Abstract
Scholars are increasingly concerned about the rising level of negativity in social media sites. This negativity has found its way into sites that are supposedly intended for prosocial civic engagement. To examine how hostility impacts behavior in a user-generated, prosocial context, an experimental study was conducted using an online petition modeled after those posted to the website Change.org. This study examines whether negativity causes a contagion effect leading to more negativity and the different types of negativity that may occur. Results suggest that when users read negative-toned petitions, a contagion effect increases both anger and anxiety. However, our findings are not consistent with previous literature that argues anger leads to increased proactive behavior. Instead, we find that while anxiety leads to an increase in petition-related action, anger does not. These findings have important theoretical and practical implications for scholars and those looking to participate in social justice via online platforms.
The summer of 2016 started off with a host of public concerns. From an incident at the Cincinnati Zoo, to the assault case involving a Stanford University swim star, to the violence in an Orlando nightclub that reignited the debate over gun control laws, the potential for civic involvement was substantial. In light of these events, Facebook and Twitter feeds erupted with affiliated hashtags and images, posted and reposted many times over: countless voices looking to affect change. Advances in social media platforms have opened up considerable opportunity for positive social change (Shirky, 2011) and prosocial behavior. Support chat rooms, social networking sites, Wikis, donation/fundraising sites and online petitions, offer users the ability to engage in behavior intended to provide help to those in need. One predominant example is the petition website Change.org. With respect to just the three aforementioned events, over 50 petitions were created, totaling well over one million signatures (Change.org). Petitioning through interpersonal exchange is a form of social action as old as the United States Constitution (Higginson, 1986), but online petitioning offers something novel—now every citizen has the opportunity to engage in prosocial behavior with minimal labor or time commitment, and without spatial constraints.
These reduced barriers to engagement, however, also open the door for anti-social behavior. Anonymity, lack of non-verbal cues, undefined consequences and weak social ties can encourage uncivil online discussion (Hill and Hughes, 1998; Papacharissi, 2002). Online incivility is a particularly pressing topic (e.g. Anderson et al., 2014; Papacharissi, 2004; Sobieraj and Berry, 2011); the negative emotions people experience, such as anger and anxiety, can have a profound effect on sociopolitical behavior (Marcus et al., 2000). While anger and anxiety may be spurred by the same uncivil discourse, the resulting behavior can be quite different (Wagner, 2014). Anger typically involves heuristic processing (Wagner, 2014), while anxiety can prompt cognitive elaboration (Valentino et al., 2008). Additionally, anger typically leads to less information-seeking but more proactive behavior, while anxiety tends to prompt information-seeking but suppresses action (Valentino et al., 2011).
Although numerous scholars, pundits, and social critics have lamented the growing negativity in online spaces (e.g. Anderson et al., 2014; Papacharissi, 2004; Sobieraj and Berry, 2011), not all negativity is created equal. In some cases, negativity may actually have positive consequences. For example, incivility can encourage participation and action (Coe et al., 2014). Although scholars have examined the divergent effects of different types of negative emotions in political communication (Valentino et al., 2008, 2011; Weeks, 2015), these distinctions have not been thoroughly examined with respect to civic engagement. Thus, examining how different forms of negativity affect behavior is crucial to understanding the potential positive and negative repercussions for online prosocial channels, such as Change.org. As far as scholars have progressed in understanding online channels and their influence on aggressive communication, further research is needed (Cicchirillo et al., 2015).
To this end, this research examines how the contextual features (petition tone, and emphasis on a culpable agent or the crisis situation) of a Change.org petition can (1) affect the type of comments a petition reader will post and (2) how the emotions elicited by these responses mediate petition-related behavior (i.e. signing the petition, forwarding the petition, or information-seeking). Using a 2 × 2 experimental design, we directly test how these contextual factors can affect petition-related behavior—providing a much-needed test of causality largely missing from the online incivility literature. In doing so, this research broadens our understanding of the influence of social cues in the language of petitions posted to Change.org and how these cues engage or disengage its users.
Positivity in digital spaces
Prosocial behavior is defined as “voluntary, intentional behavior that results in benefits for another” (Eisenberg and Miller, 1987: 92). Altruism has a long history offline, but online sources substantially reduce the geographic and temporal restraints that may have limited a person’s ability to reach out and help others (Rains and Young, 2009). While prosocial online sites are often directed externally with the intention of benefiting others, scholarly work has primarily focused on examining incivility on sites in which behavior is more self-interested, such as news and political websites and message boards (e.g. Coe et al., 2014; Papacharissi, 2004; Robertson et al., 2013). News consumption is often civic, but not inherently prosocial—when it comes to sociopolitical topics, many citizens are primarily motivated by the desire to gather support for or defend one’s beliefs and worldviews (Edgerly et al., 2014).
Conversely, digital activism networks, such as Change.org, tend to attract like-minded users, typically large numbers of anonymous strangers, with a common goal of helping others (Sproull et al., 2005). Electronic petitions present the potential for prosocial behavior that encompasses not just politics but everyday citizen concerns. It has been argued that electronic petitions are the “most prominent e-democratic innovation in the world to date” (Wright, 2012: 453).
Change.org is arguably the most well-known online space for electronic petitions (Geron, 2012). Launched in 2007, Change.org is a social network website that enables users to initiate online petitions on any subject they desire. Change.org currently boasts over 100 million users worldwide in 196 different countries, labeling itself as the “world’s largest petition platform.” Time magazine listed its founder as one of the 100 most influential people of 2013 (Time).
Much of what Change.org is meant to do is help social others with no direct connections. Despite the noble intentions of Change.org, it also provides a perfect channel for aggression and incivility. Specifically, Change.org’s comment sections are often filled with posts expressing a range of negative emotions, from hostility to despair. Online petitions as conduits for potential prosocial action are growing daily but have received little attention from scholars concerned with online incivility.
Negativity in digital spaces
There is a wealth of research on effects that come from negative discourse in traditional, sociopolitical channels (Cho et al., 2015). This incivility has, naturally, made its way into online spaces as well. Social media, for example, may promote negative messages (Robertson et al., 2013), and the Internet has been noted as a great place to air grievances (Lee and Cude, 2012). Even among the engaged and politically inclined online user, incivility is commonplace (Hutchens et al., 2015). Due to the pervasiveness of online communication and the current sociopolitical climate, every citizen has the opportunity to become pundits in their right.
Although online sites like Change.org are largely intended to help achieve positive social change, they too provide forums ripe for hostility. Unlike traditional petitions, a key component to Change.org is the ability for signers to share their reasons of support, creating an online space for community members to express themselves—both pro and anti-social. The possibilities for comment sections, like this, to support a healthy public dialogue are considerable, yet they often do not (Papacharissi, 2004). Comment sections in online sites are hotbeds of negativity (Coe et al., 2014), offering users a space to lash out (anonymously if they choose; Papacharissi, 2004) or engage in flaming (Hmielowski et al., 2014; Hutchens et al., 2015). These negative tendencies are especially pronounced when sociopolitical topics are involved.
A contextual snowball
When considering appropriate behavior, people have an innate impulse to look to others for cues. The Web is highly interactional, and users have the opportunity to express hostility with little recourse. This may produce an emotional snowball or social contagion effect (Fowler and Christakis, 2008) as users look toward others for group norms and ethical considerations (Chin et al., 2015). This is especially true when considering online comment sections. For example, when examining comments of news stories, it was found that between 20% and 50% of commenters did not respond to the journalistic news article but rather to a previously posted user comment (Ruiz et al., 2011). This suggests that commenters are often influenced by each other.
Despite this influence, commenters are typically connected by weak ties. Participating members can remain anonymous and do not have to directly observe the consequences of their behavior (Hill and Hughes, 1998; Papacharissi, 2002). When users do not fear interpersonal repercussions, weak ties can lead to a lower incentive toward ethnical behavior (Granovetter, 1973). Making matters worse, individuals often alter their own judgments in order to be consistent with group expectations (Hanson and Putler, 1996). When the views of others are already established, participants strategically shift their attitudes toward that of the known viewpoint (Tetlock et al., 1989). When faced with an online negative environment, going against your own ethics to match those of the crowd, especially when there is little possibility that you will encounter these people again, is deemed appropriate. Once a hostile norm is established, the potential for others to follow the herd increases. The social contagion effect begins.
Previous work has demonstrated the potential for the social contagion of emotions online. When researchers manipulated Facebook “News Feed” items to contain a higher percentage of either negative or positive-toned messages, emotions followed suit. Facebook users, who received less positive messages, posted less positive messages in their own News Feeds (Kramer et al., 2014). We therefore expect that the tone of a Change.org petition should affect the tone of the response of petition commenters, and when petition signers are exposed to negative-toned petitions and angry comments, they will be more inclined to emulate that behavior.
H1. After reading a negatively toned petition and comment section, readers’ comments will convey more negative emotions (compared to neutral toned petitions).
Anger or anxiety?
If there is a snowball effect impacting hostile discourse in online settings, is there a facilitator rolling the first snowball? The initial response to hostility may be the result of an emotional reaction. Emotions play a crucial mediating role in determining the behavioral response a person has to incoming information (Valentino et al., 2008). According to Marcus et al. (2000), emotions inform people whether they can rely on pre-established habits and routines to respond to an environmental stimuli or whether they need to engage in effortful processing and potentially enact a new course of action.
Contrary to conventional wisdom, heightened emotions often assist in making correct decisions (Marcus et al., 2000). In particular, negative emotions have the ability to disrupt behavioral routines and activate a conscious evaluation of relevant information. There are conflicting accounts, however, of whether negative emotions lead to prosocial (e.g. information-seeking, deliberation, political participation) or anti-social (e.g. hostility, selective exposure) behavior. These discrepancies are not necessarily at odds because heightened emotions are not universal. While Marcus et al. (2000) treated negative emotions as relatively homogenous, scholars have emphasized the important distinctions between anxiety and anger, the causes of each emotion, and the consequences on sociopolitical behavior (Valentino et al., 2008, 2011; Weeks, 2015).
Anger typically results when the cause of a threat is known (e.g. who is responsible) and/or a person feels that they have power over the situation (Wagner, 2014). Anger should rise, therefore, when a person believes a specific actor is creating an imminent threat. In the case of an online petition, we would expect that petitions that focus on the cause of a particular crisis would elicit more anger than a petition that focuses primarily on the crisis situation. Thus, we predict the following hypothesis:
H2. When a petition focuses attribution on an external agent, petition readers will post angrier messages.
Conversely, when an individual is not entirely sure who is responsible for a threat or crisis and/or they do not feel they have power over a situation, they are more likely to feel anxious and afraid (Wagner, 2014). We would expect that petitions that focus more on the imminent threat or crisis itself, rather than who is to blame for the threat, should elicit more anxiety. We therefore predict the following hypothesis:
H3. When a petition focuses on how a situation is affecting the victims, petition readers will post more anxious messages.
Anger and anxiety can also lead to very different behavioral outcomes (Wagner, 2014). Angry people assume their frustration is rooted in a known reality and, therefore, see no need to seek out more information. Rather, they tend to defer to pre-existing habits and rely on contextual heuristics (Valentino et al., 2008). If an inflammatory comment or online blog post, for example, produces anger, individuals may become more hostile, yet less inclined to counter with a thoughtful response.
Although anger tends to shut off information-seeking, it does prompt action. Because angry people are likely to think they know who is at fault and to have higher perceptions of self-efficacy, they tend to engage in more proactive and risk-seeking behavior (Valentino et al., 2011; Wagner, 2014). For example, a study found that uncivil blog commentary increased online participation despite decreasing open-mindedness (Borah, 2012). Angry users care less about accurate information and more about lashing out.
Anxiety, on the other hand, often leads to deliberative information-seeking; when people get nervous, they desire better information to guide their actions (Marcus et al., 2000), or they may take low cost, expressive actions (Valentino et al., 2011). Because fearful citizens are less sure about who is to blame, or what can be done, they are less prone to engage in proactive behavior, preferring, instead, to gather more information. If an online comment section stimulates anxiety, moderate actions, such as Googling the topic for more information or interacting with other commenters, may result. Based on these alternating actions, we predict that anger toward the content of a petition will produce a reactive response. Anxiety, however, will produce a desire to learn more about the topic.
H4. When petition readers express more anger, they should be (a) more likely to engage in petition supporting behavior (i.e. signing or forwarding the petition) but (b) less likely to pursue additional information.
H5. When petition readers express more anxiety, they should be (a) less likely to engage in petition supporting behavior (i.e. signing or forwarding the petition), but (b) more likely to pursue additional information.
Methods
A total of 400 participants were recruited through Amazon.com’s Mechanical Turk (MTurk). The sample was 50.4% male, 49.6% female, 76.8% white, 9.9% African American, 4.9% Hispanic, 6.0% Asian American. The mean age was 37.87 years (SD = 1.41). Party affiliation was 42.2% Democrat, 20.4% Republican, 30.0% independent, and 7.3% other. Because the sample was slightly skewed toward Democrats, we checked random assignment of political affiliation and found no evidence that party affiliation differed by experimental condition, χ2(9) = 5.31, p = .807.
Stimulus material
As the purpose of this study was to examine the responses posted by online petitions signers, ecological validity (replication of a petition posted to Change.org) was key. Change.org has hundreds of petitions ranging in genres and levels of importance. After examining numerous petitions posted to the website, we chose a petition requesting the Russian government grant temporary asylum to a Syrian refugee family.
This study employed a 4-cell between subjects design (Russian Neutral, Russian Negative, Refugee Neutral, Refugee Negative). To construct the tone conditions, we manipulated the language inside the petition to include either neutral or negative references to Russia or neutral or negative language in reference to the refugee situation. The same petition template was used across all conditions and great care was taken in order to keep the content as consistent as possible, aside from the manipulation variables.
In the negative conditions, the manipulation included contemptuous language (e.g. “Dictator Vladimir Putin and Minister of the Foreign Affairs, Sergei Lavrov constantly declare that ‘we must help the Syrian people’. But instead they treat them like animals”), the neutral tone condition (e.g. “Vladimir Putin and Minister of the Foreign Affairs, Sergei Lavrov declared that ‘we must help the Syrian people’, but more must be done to uphold human rights.”) did not.
To examine how the target of the crisis affected behavioral response, the petitions were either focused on the crisis situation (“they have been unable to enter the country”) or on Russian’s role in the crisis (“Russian authorities have denied them asylum and won’t allow them in the country.”).
Additionally, comment sections were manipulated to match tone of the petition. Compare a negative comment (e.g. “Putin, get off that horse, put on a shirt, and act like a human being.”) to a neutral one (e.g. “I hope that Putin decides to act and make this happen for your family. He’s a human too.”). To attempt to mirror real-world commentary, misspelled words, incorrect grammar, bolded words, capitalized words, and gratuitous punctuation was used. Many of the comments were sourced directly from real Change.org petitions.
Measures
Information-seeking
Information-seeking was measured by asking participants on a seven-point scale, “based on the previous petition, how likely would you be to seek out more information about this topic?” (M = 4.31, SD = 1.66).
Willingness to sign
Willingness to sign was measured by asking participants on a seven-point scale, “based on the previous petition, how likely would you be to sign this petition?” (M = 4.41, SD = 2.01).
Intentions to forward
Intention to forward was measured by asking participants on a seven-point scale, “based on the previous petition, how likely would you be to forward this petition to others?” (M = 3.32, SD = 1.78).
Emotional reactions
After viewing the comment section of the petition, participants were asked if they would like to add in their own comment in a free text box. Overall, 353 of the 400 participants added their own reaction to the petition. Participants who did not provide a comment were not used for subsequent analysis (see Discussion section). Comments were then coded on anger and anxiety scales. Inter-coder reliability was assessed on a 10% subsample. Results demonstrated satisfactory inter-coder reliability for anxiety (Krippendorff’s α = .81) and anger (Krippendorff’s α = .88).
Anxiety
Anxiety was operationalized as the degree to which participants’ comments expressed fear or concern about the situation. Anxiety was coded on a five-point scale with (1) indicating comments that expressed certainty that there was no cause for concern (e.g. “each country has its right to do what they feel is right for their country”) and (5) indicating comments that were very anxious and concerned (e.g. “this is terrible!”; M = 3.38, SD = 1.24). Anger was operationalized as the degree to which participants’ comments demonstrated fury, hate, frustration, and/or resentment. Anger was coded on a five-point scale with (1) indicating comments that were focused on love and kindness (e.g. “everyone has a heart including Putin”) and (5) indicating comments that were filled with anger (e.g. “this is disgusting treatment by Russian authorities”; M = 2.76, SD = 1.25). Previous research has noted that anger and anxiety tend to be correlated (Valentino et al., 2011). In our study, the strength of the correlation was weak enough that we felt justified in treating the emotions as distinct constructs (r = .20).
Analysis
Analysis was performed using the SPSS Macro PROCESS (Hayes, 2012). Specifically, mediation was estimated by examining the relationship between an X variable (tone or target), a Y variable (information-seeking, intention to sign, forwarding), and two M (mediating) variables (anxiety and anger). When tone was used as the X variable, target was used as a covariate. When target was used as the X variable, tone was used as a covariate (Figure 1).
Process model.
Indirect effects were examined using a bootstrap analysis with 1000 bootstrap samples and a 95% Confidence Interval (CI). Using this method, when an indirect path’s CI does not overlap zero (e.g. 0.1–0.3), it is considered significant, and if it does overlap zero (e.g. −.1–.3), it is considered non-significant.
Results
Tone × target interaction
We first checked whether there was an interaction between the tone and target manipulation using Process Model 8. Results demonstrated that there was not a significant interaction between tone and target on anger, b = −0.22, standard error (SE) = 0.26, t = −0.84, p = .40; anxiety, b = −0.21, SE = 0.26, t = −0.79, p = .43; information-seeking, b = −0.27, SE = 0.33, t = −0.83, p = .41; willingness to sign, b = 0.04, SE = 0.39, t = 0.10, p = .92; or intention to forward, b = 0.03, SE = 0.35, t = 0.07, p = .94.
Tone and anger
Our first hypothesis predicted that a petition with a negative tone would lead to higher levels of negativity in participants’ comments. As seen in Table 1, in the negative tone conditions, participants are significantly more likely to express anger, b = 0.70, SE = 0.13, t = 5.45, p < .000. Thus, H1 was supported.
Results for direct, indirect, and total effects of experimental manipulations.
p < .001; **p < .01; *p < .05; #p < .10.
We also predicted that the more participants expressed anger in their comments, the less likely they would be to seek out additional information, but the more likely they would be to express a willingness to sign and forward the petition.
Looking at intention to sign, there is a significant negative path from anger to intention to sign, b = −0.26, SE = 0.09, t = −2.87, p = .004, and a significant indirect path from petition tone to intention to sign, b = −0.18, SE = 0.08 (95% CI = [−0.36, −0.06]), p < .05.
There is not, however, a significant path from anger to forwarding, b = −0.12, SE = 0.08, t = −1.44, p = .15, nor is there a significant indirect path from petition tone to forwarding through anger, b = −0.08, SE = 0.06 (95% CI = [−0.22, 0.03]), ns. H4a was therefore rejected.
As seen in Table 1, there is a significant negative path from anger to information-seeking, b = −0.22, SE = 0.07, t = −3.03, p = .003. Additionally, there is a significant negative indirect path from petition tone to information-seeking, b = −0.16, SE = 0.06 (95% CI = [−0.30, −0.05]), p < .05. Thus, H4b received support.
Tone and anxiety
H1 predicted that a negative tone would lead to comments that exhibited higher levels of anxiety. As seen in Table 1, in the negative tone conditions, participants are significantly more likely to express anxiety in their comments, b = 0.27, SE = 0.13, t = 2.05, p = .040. Thus, H1 was supported.
We also predicted that anxiety would be positively related to information-seeking but negatively related to intentions to sign or willingness to forward the petition. There is a significant path from anxiety to intention to sign, but contrary to expectations the relationship was positive, b = 0.38, SE = 0.09, t = 4.41, p < .001. Similarly, there was a significant indirect path from petition tone to intention to sign through anxiety, b = 0.10, SE = 0.06 (95% CI = [0.01, 0.25]), p < .05. Additionally, there is a significant positive path from anxiety to forwarding, b = 0.32, SE = 0.08, t = 4.17, p < .001, and a significant indirect path from petition tone to forwarding through anxiety, b = 0.09, SE = 0.05 (95% CI = [0.01, 0.20]), p < .05. H5a was therefore rejected.
Looking at information-seeking, there is a significant path from anxiety to information-seeking, b = 0.28, SE = 0.07, t = 3.90, p < .001. Furthermore, our results show that there is a significant indirect path from petition tone to information-seeking through anxiety, b = 0.08, SE = 0.04 (95% CI = [0.01, 0.18]), p < .05. Thus, H5b was supported.
Target and anger
H2 predicted that a petition focused more on a culpable agent (Russia) rather than the victims (Syrian refugees) would lead to higher levels of anger in participants’ comments. In the refugee conditions, participants are significantly less likely to express anger in their comments, b = −0.37, SE = 0.13, t = −2.88, p = .004. Thus, H2 was supported.
As seen in Table 1, there is a significant negative relationship between anger and information-seeking, b = −0.22, SE = 0.07, t = −3.03, p = .003. Additionally, there was a significant indirect path from petition target to information-seeking through anger, b = 0.08, SE = 0.04 (95% CI = [0.02, 0.20]), p < .05. Meaning, when the petition focused on refugees, it tended to reduce anger, which was negatively related to information-seeking.
Additionally, there was a significant negative relationship between anger and intention to sign, b =
There was not, however, a significant relationship between anger and willingness to forward, b =
Target and anxiety
H3 predicted that a petition focused more on the victims (Syrian refugees) rather than a culpable agent (Russia) would lead to higher levels of anxiety in participants’ comments. As seen in Table 1, there is not a significant relationship between the target of the petition and anxiety, b = 0.20, SE = 0.13, t = 1.53, p = .13. We did not, therefore, find support for H3.
It is worth noting, however, that we again find a significant path from anxiety to information-seeking, b = 0.28, SE = 0.07, t = 3.90, p < .001, intention to sign, b = 0.38, SE = 0.09, t = 4.41, p < .001, and forwarding, b = 0.33, SE = 0.08, t = 4.17, p < .001. We did not, however, find any significant indirect effects (Table 1).
Table 2 displays the descriptive statistics for depentant variables by condition.
Descriptive statistics for dependent variables by experimental condition.
Discussion
The primary goals of this study were to examine the different factors that may contribute to negativity on a seemingly prosocial social networking site and how different forms of negativity connect different forms of online petition-related behavior. We argued that when petition signers were exposed to more hostile messages—in the petition itself as well as the comment sections—commenters would exhibit hostility and anxiety within their own comments. Our findings provide important insights into (1) the factors that may increase online negativity, (2) how different types of negative emotions relate to different types of online behavior, and (3) the unique context of online petitions, a growing channel in sociopolitical discourse.
First, as expected, when participants were exposed to negatively toned petitions, they tended to follow suit with a higher degree of anger. These findings provide support for the claim that in an online context, individuals often go along with the herd. When a social networking site, or user comment, sets a negative tone, other commenters are likely to emulate and express negativity themselves. Our results are particularly noteworthy because they provide one of the few direct tests of whether online negativity can lead to more anxious and angry comments from petition readers. Our study compliments previous work by experimentally demonstrating that the tone of readers’ comments does, in fact, affect the tone of online comments.
Second, this study furthers our knowledge about how different types of negative emotions prompted by online negativity correspond to different behavioral responses. In our study, these comments produced an abundance of two related, but distinct emotions—anger and anxiety. This research supports the claim that there are important distinctions between the two (Valentino et al., 2008), specifically their consequences on sociopolitical behavior. When participants were exposed to negative language and became angry, their comments were hostile, finger pointing. One commenter saying the treatment of refugees by Russian authorities was “disgusting.” Another calling Putin a “jackass.” Conversely, when negative-toned petitions produced anxiety in participants, comments reflected a heightened level of concern: one commenter writing “Those poor kids! They need to be helped!”
However, when asked if they would take action, anger and anxiety were correlated with different levels of engagement. We did find some support for our expectations that anger would be negatively related to a desire to learn more about the topic (i.e. information-seeking). This is related to previous literature that suggests when people experience anger, they are likely to feel set in their attitudes and opinions, and therefore see no need to gather additional information. Anxiety, on the other hand, was significantly related to information-seeking. When users are put in a situation in which they are fearful or concerned, they want to do something to alleviate that distress. Information-seeking is a classic uncertainty reduction strategy, and clicking a mouse to do so requires minimal effort.
Unexpectedly, however, anger was not significantly related to forwarding and negatively related to petition signing, while anxiety was positively related to petition signing and forwarding. These results are at odds with previous scholarship that argues that anger leads to proactive behavior, while anxiety leads to more passivity.
What could explain these unexpected findings? The most likely explanation is that participants did not see the petition as being directly related to the self. The majority of the literature examining the divergent effects of anger and anxiety has focused on a political context in which citizens perceived outcomes would directly affect the self (e.g. Valentino et al., 2008, 2011). Other literature, however, has demonstrated that the experience of anger is highly dependent on whether a person perceives a situation to be self-relevant. A situation need not directly affect a person, per se, as long as they perceive that the victims of a crisis share a similar group membership, thus, making their experiences self-relevant (Gordijn et al., 2001; Yzerbyt et al., 2003). When group-based anger is experienced, people become more focused on addressing their emotional distress, often through means of directing hostility toward the out-group (Van Zomeren et al., 2008). The primary determinant of whether anger leads to action, therefore, is the strength of group identification, rather than an assessment on how helpful specific actions will be in solving a problem.
Anxiety, on the other hand, has less to do with the perception of a situation being relevant to the self. Anxiety is often driven by the feeling that a certain problem or situation needs to be fixed (Gordijn et al., 2001; Yzerbyt et al., 2003). Thus, actions related to anxiety are more dependent upon whether a person perceives said actions will help address the relevant situation (Van Zomeren et al., 2008). These problem-focused coping approaches are influenced by the degree to which a person believes they have an ability to affect the situation.
These distinctions appear to be particularly relevant to the context of the petitions presented to participants in our study. The primary focus of this petition was Syrian refugees and the Russian government—topics that should not be directly related to many Americans’ self-concept. Russia is far away, and while the country and its government might solicit angry commentary that might be all. Anger that is not connected to personal involvement is more likely to dissipate quickly. Anxiety, which can be experienced without personal involvement, was exemplified through a heightened concern for the Syrian refugees and the crisis itself; signing the petition and spreading the word to others may provide the feeling that participant actions are contributory.
These findings may be particularly important for understanding how negative emotions affect sociopolitical behavior, given that the majority of the literature has largely looked at self-relevant contexts (i.e. American politics). Yet, many of the social justice causes taken up online are for social groups and/or victims only distally related to the hordes of those who end up joining the cause.
In addition to the tone of the petition, we examined how an external agent or situational factor impacted response and subsequent action—in this case the Russian authorities versus the refugee situation. As expected, when participants found a common enemy, they expressed a higher level of anger, especially when that enemy was described using hostile language. Conversely, when a family of Syrian refugees was the focus of the petition, participants’ comments were less hostile. As demonstrated by the total effects of target shown in Table 1, framing the petition around the Syrian refugees led to increases in petition-related behavior. These effects may be partially explained by the process through which focusing on social others in need, rather than a culpable agent, serves to reduce anger, which is negatively related to petition-related behavior.
In this regard, our study corroborates previous work that has lamented the deleterious effects of hostile and uncivil discourse online (e.g. Anderson et al., 2014; Papacharissi, 2004; Sobieraj and Berry, 2011). Consistent with this literature, we do find that negativity online can breed anger, which ultimately appears to stand in the way of prosocial behavior.
At the same time, our results also support our claim that while negativity may beget more negativity, its manifestations are distinct. Our research supports previous research that has highlighted the destructive effects of online negativity, yet only when negativity leads to anger. We would be remiss not to draw attention to our findings that negativity leading to anxiety is, paradoxically, positive for prosocial behavior. What we have found is that anger and blame toward an external agent—finger pointing, name-calling—ultimately is negatively related to petition-related action; despite being commonplace, it appears to be largely ineffective in promoting action on the behalf of others.
Thus, it is not negativity per se that appears disconcerting but specifically anger and hostility. Indeed, when looking at the total effects of negativity in Table 1, we see a moderately significant relationship to petition forwarding but non-significant relationships to information-seeking and petition signing. These results suggest that a negative tone does not have a particularly pronounced effect on petition-related behavior—most likely because the effects of anger and anxiety largely cancel each other out.
Today, there are countless examples of citizens expressing both concern and anger about the treatment of social others. Our study suggests that those wishing to mobilize citizens to help are better off inciting distress about the situation. Given that there is no evidence to suggest anger increases prosocial behavior, when the issue at hand is not likely to be related to petition reader’s self-concept, there appears to be little benefit to focusing on blaming a culpable agent and/or stirring up anger.
Limitations
There are a few limitations to be noted within the research. First, while we highlight incivility as contextual, context here is also a limitation. We have offered respondents one petition with a very specific theme—clearly, a different petition may lead to different results. Furthermore, the petitions directed toward Russia and those of a family of refugees were naturally constructed in slightly different ways. Negativity directed toward Russia will more easily take the form of anger than that of a family from a war torn country. However, great care was taken to include similar types of emotions in the Russia and refugee conditions. For example, the negative refugee manipulation also included angry comments, such as “… Not some of those other Islam people that just want to MURDER EVERYONE.”
Additionally, we argue that the primary reason that our petition produced unexpected results was that it focused on a distal situation with little self-relevance to petition readers. But there may have been other context-specific factors in our petition that could have also accounted for our results. More research is needed to fully examine how self-relevance interacts with negative emotions to affect petition-related behavior.
That being said, our findings are also potentially important for those looking to use online resources to achieve social justice or help bring about social change. While previous research has largely focused on negative emotions for self-relevant topics, many of the petitions posted to Change.org, are not directly related to users’ self-concept. Two of Change.org’s current featured petitions, for example, are to ban violent athletes from the National Collegiate Athletic Association (NCAA) or for President Obama to grant citizenship to a foreign born child of US parents—neither of which has direct implications on the hundreds of thousands of signatories. Such petitions tend to point fingers at others or demand unrealistic results, prompting similar comments. These petitions may, again, be better served reigning in the anger when the situation is removed from petition readers.
Additionally (while not obligatory to participation), we encouraged participants to comment, which may be abnormal behavior for some. Thus, as is typically the case with experimental research, there was a degree of artificiality in participant behavior. Despite this, commenting in these types of online spaces is quite popular and the results support the growing body of literature on online commenting and negativity (Papacharissi, 2004). However, we are inferring emotion from online commenting, which is an assumption that must be noted. While looking at comments might not give exact insight into users’ emotional states, it does lend support to our hypothesis that language tone contributes to subsequent behavior. We have good cause to believe that even if people are just emulating others, this might be consistent with real-world situations, as theories on herd mentality and social contagion argue. When individuals post comments, whether they truly believe them or not, they may feel committed to those feelings because they are aware others have observed those sentiments (McLaughlin et al., 2016). Even if they did not initially believe in their own comments, a person may adopt those negative emotions in order to allay cognitive dissonance: thus, begins the snowball effect of commenting.
Another potential issue is that 47 of the 400 participants in our study choose not to write a comment (11.7%). It is possible that those who chose not to comment were simply not experiencing much of an emotional reaction. By not including these participants in our analysis, we may be over-representing participants who experienced an emotional reaction, potentially leading to Type I error. We took several steps to allay concerns over not including non-responders in our analysis.
First, we examined whether choosing not to comment varied significantly across experimental conditions. The slight variation between conditions (Russia Negative: 14/102 [13.7%]; Russia Positive: 9/100 [9.0%]; Refugee Negative: 9/99 [9.1%]; Refugee Neutral: 6/99 [6.1%]) was not significant, χ2(3) = 3.53, p = .317.
Next, we imputed neutral emotion scores for those who did not write a comment. Our coding scheme used the scale mid-point (3) to score participants who displayed no emotion either way (e.g. “I have a very neutral feeling as of now. I would just remain silent”). After imputing scores of 3 for non-commenters, we re-ran our analysis. The pattern of results stayed the same, but some previous non-significant results became significant. Table 3 shows these results. Given these results, we concluded that the missing data did not present a serious concern.
Results with 3’s added for missing data.
p < .001; **p < .01; *p < .05; #p < .10.
Another limitation that resulted from the artificial context necessitated by the experimental design was that participants were not provided the opportunity to actually seek out additional information, or sign or forward the petition. Additionally, these behavior outcome measures relied on single-item indices, which may not accurately reflect the complex ways people engage in petition-related behavior.
Despite these limitations, our research can be seen as an important compliment to previous research that has used alternative methods but not directly tested the factors that may lead to negativity online. Previous literature has primarily focused on analyzing extant online comments or employing surveys. From an operational perspective, giving participants the freedom to comment in a pseudo real-world environment is a novel step toward understanding the factors that cause people to express negative emotions online and one we hope to see repeated in this and other contexts. Additionally, by closely emulating an actual online petition, our study did maintain a high level of ecological validity. Mimicking these online spaces through experimental methods can help further our understanding of other discursive behavior—not only with online petitions but also in other rhetorical spaces.
Conclusion
This study furthers our understanding of the growing interconnectivity between civic engagement and the online world. Previous literature studying the divergent effects of distinct negative emotions has focused mostly on politics, but everyone has a political opinion these days. Expanding our reach into user-generated platforms, devoid of traditional gate-keeping techniques, offers distinct perspectives. As we have argued, these channels fuel their own flames, and these fires have legitimate consequences for an informed and engaged citizenry. In a perfect world, social causes would encourage a democratic dialogue, but as the literature proposes this is not the norm. Rather, blog posts, comments on Twitter and Facebook, and online petitions are filled with hostility and anxiety. Contextually and theoretically, the snowball of one another’s emotions is directly impacting how people relate to social issues. This raises concerns: when those online conversations beget negative emotions, it can be counterproductive or it can be fodder for anxiety. While a proposition that online negativity can be a good thing sounds contrary, it is anger specifically that may not produce action. To genuinely produce social change, directing users toward concern and a desire to learn more through anxiety might be the best course of action.
Footnotes
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Author biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
