthier, older residents—those who may at least perceive themselves as having greater social stature in their communities—are more likely to engage in direct informal social control.

Based on data of news commenters’ demographics (Stroud et al., 2016), we also expect that high-SES, White, males will be more likely to rate and report news comments, partly as a function of who is most likely to engage in news commenting. News commenting, though, does tend to be the domain of younger Internet users (Stroud et al., 2016); thus, we expect that related activities of rating and reporting news comments will be negatively associated with age.

Efficacy is also an important predictor of informal social control. While many sociological studies examine collective efficacy—the perception that neighbors can work together to confront a neighborhood issue (Bradford and Jackson, 2016; Ohmer and Beck, 2006; Warner, 2007b, 2014)—it is also possible to examine efficacy at the individual level (Ohmer and Beck, 2006). Individual self-efficacy is one’s self-confidence in one’s abilities to execute the actions necessary to achieve desired goals (Ohmer and Beck, 2006).

Finally, specifically related to indirect, informal, social control (i.e. calling the police or reporting abusive comments to moderators), trust in authorities (e.g. the police) and perceptions of moral decline have been found to be significant predictors of individuals’ willingness to invoke authorities (Bradford and Jackson, 2016). People are not going to invoke authorities and provide assistance—for example, by reporting abusive comments—if they do not trust those authorities. For our purposes, this means trusting the news media. Individuals who perceive that social and moral values are in decline and that established social order is threatened, will also be more likely to invoke authorities “to correct what they see to be a deteriorating state of affairs … to invoke police to not only deter, but punish those they see as a threat” (Bradford and Jackson, 2016: 4). Perceptions of moral decline and a desire to reimpose moral order are associated with the authoritarian personality. Authoritarian people want to impose strict community standards and favor imposing stronger official sanctions on those who violate those community standards (Unnever and Cullen, 2010). It is also anticipated that authoritarian personality and trust in authorities (i.e. trust in news media) will be positively associated with direct informal social control (i.e. reporting news comments to moderators).

Hypotheses

H1—the frequency with which online news users (1) rate and (2) report news comments will be positively predicted by users.

H1a—race/ethnicity (White, non-Hispanic).

H1b—gender (male).

H1c—educational attainment.

H1d—household income.

H2—users’ age will negatively predict the frequency with which online news users (1) rate and (2) report news comments.

H3—the self-efficacy of online news users’ news commenting will positively predict the frequency with which online news users (1) rate and (2) report news comments.

H4a—online news users’ trust in the news media will positively predict the frequency with which online news users (1) rate and (2) report news comments.

H4b—online news users’ trust in the news media will more strongly predict the frequency with which online news users report versus rate news comments.

H5a—online news users’ authoritarian personalities will positively predict the frequency with which online news users (1) rate and (2) report news comments.

H5b—online news users’ authoritarian personalities will more strongly predict the frequency with which online news users report versus rate news comments.

Methods

This study is based on a secondary analysis of data from an online survey of 632 US-based online news users who had either commented on or “lurked” (i.e. read, but had not participated in, news comments; see Kim, Lewis, & Watson, 2018). The sample was recruited using a panel administered by Qualtrics, a US-based survey company. Participants were recruited using non-probability purposive sampling to ensure the study captured individuals who comment on the news, which previous studies suggest is only about 14% of American adults (Stroud et al., 2016).

The survey was fielded between 15 and 24 September and again between 12 and 20 October 2016.

After consenting to participate in the study, respondents were asked a series of screening questions including whether in the past 2 months they had read online news or commented on online news. Our analysis focuses on respondents (N = 334) who indicated they had engaged in both of these activities. We presume these are the most active and influential members of communities of news commenters.

Dependent variables

It is possible that similar variables predict news commenting as predict rating and reporting comments (i.e. informal social control). Thus, in order to compare factors associated with the different behaviors, three separate models will be used to predict news commenting (Model 1) from rating (Model 2) and reporting (Model 3). In Models 2 and 3, news commenting is a control variable to understand how the independent variables predict rating and reporting separate from news commenting behaviors.

Rating/direct informal control

Respondents were asked on a seven-point scale (1 = never, 7 = every day) how often they “Rate (e.g. like, thumbs up, thumbs down) comments on news websites or news applications (apps)” (M = 4.18, SD = 2.25).

Reporting/indirect informal control

Respondents were asked on a seven-point scale (1 = never, 7 = every day) how often they “Report (e.g. flag as inappropriate comments on news websites or news applications (apps)” (M = 2.40, SD = 2.03).

Commenting

Respondents were asked on a seven-point scale (1 = never, 7 = every day) how often they post comments on “news websites or news applications (apps)” (M = 3.56, SD = 2.16).

Independent variables

Self-efficacy

Given the secondary analysis of this survey data, we were limited by existing measures to capture key concepts. The survey did not include any direct questions of individuals’ self-efficacy in rating or reporting comments. We selected three items that capture individuals’ self-efficacy regarding their broader contributions to news comment sections on a seven-point scale (strongly disagree = 1, strongly agree = 7): “My comments are valuable,” “My comments are informative,” “My comments are helpful.” A mean index was created from these items (α = .892, M = 5.52, SD = .99).

Trust in news media

Respondents were asked to focus on sources of news, that is, “the actual newsrooms that produce the news” separate from platforms (print, online or mobile). Then they were asked to respond to this question: “[Do you] trust news that comes from mainstream news media (e.g. print newspapers, television).” The seven-point scale ranged from “I do not trust the news from this source at all” (1) to “I completely trust news from this source” (7) (M = 4.50, SD = 1.74). This item seeks to capture the degree to which respondents trusted news from traditional, mainstream news organizations/brands (e.g. the New York Times or NBC News) that were traditionally either print or broadcast, whether users were actually accessing that information via print/broadcast or online.

Authoritarianism

Authoritarian personality was measured by four items that asked respondents to respond to pairs of qualities that children might have and indicate which of each pair was more desirable (respondents could also select “I’m not sure,” “Both are equally important,” and “Neither is very important”) (Sears et al., 1997). Here, an (a) indicates which response is coded as authoritarian. Those qualities were “Independence or Respect of elders (a),” “Obedience (a) or self-reliance,” “Curiosity or good manners (a),” and “Being considerate or well-behaved (a).” Authoritarian responses were dummy coded “1” and all other responses “0.” Items were added together to create an authoritarian index (M = .87, SD = 1.05, α = .71).

Demographics

Respondents were also asked their age (M = 52.07, SD = 16.09), gender (male = 49.1%), race/ethnicity (White, non-Hispanic = 87.4%), household income (M = 3.35, SD = 1.41; 1 = “Less than 14,999,” 6 = “$150,000 or more”) and educational attainment (M = 4.25, SD = 1.55) [1 = “Some high school,” 7 = “Advanced degree (PhD, JD, MD, DDS, etc.)”].

Results

As shown in Table 1, Model 1 predicted 30.3% (R² = .303) of the variance in users’ frequency of news commenting; Model 2 predicted 41.9% (R² = .419) of the variance in the frequency with users rated news comments; and Model 3 predicted 51.3% (R² = .513) of the variance in the frequency with which users reported deviant news comments. In Model 1, being male (ß = .110, p = .032), having a higher income (ß = .214, p < .001), feeling greater efficacy (ß = .238, p < .001), and being younger (ß = –.211, p < .001) all predicted the frequency of news commenting. In subsequent models, the frequency of commenting is used as a control. Unsurprisingly, in Models 2 and 3, respectively, the frequency of news commenting was the strongest predictor of the frequency of rating (ß = .564, p < .001) and reporting (ß = .428, p < .001) news comments. The remainder of the analysis examines those characteristics and attitudes that predict rating and reporting news comments independent of the simple frequency of commenting, using it as a control.

Table 1.

Multiple linear (OLS) regression predicting news commenting, rating, and reporting news comments.

	Model 1^a News commenting			Model 2^b Rating comments			Model 3^c Reporting comments
	B	Std. error	ß	B	Std. error	ß	B	Std. error	ß
Gender (male)	.468	.218	.110^*	−.193	.208	−.044	.400	.178	.097^*
Race/ethnicity (White, non-Hispanic)	.473	.341	.070	−.093	.324	−.013	.568	.277	.087^*
Education	.130	.077	.094	−.068	.073	−.048	.094	.062	.071
Income	.314	.082	.214^**	.151	.079	.100	.133	.068	.094
Age	−.028	.007	−.211^**	−.004	.007	−.026	−.023	.006	−.176^**
Efficacy	.512	.109	.238^**	.159	.107	.071	.179	.092	.086
Trust in news	.025	.064	.020	.104	.060	.084	.123	.052	.105^*
Authoritarianism	.206	.108	.100	.006	.103	.003	.187	.088	.093^*
News commenting				.583	.055	.564^**	.415	.047	.428^**

F(8, 303) = 17.057, p < .001, R² = .303.

F(9, 303) = 23.515, p < .001, R² = .419.

F(9, 303) = 34.363, p < .001, R² = .513.

p < .05.

p < .01.

H1 hypothesized that the frequency with which online news users (1) rate and (2) report news comments would be positively predicted by users’ demographics. H1a, which stated that rating and reporting comments would be positively predicted by race/ethnicity, was partially supported. White, non-Hispanic commenters were significantly more likely to report comments (ß = .087, p = .042), but race/ethnicity was not significantly associated with rating comments (ß = –.013, p = .774). That is, race/ethnicity is a unique predictor of the frequency with which news users report news comments.

H1b was that gender would also significantly predict both the frequency with which news users rated and reported news comments. Again, H1b was also partially supported. Gender significantly predicted reporting news comments (ß = .097, p = .025) but not rating them (ß = –.044, p = .353). It is also worth noting that gender significantly predicted news commenting (ß = .110, p < .05), which in turn predicts rating (ß = .544, p < .001) and reporting (ß = .428, p < .001) news comments.

H1c was that educational attainment would significantly predict both the frequency with which news users rate and report news comments. H1c was not supported. Educational attainment failed to significantly predict either rating (ß = –.048, p = .348) or reporting (ß = .071, p = .132) news comments.

H1d was that household income would significantly predict both the frequency with which news users rate and report news comments. H1d was not supported. Higher income individuals were more likely to comment on the news (ß = .214, p < .001), and those who are more likely to comment on the news are more likely to rate and report news comments. But income did not predict rating (ß = .100, p = .058) and reporting (ß = .094, p = .051) of news comments independent of the frequency of news commenting behaviors.

H2 was that age would negatively predict the frequency with which news users rated and reported news comments. H2 was partially supported. Younger online news users were significantly more likely to report (ß = –.176, p < .001) but not rate (ß = –.026, p = .611) news comments. Younger online news users were also more likely to comment on the news (ß = –.211, p < .001)

H3 predicted that self-efficacy of online news users’ news commenting would positively predict the frequency with which online news users (1) rate and (2) report news comments. H3 was not supported. Self-efficacy was the single strongest predictor of the frequency of commenting (ß = .238, p < .001), which in turn was the strongest predictor of rating (ß = .564, p < .001) and reporting (ß = .428, p < .001) news comments. However, self-efficacy did not independently predict either rating (ß = .071, p = .138) or reporting (ß = .086, = <.052) news comments.

H4a was that online news users’ trust in the news media would positively predict the frequency with which online news users (1) rate and (2) report news comments. H4a was partially supported. Trust in the news media significantly predicted reporting online news comments (ß = .105, p = .018), but it did not predict rating them (ß = .084, p = .084).

H4b posited that online news users’ trust in the news media would more strongly predict the frequency with which online news users report versus rate news comments. The mixed results from H4a—that trust in the media significantly predicted reporting online news comments (ß = .105, p = .018) but not rating online news comments (ß = .084, p = .084)—supports H4b.

H5a predicted that online news users’ authoritarian personalities would positively predict the frequency with which online news users (1) rate and (2) report news comments. H5a was also partially supported. Authoritarianism positively predicted reporting news comments (ß = .093, p = .036) but not rating news comments (ß = .003, p = .950).

H5b predicted that online news users’ authoritarian personalities would more strongly predict the frequency with which online news users report versus rate news comments. The results from H5a—that authoritarianism predicted reporting news comments (ß = .093, p = .036) but not rating news comments (ß = .003, p = .950)—supports H5b.

Discussion

News comment sections perhaps have been unfairly characterized as being off-topic at best and racist and misogynistic at worst. That said, deviant commenting is a problem that news organizations and online news users must contend with. Some organizations have decided to deal with the problem by shuttering their news comment sections altogether. Such a drastic solution, however, has its own social, political, and journalistic costs, including by limiting space for public discussion and community engagement. Rather than closing down news comment sections, it would be preferable if news organizations and news users were able to cooperate to improve social control within communities of news commenters.

No one single form of social control will solve what is a complex social issue of deviant news commenting. A key contribution of this article is to suggest a typology of social control within news commenting communities. Our approach broadens the focus of the literature on comment moderation, much of which has emphasized efforts to improve news comments through design (Zamith and Lewis, 2014)—for example, by requiring individuals to sign in with a Facebook username (Curley, 2010). This type of indirect formal social control is important, but other forms of social control likely will be necessary to solve what is primarily a social, rather than a technical, conundrum.

The six-quadrant social control typology presented in this article helps to conceptualize and categorize diverse social strategies used to moderate deviant comments. It also highlights that effective social control is likely to require a balance of multiple forms of direct and indirect social control as well as cooperation between the formal involvement of authorities (i.e. journalists and/or moderators) and the informal involvement of community members (i.e. online news users). There is also a balance between approaches that are affirmative (e.g. rewarding behavior that conforms to social norms) and sanctioning (e.g. punishing behaviors that violate norms).

This study also sought to tackle what has been an important question within the sociological literature: Who will step in to intervene to discourage deviant behavior and enforce social norms within a community? Before we address this question regarding communities of news commenters, we should acknowledge some of the limitations of our study. Our analysis is based on a single, cross-sectional analysis of a secondary dataset drawn from a purposive, non-representative survey.² Some of our variables may have also been defined differently in a non-secondary data research design: Perhaps it would have been more appropriate to measure trust in news organizations separate from news content, or more specifically measure trust in comment moderators. Therefore, our data and analysis should be treated as exploratory in nature and should not be presumed to necessarily represent all US news commenters.

Nevertheless, in addition to the theory-driven typology offered by this article, the data also raise some observations worthy of thought and follow-up research, particularly on important variables such as gender and race/ethnicity variables of interest to media sociologists. Reporting comments (i.e. informal social control) specifically was positively predicted by being male and White (non-Hispanic)—those who already have greater social advantages, including being more encouraged to speak up in public forums. These findings could be partly a function of our sample, to the extent that it overrepresented White, non-Hispanic males compared with their presence in the US population (though, not necessarily, in the community of news commenters). But these findings about gender and race/ethnicity should not be surprising, as reporting comments, as opposed to rating them, involves taking a stronger personal stand. And, in particular, it involves seeking a stronger form of sanctioning (one might say, policing) of others’ activities.

We can hope that these White, male, authoritarian users are wielding their privilege to stand up for and protect the rights of women and minorities, and some likely are. But it would be naïve to think that these were primary motivators for White, authoritarian males to report abusive comments. Non-White and female users also need to be empowered to report abusive comments. This could be encouraged, perhaps, through elements of design that are a part of indirect formal control, or through efforts of moderators to make sure that women are included and respected in these spaces—a form of formal, direct social control. (Though it is also possible that it is not an issue of encouragement and empowerment, but rather that experiences of being discriminated against have, sadly, desensitized non-White, female users to abusive comments, which is why they are less likely to report them.)

In addition to race/ethnicity, trust in the news media and authoritarianism also uniquely and positively predicted the frequency with which news users report deviant comments. Reporting news comments is a request for assistance from journalists/moderators to help police deviant comments. It makes sense that users would not seek journalists’ assistance if they did not fundamentally trust them. These findings expand upon the understanding of why trust in the news media is important. Future studies might further examine the role of media trust as a social relationship, where trust can be earned (and lost) in news commenting sections, as well as other participatory media spaces. For example, it is possible that there is a reciprocal relationship (cf. Lewis et al., 2014): Those higher in trust are more likely to report comments, and, in turn, journalists’ prompt, transparent response to commenters’ reporting of abusive comments could feed back into improving trust among online news users. (On the other hand, failure to react appropriately to reports of abusive comments may further erode trust in news media.)

Evaluating the role of authoritarianism in reporting news comments is more complex. On one hand, communities benefit from having individuals who take it upon themselves, whether based on their fundamental personality or otherwise, to take an active role in social control. On the other hand, professional moderators that receive reports of abusive comments must understand that authoritarian users are apt to want to impose an especially traditional notion of proper social order. Authoritarian online news users may bristle, for example, at impassioned condemnations in comment sections to instances of sexism and racism that criticize others for not “checking” their White, male privilege, or that argue for individuals’ rights to protest by kneeling during the national anthem. (This also gets to the point that many White, male users will not have the same interest in protecting minority rights.)

The double-edged nature of authoritarianism is why social control and comment moderation need to be a collaborative, social activity, engaging and balancing multiple forms of informal and formal social control. Authorities need to be able to step in when, for example, an individual or an idea is being inappropriately profiled, to protect minority rights. Certainly, the news media have an interest in being a forum for robust discussion of issues of public concern, which is in part why comment spaces were created and persist even in the face of obvious deviance. In these commenting spaces, the media also have a strong interest in protecting the right of individuals to express non-traditional, even unpopular viewpoints, which may be erroneously reported as violating community standards. Our data raise complicated issues, including along lines of gender and race. The complexity of the issue of controlling deviant behaviors, as well as the typology of complementary forms of social control, help to illustrate that this is not a problem with a technical solution. It cannot simply be left to an algorithm to identify and remove abusive comments. Enforcing social norms in news commenting spaces necessarily requires a more complex, nuanced social process.

Footnotes

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

ORCID iDs

Brendan R Watson

Seth C Lewis

Author biographies

Brendan R Watson (PhD, University of North Carolina at Chapel Hill), is an assistant professor of journalism innovations at Michigan State University. His research interests include media sociology; community information needs and civic engagement; how emerging digital and participatory communication technologies are or are not changing the availability of, and the public’s interaction with, public affairs news and information; and quantitative research methods, particularly content analysis.

Zhao Peng (MA, Michigan State University), is a doctoral student in the School of Journalism at Michigan State University. Her research interest are in media psychology, social media effects and gender studies. She is doing research about how to change people’s attitude toward female politicians with manipulated media messages and observing people’s communication behaviors and norms on social media platforms.

Seth C Lewis (PhD, University of Texas), is the founding holder of the Shirley Papé Chair in Emerging Media in the School of Journalism and Communication at the University of Oregon, an affiliate fellow of the Information Society Project at Yale Law School, and a 2019-2020 visiting fellow at the Reuters Institute for the Study of Journalism at the University of Oxford. His present research focuses on algorithms and automation in journalism, the interplay of journalists and audiences, and questions of news, knowledge, and trust in an era of misinformation.

References

Anderson

Brossard

Scheufele

et al. (2014) The “nasty effect”: online incivility and risk perceptions of emerging technologies. Journal of Computer-Mediated Communication 19: 373–387.

Borah

(2014) Does it matter where you read the news story? Interaction of incivility and news frames in the political blogosphere. Communication Research 41(6): 809–827.

Bradford

Jackson

(2016) Cooperating with the police as an act of social control: trust and neighbourhood concerns as predictors of public assistance. Nordisk Politiforskning 3: 111–131.

Breed

(1955) Social control in the newsroom: a functional analysis. Social Forces 33: 326–335.

Cappella

Price

Nir

(2002) Argument repertoire as a reliable and valid measure of opinion quality: electronic dialogue during campaign 2000. Political Communication 19: 73–93.

Chen

YMM

(2017) Nasty online comments anger you more than me, but nice ones make me as happy as you. Computers in Human Behavior 71: 181–188.

Cho

Kwon

(2015) The impacts of identity verification and disclosure of social cues on flaming in online user comments. Computers in Human Behavior 51: 363–372.

Coe

Kenski

Rains

(2014) Online and uncivil? Patterns and determinants of incivility in newspaper website comments. Journal of Communication 64: 658–679.

Coser

(1956) The Functions of Social Conflict. Glencoe, IL: Free Press.

10.

Crawford

Gillespie

(2016) What is a flag for? Social media reporting tools and the vocabulary of complaint. New Media & Society 18: 410–428.

11.

Curley

(2010) Sun changes story commenting policy to foster responsible dialogue. Las Vegas Sun, 21 September. Available at: http://www.lasvegassun.com/news/2010/sep/21/story-commenting-policy/

12.

Deflem

(2015) Deviance and social control. In: Goode

(ed.) The Handbook of Deviance. Somerset, NJ: John Wiley & Sons, pp. 30–44.

13.

Editor & Publisher (2007) NYTimes.com outlines reader comment moderation. Editor & Publisher, 14 November. Available at: http://www.editorandpublisher.com/news/nytimes-com-outlines-reader-comment-moderation/

14.

Erjavec

(2014) Readers of online news comments: why do they read hate speech comments? Annales Histoire, Sciences Sociales 24: 451–462.

15.

Etim

(2017) The Times sharply increases articles open for comments, using Google’s technology. The New York Times, 13 June. Available at: https://www.nytimes.com/2017/06/13/insider/have-a-comment-leave-a-comment.html?_r=1

16.

Finley

(2015) A brief history of the end of the comments. Wired, 8 October. Available at: https://www.wired.com/2015/10/brief-history-of-the-demise-of-the-comments-timeline/

17.

Freelon

(2010) Analyzing online political discussion using three models of democratic communication. New Media & Society 12: 1172–1190.

18.

Goode

(2015) The sociology of deviance: an introduction. In: Goode

(ed.) The Handbook of Deviance. Somerset, NJ: John Wiley & Sons, pp. 3–29.

19.

Goodman

(2013) Online comment moderation: emerging best practices. Available at: http://www.wan-ifra.org/reports/2013/10/04/online-comment-moderation-emerging-best-practices

20.

Hwang

Kim

(2018) Influence of discussion incivility on deliberation: an examination of the mediating role of moral indignation. Communication Research 45: 213–240.

21.

Jensen

(2016) NPR website to get rid of comments. NPR, 17 August. Available at: http://www.npr.org/sections/ombudsman/2016/08/17/489516952/npr-website-to-get-rid-of-comments

22.

Kim

Lewis

Watson

(2018) The imagined audience for and perceived quality of news comments: exploring the perceptions of commenters on news sites and on Facebook. Social Media + Society 4: 1–12.

23.

Ksiazek

(2015) Civil interactivity: how news organizations’ commenting policies explain civility and hostility in user comments. Journal of Broadcasting & Electronic Media 59: 556–573.

24.

Lewis

Holton

Coddington

(2014) Reciprocal journalism: a concept of mutual exchange between journalists and audiences. Journalism Practice 8(2): 229–241.

25.

Long

(2017) Meet the New York Times’s super-commenters. The New York Times, 5 November. Available at: https://www.nytimes.com/2017/11/25/insider/new-york-times-top-commenters-profile.html

26.

McMillan

Chavis

(1986) Sense of community: a definition and theory. Journal of Community Psychology 14(1): 6–23.

27.

Moor

Heuvelman

Verleur

(2010) Flaming on YouTube. Computers in Human Behavior 26(6): 1536–1546.

28.

Mutz

(2008) Is deliberative democracy a falsifiable theory? Annual Review of Political Science 11: 521–538.

29.

Kalch

Naab

(2017) Replying, disliking, flagging: how users engage with uncivil and impolite comments on news sites. Studies in Communication and Media 4: 395–419. DOI: 10.5771/2192-4007-2017-4-395

30.

Naab

Kalch

Meitz

(2018) Flagging uncivil user comments: effects of intervention information, type of victim, and response comments on bystander behavior. New Media & Society 20: 777–795.

31.

Noci

Domingo

Masip

et al. (2012) Comments in news, democracy booster or journalistic nightmare: assessing the quality and dynamics of citizen debates in Catalan online newspapers. International Symposium on Online Journalism 2(1): 46–64.

32.

Ohmer

Beck

(2006) Citizen participation in neighborhood organizations in poor communities and its relationship to neighborhood and organizational collective efficacy. Journal of Sociology & Social Welfare 33: 179–202.

33.

Papacharissi

(2004) Democracy online: civility, politeness, and the democratic potential of online political discussion groups. New Media & Society 6(2): 259–283.

34.

Price

(1992) Public Opinion. Newbury Park, CA: Sage Publications.

35.

Prochazka

Weber

Schweiger

(2018) Effects of civility and reasoning in user comments on perceived journalistic quality. Journalism Studies 19(1): 62–78.

36.

Rainie

Wellman

(2012) Networked: The New Social Operating System. Cambridge, MA: The MIT Press.

37.

Reich

(2011) User comments. In: Singer

Hermida

Domingo

et al. (eds) Participatory Journalism: Guarding Open Gates at Online Newspapers. Somerset, NJ: John Wiley & Sons, pp. 96–117.

38.

Robinson

(2015) Redrawing borders from within: commenting on news stories as boundary work. In: Carlson

Lewis

(eds) Boundaries of Journalism: Professionalism, Practices and Participation. New York: Routledge, pp. 152–168.

39.

Sampson

Morenoff

Gannon-Rowley

(2002) Assessing “neighborhood effects”: social processes and new directions in research. Annual Review of Sociology 28(1): 443–478.

40.

Sampson

Raudenbush

Earls

(1997) Neighborhoods and violent crime: a multilevel study of collective efficacy. Science 277(5328): 918–924.

41.

Santana

(2011) Online readers’ comments represent new opinion pipeline. Newspaper Research Journal 32(3): 66–81.

42.

Santana

(2014) Virtuous or vitriolic: the effect of anonymity on civility in online newspaper reader comment boards. Journalism Practice 8(1): 18–33.

43.

Santana

(2016) Controlling the conversation: the availability of commenting forums in online newspapers. Journalism Studies 17(2): 141–158.

44.

Sears

Van Laar

Carrillo

et al. (1997) Is it really racism? The origins of white American opposition to race-targeted polices. Public Opinion Quarterly 61: 16–53.

45.

Simmel

(1955) Conflict and the Web of Group Affiliations (trans. Wolff

). Glencoe, IL: Free Press.

46.

Singer

Domingo

Heinonen

et al. (2011) Participatory Journalism: Guarding Open Gates at Online Newspapers. Somerset, NJ: John Wiley & Sons.

47.

Steenbergen

Bächtiger

Spörndli

et al. (2003) Measuring political deliberation: a discourse quality index. Comparative European Politics 1: 21–48.

48.

Stromer-Galley

(2007) Measuring deliberation’s content: a coding scheme. Journal of Public Deliberation 3: 1–37.

49.

Stroud

Muddiman

Scacco

(2017) Like, recommend, or respect? Altering political behavior in news comment sections. New Media & Society 19(11): 1727–1743.

50.

Stroud

Van Duyn

Peacock

(2016) News commenters and news comment readers. Report, Center for Media Engagement, The University of Texas at Austin, Austin, TX, 14 March. Available at: https://mediaengagement.org/research/survey-of-commenters-and-comment-readers/

51.

Uneven

Cullen

(2010) The social sources of Americans’ punitiveness: a test of three competing models. Criminology 48: 99–129.

52.

Warner

(2007a) Directly intervene or call the authorities? A study of forms of neighborhood social control within a social disorganization framework. Criminology 45: 99–129.

53.

Warner

(2007b) Robberies with guns: neighborhood factors and the nature of crime. Journal of Criminal Justice 35: 39–50.

54.

Warner

(2014) Neighborhood factors related to the likelihood of successful informal social control efforts. Journal of Criminal Justice 42: 421–430.

55.

Watson

Myers

Kim

et al. (2017) Sounding off on news commenting: what value does it have for civic and political participation? Paper presented at the international communication association conference, San Diego, CA.

56.

Watson

Zamith

Savannah

et al. (2018) Are demographics adequate controls for cell-phone-only coverage bias in mass communication research? Journalism & Mass Communication Quarterly 92: 723–743.

57.

Woldoff

(2002) The effects of local stressors on neighborhood attachment. Social Forces 81: 87–116.

58.

World Association of Newspapers and News Publishers (2016) The 2016 global report on online commenting. Available at: http://www.wan-ifra.org/reports/2016/10/06/the-2016-global-report-on-online-commenting

59.

Zamith

Lewis

(2014) From public spaces to public sphere: rethinking systems for reader comments on online news sites. Digital Journalism 2: 558–574.