User Experiences with Editorial Control in Online Newspaper Comment Fields

Abstract

This article investigates user experiences with editorial control in online newspaper comment fields following the public backlash against online comments after the 2011 terror attacks in Norway. We analyze data from a survey of online news consumers focusing on experiences and attitudes towards editorial control set against a spectrum between “interventionist” and “noninterventionist” positions. Results indicate that interventionist respondents rate the quality of online comments as poor, whereas noninterventionist respondents have most often experienced being the target of editorial control measures and feel that editorial control has intensified after the terror attacks. We conclude that newspapers should pay attention to the different needs of participants when devising strategies for editorial control. Media professionals should also consider changes to increase the transparency of moderation practices.

KEYWORDS:

• comment fields
• communication design
• editorial control
• journalism
• moderation
• online comments
• user experience

Introduction

Online comment fields on mass media news websites have long been subject to critical scrutiny from media scholars, journalists and society at large. Following the digital turn in the 1990s, the media industry actively pursued participatory formats that would enhance user involvement and public debate. Although this trend was inspired by the possible democratic and economic gains from facilitating such activities, media professionals have also expressed much skepticism towards online comments, and a sizable scholarship has explored how editors have struggled with the administration of said functionality (Ihlebæk and Krumsvik 2015; Singer et al. 2011).

While comment fields have been frequently investigated from the point of view of media institutions, this article aims to explore the opinions and experiences of editorial control among the users. For comment fields to serve their purpose as sites for democratic debate, it is important that editorial control is exercised in a manner that is perceived as fair and balanced by the users. Earlier research has indicated that editorial control is often practiced through tacit expectations and unwritten rules and may be hard for participants to understand (Ihlebæk and Ytreberg 2009; Springer, Engelmann, and Pfaffinger 2015). Since comment fields are lowering the barriers to participate in public debate, it is reasonable to assume that some users may have a harder time understanding and complying with these informal rules than others. Furthermore, given the widespread perception that online newspaper comment fields are dominated by harassment directed at particular groups—e.g. women and minorities (Gardiner et al. 2016; Johansson 2013)—it could be expected that those belonging to such groups may have different views on the need for editorial control than those who do not. Knowledge about such differences may be of help to newspapers when shaping their strategies for editorial control and for the facilitation of diverse and inclusive debates.

We address these issues with findings from a survey among contributors to online newspaper comments fields, conducted in the middle of a public backlash against such online comments following the 2011 terror attacks in Norway. The devastating killing spree, in which 77 people were killed and several hundred wounded, was conducted by a terrorist who was inspired by right-wing ideologies and a conspiracy theory known as “Eurabia” (Strømmen 2011). Media reports about the perpetrator’s online activities led to much public criticism of online comments fields, which were seen as sites of a destructive debate culture. Previous studies (Figenschou and Thorbjørnsrud 2016; Ihlebæk, Løvlie, and Mainsah 2013) have illuminated how the terrorist attacks led to significant adjustments in the ways that editorial control with online comments was carried out in the newspapers.

While the situation described above presents a case of unusually strong public controversy about online comments, the arguments brought forward are mostly centered on points of criticism that have been debated in many western countries since online comments became common in the early 2000s. More recently, current political events have led to debates about “echo chambers” (Sunstein 2007) and “filter bubbles” (Pariser 2011) to rise high on the public agenda in both the US and many European countries. While recent debates about these phenomena have tended to focus on the role of social media, many have also questioned the contributions of comment fields on news media websites. Political institutions in several European countries have proposed control measures intended to combat the phenomenon increasingly known as “fake news” (Jackson 2017; Morozov 2017). With these recent developments in mind, we argue that it is vital to gain more knowledge about the ways that editorial control measures are experienced by those who choose to contribute such online comments. Relating back to our specific case, the post-terror backlash against online commenting in Norway after July 22, 2011 may be seen as an amplification of a long-standing professional and academic debate about online comments. It thus presents an interesting case where disagreements regarding online comments were put high enough on the public agenda to ensure that most users were likely to have an awareness of the controversies. With this in mind, the survey reported on in the paper at hand offers a particularly interesting opportunity to study ordinary users’ opinions and experiences of editorial control.

Using data from the survey, we investigate the following two exploratory research questions:

RQ1: What variables influence users’ opinions and experiences of editorial control with online comment fields?

RQ2: What variables influence users’ perception of changes in editorial control after the 2011 terror attacks?

RQ1 addresses not only the users’ attitudes towards editorial control, but also their perceptions of how control is exercised, as well as the effects of the control on their own opportunities to speak their minds freely. RQ2 addresses whether or not the users have experienced changes in the editorial control after the terrorist attacks, and if so, how those changes have affected them.

Given that our survey took place over a year after the terror attacks, it should be noted that we are investigating the respondents’ perception of changes having taken place, rather than the reality of such changes. We argue that whether or not the respondents recall changes accurately, their feeling that a change has taken place is important in and of itself, as this perception might contribute to a general feeling of skepticism and suspicion towards editorial control. This will be elaborated further later on.

In the study, we follow industry parlance in using “comment fields” as a general term which refers both to “below the line” comment fields placed under a journalistic article, as well as discussion forums where users can start their own threads and which are run as separate websites or subsections of a newspaper’s website (cf. Graham and Wright 2015).

Ambivalence Towards Comment Fields

User participation through online comment fields has long been the subject of controversies, and there has been disagreement in the scholarly community as well as in the media about the democratic potential and value of services like these (e.g. Graham and Witschge 2003; Janssen and Kies 2005; Jönsson and Örnebring 2011; Papacharissi 2004; Ruiz et al. 2011; Toepfl and Piwoni 2015; Wright and Street 2007). Scholars have pointed towards several challenges with online discussion formats like the rise of echo chambers and polarization of opinions (Sunstein 2007), “flaming” (e.g. Hutchens, Cicchirillo, and Hmielowski 2015; Lee 2005; Santana 2014), “trolling” (e.g. Binns 2012; Coleman 2012; Hardaker 2010) and harassment of women (e.g. Biber et al. 2002; Herring et al. 2002; Jane 2014).

The problematic aspects of online discussion formats are also reflected among media professionals. Studies have shown that media professionals have been reluctant to fully engage in dialogue with the users (e.g. Domingo 2008; Hermida and Thurman 2008; Jönsson and Örnebring 2011; Martin 2015; Singer et al. 2011). A relatively widespread consensus has nevertheless developed among editors and journalists that newspapers should enable user participation through comment fields (Domingo, Quandt, and Heinonen 2008; Ihlebæk and Krumsvik 2015; Weber 2014). For example, Reich (2011) argues that a likely explanation for this consensus is that online comment fields do not challenge the journalistic authority in the same way as other forms of audience participation—such as public or participatory journalism. Furthermore, participatory strategies have been defended on the basis of the democratic responsibility of the news media to facilitate the public debate, as well as the economic potential of user involvement on new platforms (Ihlebæk and Krumsvik 2015). In a Swedish study, Bergström and Wadbring (2015) investigated attitudes towards online comments in the general public as well as among journalists. They found that a large proportion of the general public had no opinion on reader comments, but that those who did were more positive than journalists towards such comments. The public also tended to agree that the journalists should join the commentary threads, whereas the journalists were more inclined than the general public to support “censoring” of abusive comments. Similarly, a study of two British local newspaper websites by Canter (2013) found that journalists were conflicted in their opinions about the value of online comments, in particular with regard to the perceived risk of brand damage. Conversely, Graham and Wright (2015) found that journalists at the British newspaper The Guardian valued online comment fields quite positively. The result from this latter study might seem a bit odd, given that the bulk of research seems to indicate a certain degree of skepticism towards commenting among media professionals. It might be that the empirical material—“eightyfive articles that focused on the UN Climate Change Summit” (Graham and Wright 2015, 317)—had an influence on this more positive view. Our current research focus, dealing with these issues in post-terror Norway, could thus be seen as providing a contrasting case.

Turning our attention to the contributors of online comments, several studies have attempted to uncover who participates in online commenting and why (e.g. Albrecht 2006; Baek, Wojcieszak, and Delli Carpini 2012; Mitchelstein 2011). Mitchelstein’s (2011, 2020) study of commentators in Argentinian newspapers found that commenting was linked to self-expression needs and the desire to voice their opinions. In a study conducted by Springer, Engelmann, and Pfaffinger (2015, 812), where they explore the motivations and gratifications amongst German commenters, lurkers and non-users, they find that commenters are to a large degree motivated by social factors like taking part in journalism, as well as discussing with other users, however they do not “obtain cognitive gratification”. Furthermore, readers and lurkers are driven by the entertaining elements of reading comments, including discussions of low quality, while non-users consider it to be “a waste of time” (812). They argue that to strengthen public engagement, news organizations should strive for transparency in privacy policies as well as in moderation practices, as well as encourage engagement from journalists. However, a study by Ihlebæk and Larsson (2016), exploring social media guidelines in news organizations, indicates that motivating journalists to interact with users are not highly prioritized in Norwegian news organizations and that utilizing distribution in social media is higher up on the agenda. Krumsvik’s (2017) study with media leaders supports this claim. He argues that economic imperatives have become more important as newspapers increasingly view users as distributors and not co-producers.

From a gender perspective, studies also indicate that a larger proportion of those who contributed to newspaper comment fields tended to be males and participants with lower education than among those who engaged in discussion on those topics on social media or in offline settings (Enjolras et al. 2013). Women were particularly absent in online discussions about immigration and politics. The authors of the study argue that these gender patterns mirror the skewed gender balance found in other public debate formats, like newspaper op-ed pieces or broadcast debates. Based on their findings, Enjolras et al. warn that online forums might increase these types of differences rather than decrease them. In a similar vein, Nussbaum (2010) argues that the reluctance of many women to participate in online forums is linked to the risk of experiencing harassment. Similarly, other studies have shown that while men experience threats more often than women, women are more prone to gendered and sexual responses, and tend to limit their own utterances to a larger degree than men (e.g. Enjolras and Steen-Johnsen 2014; Pew Research Center 2014; Staksrud et al. 2014). Based on the different consequences men and women might experience when participating in the public debate, it is interesting to see whether or not this may influence their views on editorial control.

Interventionist and Noninterventionist Editorial Strategies

How to best facilitate and utilize user comment fields has proved to be a difficult task for news organizations. The dilemma is based on the (sometimes) paradoxical motives of enhancing both quantitative and qualitative dimensions of the debate. Several studies have pinpointed how editorial measures like identification, moderation practices and issue controversy influence the quality of online debates (Berg 2016; Ihlebæk and Krumsvik 2015; Jensen 2003; Nagar 2011; Singer et al. 2011; Trénel 2009; Wright and Street 2007). While some studies show that anonymity might have a negative impact on civility (Coleman and Moss 2012; Santana 2014), others indicate that the topic discussed also plays an important part (Berg 2016). The ideals for online debate differ within particular traditions. While some justify a liberal stance on freedom of speech and emphasize the importance of arguments (civil or uncivil) being met with counter-arguments in the public, others defend the need for applying deliberative ideals as a reference point (see, for instance, Kies 2010; Stromer-Galley 2007). Furthermore, news organizations operate within particular legal and ethical contexts that influence the level of editorial control (Ruiz et al. 2011; Woods 2012). In Norway, editors can be held accountable for utterances in the publication they are responsible for, both through the law and through the self-regulatory ethical guidelines of the press as is common in the Northern European Democratic-Corporatist media system (Hallin and Mancini 2004).

Building on earlier work by Jensen (2003), we have previously suggested a model that describes different approaches to editorial control in online newspaper comment fields as a continuum between two poles: interventionist, indicating a high level of editorial control, and noninterventionist, referring to little or no regulatory measures (Ihlebæk, Løvlie, and Mainsah 2013) (see Table 1). We have used this model to analyze editorial control strategies from the point of view of newspaper editors and moderators. In the paper at hand, we will use the same model to explore users’ attitudes towards editorial control.

TABLE 1 Attitudes towards editorial control with comment fields

Interventionist strategies	Noninterventionist strategies
Pre-control	Post-control
Identification required	Anonymity allowed
Active moderators	Passive (or absent) moderators
Topics restricted	Users decide topics
Time restrictions	No time restrictions

A noninterventionist approach implies that the editors impose as little editorial control as possible. Employed in an online system for commenting, such an approach may entail that participants are allowed to be anonymous (often, by using pseudonyms), that contributions are published without any prior approval by a moderator, that moderators remain passive and only intervene if absolutely necessary, that users have a high degree of freedom to influence the topic for discussion, and that there are no time restrictions on debate (e.g. keeping debates open when no editors are available to potentially intervene). Nevertheless, most noninterventionist approaches do not entail absolute absence of control, given the legal and ethical frameworks that guide journalistic activity in most media systems. Conversely, an interventionist approach could entail the use of active moderators that approve all contributions before publication (pre-moderation), requiring users to register and use their real names online, limitations on the possibility for the user to decide or influence the topics for discussion, as well as time restrictions for participation.

For newspapers, the choice of editorial principles depends on several democratic, economic, legal and ethical factors (Singer et al. 2011). From a democratic point of view, a noninterventionist approach can be justified on the basis of utilitarianism and the ideal of a free marketplace of ideas. In this view, less control is seen as a way to lower the threshold for participation, thereby strengthening the freedom of debate. Furthermore, a noninterventionist approach can be defended on economic grounds—less control requires fewer editorial resources and might lead to more website visitors. However, a noninterventionist approach also increases the risk that users may cross legal and ethical lines, potentially violating norms against harassment, threats or other forms of unsuitable behavior that could potentially harm the brand. Aside from the potential individual harm caused by such comments, the lack of control may also lead to a rough debate climate causing the quality of the debate to suffer, potentially discouraging some readers from participating. An interventionist approach can be employed strategically to enhance the deliberative quality of the debate, and to ensure that both legal and editorial responsibilities are upheld. However, in order to impose efficient control mechanisms, resources such as technology and manpower are needed. Furthermore, journalists tend to be dissatisfied about having to take active roles in the comment fields (e.g. Bergström and Wadbring 2015; Ihlebæk and Krumsvik 2015).

More often than not, newspapers in Norway and beyond have tended towards the noninterventionist pole of the spectrum, allowing post-moderation and anonymous participation while at the same time implementing some sort of moderation—a task often fulfilled by journalists or employed moderators (Ihlebæk and Krumsvik 2015; Singer et al. 2011). Such a laissez-faire attitude towards online commenting has been motivated by economic interests as well as by arguments regarding democracy and free expression. As we shall see, the 2011 terror attacks in Norway intensified debates about editorial control with online comment fields, leading to several changes in the administration of user comments in Norwegian newspapers.

Shifting Editorial Strategies After the Terror Attacks

The terror attacks of July 22, 2011 sent shockwaves throughout Norwegian society on many levels. Several studies have investigated how the news media responded to the attacks and the public debate that followed (e.g. Andenæs 2012; Backholm and Idås 2015; Figenschou and Beyer 2014; Hervik and Meret 2013; Jupskås 2013; Kammer 2013; Lund and Olsson 2016). Of particular interest here is the high attention given to online comment fields in the time after the attacks, by scholars as well as media, politicians and other public figures.

In light of reports about the terrorist’s activities in various online forums, many criticized what they saw as a destructive culture in online debates—as well as too lax editorial control in online media (e.g. Eide, Kjølstad, and Naper 2013; Skogerbø 2013). Some even suggested that this online culture might be to blame for the attacks (e.g. Brandtzæg 2011; Øgrim 2011). Others expressed concerns that online comment fields might provide platforms for extremists to develop and amplify radicalized views together with the like-minded (Strømmen 2011). Although the main sources of inspiration cited by the terrorist were radical anti-Islamic and anti-immigration forums, the public criticism was also directed towards the comments fields of mainstream online newspapers.

In order to investigate whether newspapers changed their editorial strategies regarding online comments in light of these controversies, an interview study was conducted with respondents from the four Oslo-based newspapers shown in Table 2, nearly one year after the attacks (Ihlebæk, Løvlie, and Mainsah 2013).

TABLE 2 Newspapers in the study

Website	Average daily readers (2012)	Political orientation
Verdens Gang: vg.no, vgd.no	1,849,000	Center-right
Dagbladet: db.no	1,213,000	Center-left
Aftenposten: ap.no	745,000	Center-right
Vårt Land: vl.no, verdidebatt.no	29,000	Christian/Center-right

Source for readership numbers: TNS Gallup and http://www.medienorge.uib.no/statistikk/medium/avis/253.

Three of the newspapers (VG, Dagbladet and Aftenposten) are the three largest online newspapers in Norway by readership, whereas the fourth (Vårt Land) is a niche newspaper based on Christian values, which represented an interesting edge case as they had taken an unusually interventionist approach to online comment fields two years before the terror attacks. Results indicated that those newspapers that had the most noninterventionist approach to editorial control implemented stricter measures after the attacks. Within a short time after the attacks, the comment fields were entirely or partially closed on many Norwegian news websites, including the three largest newspapers mentioned above. According to the informants, the main reason for this measure was a feeling of lack of oversight and control among the editors in a situation of extreme uncertainty, not just regarding editorial control but the physical safety of staff and citizens. (The main offices of two of the newspapers were within the blast zone of the bomb explosion and were severely damaged; the other two newspapers also evacuated staff due to reports about a second bomb.) Many of these short-term measures were reverted after some weeks. However, one year after the attacks it was clear that all the three largest newspapers had implemented some lasting changes to strengthen their control with online comment fields, such as demanding user identification (VG), closing down an open discussion forum (Aftenposten), as well as strengthening their moderation services and restricting debate on particularly sensitive topics.

It is important to note that all of the newspapers stated that they had not changed the rules for what was allowed to be written in comment fields, nor the policies concerning when and how to intervene. However, based on the other changes mentioned above, the study concluded that there was a shift towards more interventionist strategies for editorial control in the leading Norwegian newspapers in the first year after the attacks. In the following years, this tendency towards increased intervention appears to have continued: several more leading newspapers (including Aftenposten) now demand identification, and Dagbladet has experimented with more active moderators, before deciding in 2016 to close their online comment fields in order to focus their efforts on social media, in particular Facebook (Ramnefjell 2016). Some of these tendencies have also been documented by Figenschou and Thorbjørnsrud (2016).

Method

Based on the research reviewed above, an online survey was conducted among participants in the comment fields of the four newspapers in Table 2. The participants were invited through a link from each newspaper to the online survey system Questback. The survey featured 22 questions and was open from September, 17 to October 3, 2013. During this time period, 3470 answers were received. In the present article we use answers from some questions in the survey, seen in Table 3, as independent variables in a regression analysis in order to explore the influence of these variables on the respondents’ opinions of and experiences with editorial control (formulated as dependent variables; see Table 4). The survey featured a series of questions that are not included here and that dealt mainly with user experiences of repercussions from writing online newspaper comments. Results from these questions have been presented elsewhere (Løvlie and Ihlebæk 2016). Questions and answers from the survey have been translated from Norwegian by the authors.

TABLE 3 Summary of the independent variables

Items included	Measurement
Age	1 = “Under 18” (3%) 2 = “18–29” (37%) 3 = “30–44” (36%) 4 = “45–60” (17%) 5 = “Over 60” (7%) (N = 3460)
Gender	1 = “Female” (22%) 2 = “Male” (78%) (N = 3413)
Education	1 = “Elementary school” (4%) 2 = “High school” (27%) 3 = “Up to four years of university” (37%) 4 = “More than four years of university” (32%) (N = 3450)
Frequency: “How often do you participate in online discussion, writing comments or posts, or making other contributions?”	1 = “Never” (18%) 2 = “Sometimes, but less than once a week” (45%) 3 = “Once or several times per week” (28%) 4 = “Once or several times per day” (8%) [Discarded: “Don’t know” (1%)] (N = 3452)
Quality: “How do you feel that the online debates work on the websites where you participate, judged against your own opinion about what constitutes a good or a poor debate?”	1 = “Very poorly” (14%) 2 = “Mostly poorly” (31%) 3 = “Neither good nor bad” (36%) 4 = “Mostly well” (18%) 5 = “Very well” (2%) (N = 3329)
Experience of moderation: “Have you ever noticed that one of your own comments has been removed or edited by the moderators?”	1 = “Never” (53%) 2 = “Once or a few times” (31%) or “Often” (8%) [Discarded: “I don’t know” (7%)] (N = 3290)
Change of presence: “Do you think that moderator and journalist presence in online comments has changed since July 22, 2011?”	1 = “They have become much less active” (1%) 2 = “They have become somewhat less active” (3%) 3 = “I haven’t noticed any change” (56%) 4 = “They have become somewhat more active” (29%) 5 = “They have become much more active” (11%) (N = 3383)
Change of freedom: “Do you feel that your opportunities to speak your mind freely in online comments have become better or worse after July 22, 2011?”	1 = “Much worse” (26%) 2 = “Somewhat worse” (22%) 3 = “The same” (47%) 4 = “Somewhat better” (4%) 5 = “Much better” (1%) (N = 3365)
Website: “In which online newspapers/online media have you participated with online comments?” (Multiple options allowed. In the regression analysis, this is recoded as a set of binary variables: 0 = Respondent does not debate on this site, 1 = Respondent debates on this site)	VG: 57% Dagbladet: 70% Aftenposten: 51% NRK website: 27% TV2 website: 19% Nettavisen: 20% Vårt Land (verdidebatt.no): 5% Dagsavisen (nyemeninger.no): 6% Hegnar Online: 7% Document.no: 5% Others: 33% (N = 2938)

TABLE 4 Summary of the dependent variables

Items included	Measurement
Rules: “In your opinion, are the limits for what is allowed to be written in online comments on the sites where you yourself participate appropriate, or too strict, or too allowing?”	1 = “Far too allowing” (8%) 2 = “A little too allowing” (19%) 3 = “Just appropriate” (29%) 4 = “A little too strict” (24%) 5 = “Far too strict” (21%) (N = 3355)
Presence: “Do you want the moderators and/or journalists to be present and active in the comments section, in order to answer questions and such?”	1 = “Yes” (74%) 2 = “No” (15%) [Discarded: “Don’t know” (11%)] (N = 3416)
Anonymity: “Do you usually comment using your real name, or anonymously/with a made-up name (pseudonym)?”	1 = “Real name” (25%) 2 = “Anonymously/made-up name (pseudonym)” (43%) [Discarded: “Both” (29%) and “Don’t know” (4%)] (N = 3086)

We should point out a few circumstances that may have had an effect on the results from our survey. Our sample is self-selected, so we cannot draw conclusions about the general population. Similarly, our respondent group is characterized by an overrepresentation of males, respondents between the ages of 18–44 and of university graduates when compared to the Norwegian population at large (cf. population statistics from Statistics Norway, ssb.no). As such, while it is difficult to relate the findings here to the broader populace, we argue that the provided results nevertheless provide useful insights into the views held by a respondent sample that is likely to have vested interests in issues like the ones under scrutiny here—online debate culture, freedom of speech, etc. Due to differences in the ways the invitation to participate in the survey was posted in the four newspapers, the invitation was visible to a much larger share of the readers in Dagbladet and Aftenposten than in the other two newspapers. As such, readers of these two latter publications are likely to be overrepresented in our sample.

As previously noted, in the time after the terror attacks there were fervent public debates about the quality and perceived problems with online comment fields. The data collection period was initiated about one month after the trial against the terrorist, which coincided with the release of a parliamentary report into the events of the 2011 attacks. Given this context, it is likely that respondents have had a high degree of awareness towards controversies surrounding online comment fields.

As for dependent variables, these are outlined in Table 4.

In the analysis that follows, we will use the three questions outlined in Table 4 as indicators of the respondent’s attitude towards three dimensions of editorial control, in accordance with the interventionist–noninterventionist model introduced above. The first question included, labeled “rules” in Table 4, gauges whether the respondent finds the limits for what one may (or may not) write too strict or too allowing, serving as a measure of the respondent’s attitude towards the rules set by the publication both through the explicit set of rules, and the implicit limits expressed through the moderation practices. The second question, labeled “presence”, addresses the respondent’s attitude towards an issue that has often been raised in debates about online comments: namely the idea that journalists and moderators ought to participate more actively in the discussions on their website. Finally, the third variable, “anonymity”, gauges whether the respondent tends to identify with his or her real name or use a pseudonym. Each of the questions are connected with our model: if a respondent thinks the rules are too strict, does not wish the moderators to be present and prefers to use a pseudonym, that respondent should be considered to belong to the noninterventionist side of the spectrum. Conversely, respondents who answer in the opposite should be considered as interventionists.

Results

In the following analysis, we will treat each of the questions presented in Table 4 as dependent variables and test for the influences of the independent variables defined in Table 3. The “rules” question is analyzed by means of multiple regression analysis, whereas logistic regression is used for the variables called presence and anonymity. Table 5 gives an overview of the performed analyses.

TABLE 5 Results from regression analyses of the variables “rules”, “presence” and “anonymity”

	Rules	Presence	Anonymity
Age	−0.05***	0.21**	−0.15*
Gender	0.17***	−0.17	0.02
Education	−0.02	−0.05	0.27***
Frequency	0.02	−0.02	−0.03
Quality	0.25***	0.16**	−0.02*
Experience of moderation	0.32***	0.54***	0.81***
Change of presence	0.20***	0.57***	−0.03
Change of freedom	−0.57***	−0.28***	−0.31***
Website
VG	−0.01	0.16	−0.70***
Dagbladet	0.03	−0.40**	0.27*
Aftenposten	0.02	−0.26*	0.05
NRK	−0.05**	0.13	0.01
TV2	−0.01	−0.24	−0.11
Nettavisen	0.04*	−0.06	−0.76***
Vårt Land (verdidebatt.no)	−0.02	−0.27	−0.89***
Dagsavisen (nyemeninger.no)	−0.04*	−0.64**	−0.90***
Hegnar Online	0.05**	0.43*	0.71**
Document.no	0.03	0.01	0.67*
Others	−0.04*	−0.14	−0.46***
R^2 measures	R^2 = 0.44, adjusted R^2 = 0.43	Cox & Snell R^2 = 0.09, Nagelkerke R^2 = 0.15	Cox & Snell R^2 = 0.15, Nagelkerke R^2 = 0.20

Values are adjusted beta.

*p < 0.05, **p < 0.01, ***p < 0.001.

As shown in the first row of Table 5, age emerges as a significant factor in all three analyses. However, the results point to differing directions on the proposed interventionism spectrum—given the results presented in Table 5, younger respondents tend to view the rules as too strict and tend to participate anonymously, both of which indicate a noninterventionist attitude. However, comparably younger respondents also lean towards moderators to be present—defined as a sign of interventionism in the study at hand. Such a possible inconsistency could depend on a number of factors—perhaps related to online conventions that young people have adapted to a larger degree than older respondents, invoking expectations of anonymity and a large degree of freedom, but also of active moderators who can respond to user input. Focusing on significant results only, the subsequent two demographic variables—Gender and Education—indicate that males tend to think the rules are too strict, and respondents with higher education tend to use pseudonyms. Surprisingly, there is also no effect of frequency of debate on any of the three dependent variables, indicating a limited association between how often a respondent contributes with online comments, on the one hand, and their attitudes towards editorial control, on the other.

Regarding respondent views of debate quality, we again see significant but somewhat contradictory results. For the “Quality” variable, then, respondents who respond that debates do not work well tend to think the rules are too allowing and that moderators should be present, placing them on the interventionist side of the spectrum. The association with the anonymity variable points in the opposite direction, indicating that those who think the debates do not work well tend to participate anonymously; however, this effect is quite weak.

The next independent variable—labeled “Experience of moderation” in Table 5—results in highly significant effects in line with our model as presented in Table 1. Specifically, respondents who have experienced that their comments have been edited or removed tend to think the rules are too strict, they are opposed to moderators being present and they tend to write anonymously, all positions that are associated with noninterventionism.

The two subsequent variables gauge respondent perception of changes after the 2011 terror attacks. “Change of presence” shows highly significant and positive effects in relation to our first two dependent variables. This finding indicates that those who find that moderators have become more active tend to think that the rules are too strict. Similarly, these respondents do not wish for moderator presence, placing those who answered in this way on the noninterventionist side of the spectrum. These findings show a strong association between noninterventionist attitudes and a perception of change towards more interventionist practice in the newspapers. Such an interpretation is supported by the findings for the next variable—labeled “Change of freedom”. Here, we see that those who find that their freedom to speak their minds has deteriorated after the terror attacks associate with the noninterventionist stance on all three dependent variables. Bearing in mind that answers to both of these questions were heavily skewed (cf. Table 3), it seems that there is a large group of respondents who perceive editorial control as having increased after the terror attacks. Moreover, this perception seems to be strongest among those respondents who are opposed to editorial control.

Finally, the last 11 variables featured in Table 5 provide results related to which websites the respondents reported to write comments on. (Note that although the invitation to participate in the survey was only posted on four of these websites, the question included other websites as well, and respondents could choose several websites.) Unsurprisingly, we see strong associations with the anonymity variable, since some of the websites (such as VG, Vårt Land and Dagsavisen) did not allow anonymous comments at the time. The effects on the other two dependent variables seem to offer two main interpretations. The first is that respondents choose to write comments on websites whose policies for editorial control match their own preferences. This is supported by the association between noninterventionist stances and participation on websites such as Hegnar Online and Nettavisen, which have long been regarded as having an unusually noninterventionist policy regarding editorial control with online comments; and by the association between the “rules” variable and participation in Dagsavisen, known for interventionist policies. Be that as it may, the suggested pattern is not entirely clear when regarding the other websites (in particular for Dagbladet, known for featuring noninterventionist policies). Secondly, taking the political orientation of the websites into account allows us to consider whether respondents from right-wing and left-wing websites differ in their attitudes towards editorial control. Among the six websites listed in Table 5 with significant results for the “rules” or “presence” variables, Hegnar Online and Nettavisen are typically considered to be right of center in Norwegian politics, whereas Dagsavisen and Dagbladet are associated with the center-left, while NRK and Aftenposten are closer to the center. As shown in Table 5, respondents from the two center-right websites indicate noninterventionist attitudes, while respondents from the other four websites have given more interventionist answers, and this effect is strongest for the two websites from the center-left.

Discussion

Regarding RQ1, the results presented above seem to indicate that respondents diverge along an interventionist–noninterventionist axis with particular regard to two aspects. First, their perception of the quality of debates, and second, their experience of moderator interventions. Regarding quality, we see that those who rate the quality of the debates as poor tend to lean towards the interventionist side of our model. In other words, those who see a problem with the quality of the debates tend to feel that there is a need for stronger editorial control regarding rules and moderator presence. This mirrors much of the public criticism of online comment fields that was raised after the 2011 terror attacks, and could be seen as an indication that the respondents on the interventionist side of the spectrum tend to agree with opinion leaders and media commentators more than the noninterventionists.

As for experiences of moderator interventions, we see that such experiences are clearly associated with noninterventionism for all our three dependent variables. This finding should be seen in light of another question asked in the survey. In order to test an observation from earlier research that moderation practices suffered from a lack of transparency (Ihlebæk and Ytreberg 2009; see also Springer, Engelmann, and Pfaffinger 2015), we asked a follow-up question to those respondents who answered affirmatively to the question regarding experiences of moderation. This question was phrased as follows: “Has it ever occurred, when one of your comments has been removed or edited, that you could not understand why this was done?” An overwhelming majority—84 percent—of those who had experienced editorial intervention indicated that they had experienced difficulty understanding the reasons for the intervention. As an attempt at interpreting these results, one might speculate that a causal relation is at play here. If one does not understand why moderators edit or remove one’s contributions, one may easily get the impression that editorial interventions are arbitrary or hostile, and therefore become skeptical towards editorial control. Perhaps this lack of understanding might also contribute to the frequent assertions of “politically correct” censorship that often comes across in online comments. As noted elsewhere (Løvlie and Ihlebæk 2016), we have seen indications in qualitative analysis of free-form answers to other questions in the survey that many respondents are suspicious about the fairness of and the motives behind editorial control. However, although we have fashioned this as an independent variable for the purpose of analysis, we do not have data to assert a causal relation. It is indeed possible that the influence is mediated by a third factor: that those initially skeptical of editorial control are more likely to be the target of editorial interventions—for instance, if respondents who have noninterventionist attitudes also have a tendency towards discussing controversial opinions or topics, a confrontational style of debating, lack of digital literacy or understanding of editorial policies. These issues could serve as useful starting points for future research.

Finally, given the attention often devoted to gender when it comes to issues like those discussed here, it is surprising to note that we only find a significant effect of gender on the variable gauging the limits for what one may write, with female respondents tending to view such limits more favorably than males. However, this particular finding is supported by a similar finding in Swedish data from 2011 reported by Bergström and Wadbring (2015, 9–10), who found that women were more likely than men to agree that abusive comments should be censored. One possible interpretation of these combined observations could be to see them in light of Enjolras et al. (2013, 122–123), who find that men more often than women participate in political debate online—except on Facebook, where women report as being equally active. Enjolras et al. interpret this as reflecting not only that women use Facebook more than men, but also that women prefer to debate with people they know well. Earlier findings indicating that women receive more gendered and sexual harassment and limit their own utterances to a larger degree than men suggest that women may, to a larger degree than men, consider the act of writing comments on newspaper websites as a risky activity (Staksrud et al. 2014). With these findings in mind, it is not surprising that female respondents are more inclined than males to desire stricter rules and censoring of abusive comments. However, it is surprising that no significant effects of gender on the questions regarding moderator presence and anonymity were found in the paper at hand. Moreover, given other findings that contradict the idea that female respondents receive more harassment and threats than males (Pew Research Center 2014), one might consider an alternative explanation: maybe there is no clear difference between female and male respondents in their view of risk connected with writing online comments—rather, the difference is perhaps better understood in relation to male and female views on the appropriate limits for utterances in online comments. Our data simply suggest that our male respondents desire looser limits for what one may write in online comments than our female respondents did.

Turning to RQ2, our results indicate that those respondents who feel that editorial control has increased following the terror attacks have a clear tendency to be noninterventionists. In other words, those who were the most skeptical of editorial control had the strongest impression that such control had increased after the terror attacks. In this context, it is interesting to note that respondent experience of changes after the terror attacks seem to be stronger than changes as perceived by the newspaper editors and moderators focused on in our earlier interview study (Ihlebæk, Løvlie, and Mainsah 2013). In that study, all the respondents asserted that they had not made any changes in the rules dictating what could be written in comment fields after the terror attacks—even though some indicated they had tightened the control in other ways, such as limiting which articles users could comment on and hiring extra moderators to overlook the debates. We have no independent data to assess the actual magnitude of changes, so we cannot determine which is more accurate—the small changes described by editorial informants in the aforementioned study, or the massive changes suggested by the survey respondents in the current one. It is certainly possible that the editors and moderators may have underestimated the changes that had occurred—perhaps because the changes had been tacit, unconscious or because the informants might have had an interest in countering a perception of insufficient editorial control prior to the terror attacks. Conversely, respondents might exaggerate the changes due to the public controversies surrounding comment fields, misremembering the situation prior to the terror attacks or perhaps also due to a feeling of being censored or even ostracized due to their controversial opinions. Regardless of which view one takes, the fact that many respondents seem to experience these changes as a strong curbing of their freedom to speak their minds seems to indicate a problematic climate for debate and lack of understanding between the newspapers and their audiences.

While the study at hand has provided important insights into online comment fields, it has limitations that need to be duly addressed. First, the results from our analysis should be read with some caution, due to the fact that the sample is self-selected. Second, a number of factors could influence respondents statements about editorial control which we have not been able to include in this analysis. For example, due to the way that the survey was set up, we could not directly identify from which newspaper website the respondent clicked on the link to the survey. Therefore, we had to rely on the respondents’ own statement about which website they usually wrote comments on. This has arguably limited our ability to analyze the influence of the website on the respondents’ experiences with editorial control, and we urge those interested in these issues to take such issues into account in future, similar projects. Third, while we have asked questions about the respondents’ general views and practices regarding comment fields on newspaper websites, we have not explored their views and practices in relation to particular article topics or debate themes, such as politics, immigration or terrorism. As such, the results presented here should be supplemented with studies probing into particular websites and topics. Fourth, it is likely that the strong controversies surrounding online comments at the time of the survey may have affected the answers provided. It would be beneficial to conduct repeated surveys in order to assess whether the findings in this article hold true over time, or whether recent developments have changed user perspectives on editorial control.

In this study, we have expanded our use of the interventionist–noninterventionist model to not only describe editorial control strategies, but also the users’ experience of editorial control. Further work should aim to develop this model further, perhaps by adding more dimensions to the model and consider more nuances in the variety of editorial strategies and user experiences. Further research should also explore the growing role of social media platforms such as Facebook and Twitter, and the commenting practices and editorial control strategies for newspapers publishing on these platforms.

Conclusion

Above we have explored a model for analyzing users’ attitudes towards editorial control with comment fields along a spectrum from interventionism to noninterventionism. We find that among the respondents to our survey, the interventionists tend to be most critical towards the quality of online comments, whereas the noninterventionists have most often experienced being the target of editorial control measures. The noninterventionists also have a stronger tendency to report that their freedom to speak their mind has deteriorated after the terror attacks.

The observation that female and male respondents desire differing limits for acceptable utterances in online comments could be seen as a conflict of interest between female and male respondents—and as a dilemma for media professionals. Bearing in mind that males are in the majority among contributors to online newspaper comment fields, media professionals might be tempted to listen to the plurality of respondents who think the limits are too strict and decide to implement noninterventionist rules and moderation practices. However, that would run contrary to the wishes of the interventionists, where females are overrepresented. A strategy for broadening participation in the comment fields might need to include measures to improve the experience for female participants, which according to our results might imply stronger rather than looser limitations. For different newspapers this strategic dilemma might weigh differently depending on their prioritized target groups, but for society as a whole democratic ideals dictate that we should strive for the broadest possible participation in public debates.

Taken together, our findings seem to suggest that many of our respondents have a strong experience of antagonism towards moderators and editorial control measures. In particular, the fact that a very large majority of those who have experienced editorial interventions report that they could not understand why these interventions occurred suggests that newspapers should consider changes to increase the transparency of moderation practices, in order to improve the user experience and reduce distrust between participants and moderators. The main challenge in achieving this is that it requires resources in the form of redesigned systems and/or more active moderators, as indicated by recent shifts in strategy for one of the newspapers in our sample, Dagbladet. In November 2014, the newspaper hired a new group of moderators and announced a new editorial strategy for their comment fields, where one of the core elements was that the moderators should have an active presence and give explanations to the contributors when comments were deleted (Suvatne 2014). However, after following this strategy for a little over a year, the newspaper recently announced that they would now shut down their comment fields due to a need to prioritize resources towards social media (Ramnefjell 2016). This example indicates that (at least some) newspaper professionals do recognize a need for more transparency in their moderation practices. This need is not likely to go away even if comments are shifted to Facebook or other social media.

Given the high level of attention currently being given towards various control measures in order to combat problems such as “fake news” (Jackson 2017; Morozov 2017), it is important to gain further understanding not just of the efficacy of editorial control with online comments—but also of the ways in which such control measures are experienced by users. If the way in which editorial control is practiced by online newspapers is contributing to feelings of distrust and animosity among users, this would be damaging not just for the newspapers’ relationship with their readers—but also for the role of newspapers as facilitators of democratic debate.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Council for Applied Media Research.