To comment or not to comment?: How virality, arousal level, and commenting behavior on YouTube videos affect civic behavioral intentions

Computers in Human Behavior 51 (2015) 520–531

Contents lists available at ScienceDirect

Computers in Human Behavior journal homepage: www.elsevier.com/locate/comphumbeh

To comment or not to comment?: How virality, arousal level, and commenting behavior on YouTube videos affect civic behavioral intentions Saleem Alhabash a,⇑, Jong-hwan Baek b, Carie Cunningham c, Amy Hagerstrom d a Department of Advertising + Public Relations, Department of Media and Information, College of Communication Arts & Sciences, Michigan State University, 404 Wilson Road, Communication Arts & Sciences Building, Rm. 313, East Lansing, MI 48824-1212, USA b Department of Media and Information, College of Communication Arts & Sciences, Michigan State University, 404 Wilson Road, Communication Arts & Sciences Building, Rm. 409, East Lansing, MI 48824-1212, USA c School of Journalism, College of Communication Arts & Sciences, Michigan State University, 404 Wilson Road, Communication Arts & Sciences Building, Rm. 305, East Lansing, MI 48824-1212, USA d School of Communications, Grand Valley State University, 126 LSH, Allendale, MI 49401, USA

a r t i c l e

i n f o

Article history: Available online 5 June 2015 Keywords: Cyberbullying YouTube Civic behavioral intentions Viral behavioral intentions Persuasion LC4MP

a b s t r a c t An experiment investigated the effects of commenting behavior, virality, and arousal level on anti-cyberbullying civic behavioral intentions. Participants (N = 98) were exposed to cyberbullying-related YouTube videos that varied in arousal level (low vs. high), number of views (low vs. high), and commenting behavior where they either commented on the video or did not comment after watching it. Participants expressed greater Civic Behavioral Intentions (CBI) upon exposure to highly than lowly arousing videos. Additionally, they expressed greater CBI when instructed to comment on highly arousing videos with high than low views, while those who did not comment on highly arousing videos expressed greater CBI upon exposure to videos with low than high views. As for lowly arousing videos, participants who were instructed to comment expressed greater CBI when the video had low than high views, while those who did not comment did not differ in CBI as a function of the number of views. Viral behavioral intentions (VBI) were the strongest predictors of CBI with degrees that varied as a function of commenting behavior, virality, arousal level, and the interactions among them. Results are discussed within the framework of the relationship between online engagement and offline civic action. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Bullying and cyberbullying have been characterized as phenomena that prevail in schools. Between a third and a half of adolescents report having been cyberbullied, and 80% of them observe others being cyberbullied (BullyingStatistics.org, 2013; Hinduja & Patchin, 2010; i-SAFE, 2004; Lenhart, 2007; National Crime Prevention Council, 2010; Patchin, 2010; Statistics Brain, 2013; Webster, 2010). A recent study found that about one-fifth of college students have been bullied and cyberbullied and about 70% of students observed others being cyberbullied, indicating the problem’s prevalence on college campuses (Alhabash, McAlister, Hagerstrom, Quilliam, Rifon, & Richards, 2013). Cyberbullying refers to the use of information communication technologies (ICTs) to perform repeated intentional acts of direct

⇑ Corresponding author. Tel.: +1 517 432 2178. E-mail address: [email protected] (S. Alhabash). http://dx.doi.org/10.1016/j.chb.2015.05.036 0747-5632/Ó 2015 Elsevier Ltd. All rights reserved.

(e.g., repeated direct attacks) or indirect (e.g., posting harmful messages) aggression that reflect a power imbalance between the offender and the victim (Langos, 2012; Patchin & Hinduja, 2006; Wade & Beran, 2011). Compared to offline bullying, cyberbullying is often facilitated by anonymity or the perception of anonymity, has less parental oversight, lacks time and space restrictions, is accessible by large audiences, is maintained online, and has severe consequences (Patchin & Hinduja, 2006; Raskauskas & Stoltz, 2007; Strom & Strom, 2005). While the prevalence of offline bullying declines with age, cyberbullying happens among older youth. Because cyberbullying allows one to be anonymous and, unlike offline bullying that is space- and time-constrained, it lives longer in the online sphere, resulting in more devastating consequences in relation to depression, psychosomatic problems, lowered self-esteem, suicidal behavior, and poor school performance; such consequences are often underestimated by bullies, parents, school administrators, and victims themselves (Kiriakidis & Kavoura, 2010; Nunnally, 1967; Patchin & Hinduja, 2006; Strom & Strom, 2005; Wade & Beran, 2011).

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

While much of the research on offline bullying and cyberbullying focuses on middle and high school students (e.g., Kiriakidis & Kavoura, 2010; Nunnally, 1967; Patchin & Hinduja, 2006; Strom & Strom, 2005; Wade & Beran, 2011), little research explores this phenomenon among college students, and much more scarcity is found in relation to intervention programs addressing this issue. Past research shows that between two to three in every 10 students report being cyberbullied during college, with about one in every 10 students reporting they have cyberbullied others (Kowalski, Giumetti, Schroeder, & Reese, 2012; MacDonald & Roberts-Pittman, 2010). Zalaquett and Chatters (2014) found high associations between offline bullying and cyberbullying victimization in high school and college; where those bullied in college also reported being bullied during high school. Offline bullying and cyberbullying among college students is associated with negative psychological and health outcomes. College students who are victims of cyberbullying report greater levels of anxiety; depression; suicidal ideation, planning, and attempts; lower self-esteem; poorer health indicators; and poorer academic performance compared to non-victims (Kowalski & Limber, 2013; Schenk & Fremouw, 2012). A growing number of offline bullying and cyberbullying victims and sympathizers have resorted to social media to discuss such incidents. Last year, a YouTube video by Amanda Todd became viral as she talked about being bullied and cyberbullied leading up to her suicide. YouTube and other social media are platforms where cyberbullying takes place, yet they can also be used to raise awareness and change users’ attitudes and behaviors in relation to cyberbullying. Because college students who are cyberbullied often cope with the effects of cyberbullying on their own without utility of community resources, understanding how to effectively advocate against cyberbullying through social media platforms is an essential intervention strategy. Moreover, while dealing with psychological and physical health consequences of cyberbullying is essential to the individual, community awareness and civic actions are needed to realize social change regarding the prevalence of bullying and cyberbullying. The current study investigates the effects of user-generated YouTube videos on anti-bullying/cyberbullying Civic Behavioral Intentions (CBI), as a means for reducing the prevalence and effects of offline bullying/cyberbullying. Civic behavioral intentions are defined as intentions to perform civic actions geared toward raising awareness and affecting policy changes in relation to cyberbullying (see Appendix A). Using persuasion models and theories of emotional and excitation transfer, the study explores the effects of the intensity of emotional tone (level of arousal), video virality (number of views), and commenting behavior on CBI.

2. Literature review 2.1. YouTube and virality YouTube is the third most-visited website in the United State and worldwide, with over one billion monthly visitors who watch more than six billion hours of video monthly, upload 100 h of new video every minute, and are highly engaged in liking, sharing, and commenting on videos on YouTube and other social networking sites (Alexa, 2013; Cheng, Liu, & Dale, 2013; Glenn, 2013; Thelwall, Sud, & Vis; 2012; YouTube, 2013). Young adults (18– 34 years old) are the highest adopters and most frequent users of YouTube, who comprise two-thirds of YouTube and watch YouTube videos more than any cable TV channel (Glenn, 2013; Lenhart & Madden, 2007; Purcell, 2013; YouTube, 2013). YouTube is only but one platform for video sharing and viewing. On websites like YouTube, users can upload videos, interact with

521

video content by sharing videos with their online and offline social networks, like, dislike, and comment on videos. These online behaviors can be understood through the framework of virality. While the number of views is the most common indicator of a YouTube video’s virality, Alhabash and McAlister (2014, p. 3) argue for a tripartite approach to defining virality: Affective evaluation is defined as the explicit emotional responses visible to other users (e.g., likes, dislikes). Viral reach refers to both sharing and viewership of content (e.g., views, shares). Message deliberation refers to discussions and comments on online content, which can also entail affective evaluation. Past research argued that emotional engagement plays a critical role in driving virality (e.g., Hagerstrom, Alhabash, & Kononova, 2014; Eckler & Bolls, 2011; Kirby, 2004; Phelps, Lewis, & Mobilio, 2004). While positive messages have a greater chance of virality than negative ones, emotional intensity in both negative and positive messages increases the likelihood of virality as well as other persuasion outcomes like attitude and behavior changes (Hagerstrom et al., 2014; Eckler & Bolls, 2011; Kirby, 2004; Phelps et al., 2004). In the following sections, we provide three theoretical explanations for how and with what effects online content goes viral, using theories of limited capacity, excitation transfer, and social norms. 2.2. Virality: a limited capacity take Berger and Milkman (2011) found that the presence of intense emotions in New York Times news articles (e.g., anxiousness, awe, anger, and surprise) positively correlated with the number of times they were shared. This is supported by other studies, where emotionality of content predicted virality (Hagerstrom et al., 2014; Eckler & Bolls, 2011). The question, therein, lies in attempting to explain why emotional content, both positive and negative, has a greater likelihood of virality than neutral content. Eckler and Bolls (2011) suggested that forwarding intentions – which we term as viral behavioral intentions – are sensitive to activation of appetitive and aversive motivational systems. Appetitive motivation deals with an individual’s motivation to approach external stimuli (e.g., approaching food and sex), while aversive motivation refers to withdrawal of resources to maintain survival and escape danger (e.g., running away from a roaring lion) (Lang, 2000, 2006). The two motivational systems work in parallel to guide our central nervous system’s responses to external stimuli (Lang, 2000, 2006). The limited capacity model of mediated motivated message processing (LC4MP; Lang, 2000, 2006) builds on past research in cognitive psychology to explain how humans respond to mediated communication. LC4MP rests on five major assumptions (Lang, 2000, 2006). First, Lang argues that humans are information processors with a limited cognitive capacity. Second, information processing is pertinent to activation of the appetitive and/or aversive motivational systems. Third, humans receive media messages in different formats (words, still pictures, moving pictures, etc.) through sensory channels of message reception (eyes, ears, touch). Fourth, information processing takes place over time (as little as seconds and milliseconds). Finally, humans interact with the communication message in multiple ways. These assumptions provide an understanding of how selective we, as human beings, are when we are faced with external stimuli, thus the description of humans as cognitive misers (Fiske & Taylor, 1984). We employ shortcuts to information processing tasks that minimize the use of cognitive resources. Environmental factors trigger uncontrolled, automatic information processing due to

522

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

motivational activation, where its direction and intensity guide our cognitive, affective, and behavioral responses to external stimuli (Bolls, 2010; Cacioppo, Gardner, & Berntson, 1999). Our responses and actions follow activation of the appetitive (approach, e.g. approaching food and sex) and/or aversive (withdraw, e.g., running away from danger) motivational systems (Lang, 2000, 2006). Based on this logic, intensely emotional content triggers greater motivational activation and subsequent cognitive, affective, and behavioral responses. Therefore, the increased magnitude of activation (as opposed to responses to mildly intense content) increases the chances of behavioral responses, starting with viral behaviors carried online, such as liking, sharing, or commenting on an online piece of content. 2.3. Virality: a form of excitation transfer Another explanation for virality stems from the excitation transfer approach to understanding emotions. Excitation transfer posits that residual excitation from a previous stimulus may escalate a subsequent emotional state (Zillmann, 1971). After experiencing physiological arousal or excitation during exposure, residual excitement does not decline immediately, and may get intensified with a subsequent exposure, often resulting in an intensified emotional state or response. Excitation transfer has been examined in the context of exposure to film, television, and music, among others (Cantor & Zillmann, 1973; Hansen & Hansen, 1990; Zillmann, Mody, & Cantor, 1974). Zillmann (1971) exposed participants to movie clips (e.g., educational movie, drama, sexual movie) and asked them to give an electric shock to confederates who answered a question wrong, where shock intensity was highest after viewing sexual clips. Cantor and Zillmann (1973) showed evidence of excitation when participants had more intense positive responses to music upon exposure to highly than mildly arousing clips. We argue that viral behaviors (e.g., likes, shares, and comments) are carried out during the window of excitation residual following exposure to online content and are facilitated by ICT affordances (e.g., constant access to social media), ease of use (e.g., habitual performance of viral behaviors), and access (e.g., use of mobile devices to access social media sites). When one sees a video that increases excitation and arousal, he/she reaches a point when the residual excitation needs to be released. Having a keyboard and knowing which buttons to push facilitate excitation transfer. In turn, excitation transfer leads the user to click on the like or share buttons, or even expend more cognitive resources in commenting on content, as well as take actions offline to release excess arousal resulting from message exposure. While arousal is often regarded as an outcome depiction of an individual’s emotional response, mediated communication messages can be structured in a way to be more or less arousing. Generally, arousal refers to the intensity of the emotional response, where the emotional response can be positive and/or negative (Bradley & Lang, 1994). In the current study, we regard emotional arousal as a function of the video itself, in that highly arousing videos are ones with highly intense emotions, and vice versa: lowly arousing videos are those with mild emotions. Due to the nature of the topic of cyberbullying, we are limiting the valence of emotional appeal to negative emotions. Hence, the study primarily deals with lowly and highly arousing negative videos related to cyberbullying. Based on the excitation transfer hypothesis, we argue that an increase in the intensity of the emotional appeal of the video would increase participants’ experienced arousal, which would then transfer into the expression of intentions to engage in civic actions related to cyberbullying. We hypothesize: H1a. Participants will report greater Civic Behavioral Intentions (CBI) upon exposure to highly arousing videos compared to lowly arousing ones.

Thelwall et al. (2012) found that commenting on news and political videos on YouTube tends to be negative rather than positive, thus suggesting that individuals engage in forms of releasing the emotional arousal experienced during message exposure. It is unclear, however, what impact the act of commenting has in relation to performing behaviors related to the message that elicited the commenting in the first place. The current study manipulates commenting behavior, where participants either comment on YouTube videos or not. There are two competing arguments about the effect of commenting behavior on CBI. First, commenting on content online could release the residual excitement faster, which hinders offline behavioral engagement. Second, commenting could prohibit users from resolving their excitement, as it facilitates greater engagement with the topic and, through greater deliberation, excitation might increase, which then increases the chances of offline behaviors. In other words, commenting on a video might induce further elaboration of the message, thus leading to greater involvement with the topic and, therefore, generating greater chances for offline and online behaviors resulting from message exposure. We follow this second line of reasoning to hypothesize: H1b. Participants instructed to comment will report greater CBI than those who were not instructed to comment on cyberbullying videos. 2.4. Virality: a social norm Virality, as a message feature (e.g., the number of views, likes, dislikes, comments), can be regarded as an expression of social norms. Social norms are integral to the theory of planned behavior (Ajzen, 1991), social cognitive theory (Bandura, 1986), and social comparison theory (Festinger, 1954). Cialdini, Reno, and Kallgren (1990) identified two types of social norms: (1) descriptive norms that refer to perceived prevalence of a behavior or attitude in a group or society; and (2) injunctive norms that refer to perceived acceptance of the behavior or attitude. A number of factors have been found to moderate the effects of social norms on behavior. For example, the amount of attention paid to norms can play an important role in the relationship between norms and actual behaviors, as can the clarity of the normative information and believability of disseminated information about norms (Cialdini et al., 1990). In the online world, social norms can be manipulated through message content (e.g., ‘‘90% of college students do not cyberbully one another’’). However, it is also possible to observe an effect of social norms by algorithmic measures of message popularity and virality. Virality of online content (i.e., high likes, shares, and comments) reflects acceptance and prevalence of the advocated behavior/attitude. These metrics of virality are often correlated, where a video or an online messages with a high number of likes, also has high shares and views. However, there is inconsistency in the industry with regard to specifying a single metric for assessing whether a piece of online content is viral or not. For example, AdWeek regards the number of shares as the metric used to assess the virality of an online advertisement (Nudd, 2014), while AdAge (2015) highlights the number of views as a metric for assessing an advertisement’s virality. On YouTube, the number of views is the most prominent metric – visually speaking – to users. The number is displayed in large font right beneath the video (left-side). The number of likes and dislikes are displayed under the number of views in smaller font, and the number of comments is displayed farther down the page, where a user needs to scroll down to the comments section in order to see that number. To this end, in the current study we chose to only manipulate the number of views as a metric of virality, considering its visual prominence in the YouTube environment.

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

523

Therefore, a high number of views, we argue, signifies acceptance of the message by others (high social norm), which would in turn affect the message’s persuasiveness. We hypothesize:

3. Method

H1c. Participants will report greater CBI upon exposure to videos with a high than low number of views. The study further asked:

The study employed a 2 (arousal: high vs. low)  2 (commenting behavior: commented vs. did not comment)  2 (number of views: low vs. high)  2 (repetition) mixed factorial design. Arousal was manipulated within subject, where participants were exposed to videos with low and high levels of arousal. Arousal level refers to the intensity of the video’s negative emotional appeal. For example, a highly arousing video includes intense music coupled with moving narration and/or text, while lowly arousing videos tend to highlight the issue of cyberbullying in a lighter manner. In other words, highly-arousing videos are what is referred to as ‘tear-jerkers,’ while lowly-arousing videos portray the topic with milder emotions. Commenting behavior was manipulated between subjects, where participants were randomly assigned to either comment or not comment after viewing the video. Number of views was manipulated between subjects, where participants were randomly assigned to view the same videos either with a high or low number of views. The number of views was placed beneath the video, simulating a YouTube page. Low views ranged from 243 to 298 views (M = 270.33, SD = 20.82), while high views ranged between 7,178,983 and 7,342,765 (M = 7,254,608, SD = 74.123). The choice of the number of views and its range was carried out at random with the purpose of distinguishing low and high number of views in each of the conditions. The arousal factor was represented with two repetitions at each level, where participants viewed four videos in total (2 highly arousing and 2 lowly arousing) in one of the following four conditions: (1) high views and commenting instructions; (2) low views and commenting instructions; (3) high view and no commenting instructions; and (4) low views and no commenting instructions. Participants (N = 95) were recruited through a subject pool at a large U.S. Midwestern university, and were rewarded with extra credit. The sample was evenly split in terms of gender (females = 48.42%). Participants had an average age of 21 (M = 20.65, SD = 1.70) and were mostly freshmen and sophomores (61.05%). As for the ethnic make-up of the sample, the majority reported being White/Caucasian (55.79%), followed by Asian (28.42%), African American (11.58%), Native Hawaiian or other Pacific Islander (1.05%), and other ethnicities (2.11%; 1.05% declined to report their ethnicity).

RQ1. How will arousal level, commenting behavior, and number of views interact in affecting CBI? 2.5. Predicting Civic Behavioral Intentions (CBI) Dual processing models like the elaboration likelihood model (ELM; Petty & Cacioppo, 1986) and the heuristic-systematic model (HSM; Chaiken, 1980; Chaiken & Troppe, 1999) argue that information processing of persuasive arguments follows activation of central route (systematic) or peripheral route (heuristic) to processing. The two models imply that there’s a direct relationship between attitude formation and change, and ultimately behavior change. Contrary to ELM and HSM, the theory of planned behavior (TPB), derived from the theory of reasoned action (TRA; Ajzen, 1991; Ajzen & Fishbein, 1980), suggests attitudes are not directly related to behaviors, but rather this relationship is mediated by behavioral intentions (Ajzen, 1991; Ajzen & Fishbein, 1980). Behavioral intentions, defined as an individual’s desire to perform the behavior, are affected by the individual’s attitude toward the behavior, subjective/social norms, and perceived behavioral control (Ajzen, 1991; Ajzen & Fishbein, 1980). In the context of online communication, individuals are faced with a plethora of messages, which adds another layer of complexity to understanding the relationship between attitudes and behaviors. Thus, the importance of attitudes decreases at the expense of viral behaviors. In other words, because Internet users are exposed to a lot of information, they might develop favorable attitudes toward an online message, which can lead to viral behaviors, yet these favorable attitudes are not sufficient to result in offline behavioral changes. In contrast, viral behaviors elevate the message’s importance, resemble personal endorsement of its arguments and, therefore, can more strongly predict offline behaviors. Attitudes are defined as affective, cognitive, and behavioral evaluations of an object or a person (Maio & Haddock, 2007). In the current study, we differentiate between attitudes toward the video (AV) and attitudes toward the issue (AI). AV are analogous to attitudes toward the advertisement (Choi, Miracle, & Biocca, 2001) and refer to participants’ evaluations of the anti-cyberbullying video. AI are similar to attitudes toward the brand (Choi et al., 2001) and refer to an individual’s evaluation of the issue presented in the video; in this case, anti-cyberbullying. Finally, we define viral behavioral intentions as the desire to enact online behaviors that contribute to a message’s virality, such as pressing the like button, sharing the video, and commenting on it. We hypothesize: H2. Compared to attitudes toward the video (AV) and attitudes toward the issue (AI), viral behavioral intentions (VBI) will be the strongest predictor of CBI. We also explore how attitudinal variables and VBI predict offline behavioral intentions (CBI) as a function of arousal level, commenting behavior, and number of views. We ask: RQ2. How will AV, AI, and VBI predict CBI as a function of the level of arousal in videos, commenting behavior, number of views, and the interactions among them?

3.1. Study design and participants

3.2. Operational measures 3.2.1. Attitudes toward the video (AV) and attitudes toward the issue (AI) Upon viewing each video, participants indicated their attitudes toward video and their anti-bullying/cyberbullying attitudes (i.e., attitudes toward the issue) using three seven-point semantic differential scales: bad/good, negative/positive, unfavorable/favorable (Choi et al., 2001; Coulter & Punj, 2007; MacKenzie & Lutz, 1989; Muehling, 1987). Upon satisfactory factor and reliability analyses (see Appendix A), AV and AI were computed for each message.

3.2.2. Viral behavioral intentions (VBI) Participants indicated their agreement/disagreement with 11 statements using seven-point scales anchored with ‘‘Strongly Disagree’’ and ‘‘Strongly Agree’’ (Alhabash et al., 2013; Eckler & Bolls, 2011). Upon satisfactory factor and reliability analyses (see Appendix A), VBI was computed for each message.

524

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

Fig. 1. Sample experimental stimuli.

3.2.3. Civic Behavioral Intentions (CBI) Participants rated their agreement/disagreement to seven items using seven-point scales anchored with ‘‘Strongly Disagree’’ and ‘‘Strongly Agree,’’ which were created specifically for this study. Upon satisfactory factor and reliability analyses (see Appendix A), CBI was computed for each message. 3.3. Stimuli Stimuli for this experiment were selected from a pool of 12 videos that were pretested with an independent sample of participants (N = 11). Pretest participants evaluated the arousal level, positivity and negativity of each video (using nine-point semantic differential scales; Bradley & Lang, 1994). The selected videos comprised of two videos that rated highest on arousal and other two that rated lowest on arousal. Selected videos were comparable on positivity and negativity and had an average length of 168 s. The videos primarily focused on raising awareness about the issue of cyberbullying. All videos were user-generated, and presented a clear anti-cyberbullying sentiment (see Appendix B for a description of each video). The original video titles were maintained in the experiment with minor modifications to enhance ecological validity. During the experiment, participants saw four videos with a background identical in look and feel to YouTube (see Fig. 1). Participants either saw the videos paired with low or high views. 3.4. Procedure Participants completed the study in groups of 1–20 in a computer lab, where they used a desktop computer equipped with a 1700 monitor, keyboard, mouse, and headphones to complete the

study via Qualtrics.com. Upon providing consent, they completed a psychological task and answered questions about their social media use and their experiences with offline bullying and cyberbullying. Participants were then randomly assigned to one of the four conditions (commenting/high views; commenting/low views; no commenting/high views; and no commenting/low views), and were made to believe that videos have been embedded in the experiment directly from YouTube. After viewing each video, they answered questions related to the predictors and dependent variable. Finally, participants provided demographic information, were debriefed, provided with extra credit receipts, and dismissed. 4. Results 4.1. Descriptive results The majority of participants reported that they watch videos on YouTube (98.00%) and have active user accounts on YouTube (74.48%). Participants reported they have been bullied (24.48%), cyberbullied (21.43%), have cyberbullied someone else (19.39%), and observed others being cyberbullied (58.16%) in the past couple of months.1 Social networking sites (SNSs) and text messaging were the platforms where participants were most frequently cyberbullied (SNSs = 39.80%; texting = 21.42%), cyberbullied others 1 Independent samples t-test with cyberbullying experiences (i.e., being cyberbullied, cyberbullying others, and observing others being cyberbullied) as independent variables and AV, AI, VBI, and CBI as dependent variables were conducted and showed no significant differences between those who reported experiencing cyberbullying as a victim, perpetrator, or observer, and those who did not report such experiences. Therefore, these variables were excluded from further analyses as potential covariates.

525

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

(SNSs = 29.59%; texting = 19.39%), and observed others being cyberbullied (SNSs = 71.4%; texting = 21.42%).

4.2. Effects of arousal level, commenting behavior, and virality The first set of hypotheses predicted significant main effects for arousal level (H1a), commenting behavior (H1b), and virality (H1c) on CBI. Additionally, RQ1 asked about the effect of the interaction among arousal level, commenting behavior, and virality on CBI. Data for CBI were submitted to a 2 (arousal level)  2 (commenting behavior)  2 (number of views)  2 (repetition) ANOVA with repeated measures on arousal level and repetition. In support of H1a, results indicated that participants expressed greater CBI upon exposure to highly arousing videos (M = 4.25, SD = 1.69) than lowly arousing videos (M = 4.02, SD = 1.71), F(1, 94) = 7.28, p < .01, g2p = .07. The main effects of commenting behavior, F(1, 94) = .001, ns, and virality, F(1, 94) = 3.36, ns, were not significant. Based on this, H1b and H1c were not supported. Finally, the three-way interaction among arousal level, commenting behavior, and virality was significant, F(1, 94) = 7.22, p < .01, g2p = .07 (see Fig. 2). With regard to highly arousing videos, participants who were instructed to comment on videos expressed greater CBI upon exposure to videos with high views (M = 4.34, SD 1.58) compared to those with low views (M = 4.27, SD = 1.80), while participants who did not comment on videos expressed greater CBI after seeing videos with low views (M = 4.38, SD = 1.56) compared to videos with high views (M = 3.99, SD = 1.89). On the other hand, with regard to lowly arousing videos, participants who commented on the videos expressed greater CBI after seeing videos with low views (M = 4.25, SD = 1.76) than high views (M = 3.67, SD = 1.55), while those who did not comment on videos expressed somewhat similar CBI upon exposure to videos with low views (M = 4.17, SD = 1.57) and high views (M = 4.01, SD = 1.97).

4.3. Predicting CBI To test H2 and answer RQ2, we ran identical multiple linear regression models with AV, AI, and VBI as predictors, and CBI as a criterion. First the model was run with all messages (H2). Subsequent results showcase the same regression model as a function of arousal level, commenting behavior, and the number of views. All models were significant, and across all models, AV and AI did not significantly predict CBI and, thus, the following report of regression results focuses on the strength of association between VBI and CBI.

4.3.1. All messages H1b predicted that VBI would be the strongest predictor of CBI compared to AV and AI. The regression model for all messages combined was significant and fully explained 54% of the variance in CBI. VBI was the strongest predictor of CBI (b = .75, p < .001), thus supporting H2. Both AV and AI were not significant predictors of CBI (see Table 1).

4.3.2. Arousal level The explanatory power of the regression model was slightly higher for highly arousing (54%) than lowly arousing (50%) videos. The strength of association between VBI and CBI was also higher for highly arousing (b = .74, p < .001) than lowly arousing (b = .71, p < .001) videos (see Table 1).

4.3.3. Commenting behavior The regression model for participants who did not comment on the videos had a larger explanatory power (58%) compared to the model for those who commented on the video (49%). VBI had a stronger association with CBI in the no commenting (b = .80, p < .001) compared to the commenting (b = .70, p < .001) condition. The strength of VBI–CBI association was higher for lowly (b = .79, p < .001) than highly (b = .76, p < .001) arousing videos when participants did not comment on the videos. However, for participants in the commenting condition, the VBI–CBI association was stronger upon exposure to highly arousing videos (b = .75, p < .001) than lowly arousing ones (b = .61, p < .001; see Table 2).

Table 1 Regression results for the relationship between attitudes toward the video, attitudes toward the issue, and viral behavioral intentions, and civic behavioral intentions, and by arousal level.

Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI) Model statistics R (R2adj) df F Notes.  p 6 .10, ⁄p 6 .05, *** p 6 .001.

⁄⁄

p 6 .01.

Fig. 2. Effect of three-way interaction for commenting behavior.

All messages b (SE)

Low arousal b (SE)

High arousal b (SE)

.01 (.10) .02 (.06) .75 (.08)***

.04 (.10) .08 (.06) .71 (.09)***

.05 (.09) .06 (.04) .74 (.08)***

.75 (.54) 3, 94 39.292***

.72 (.50) 3, 94 33.17***

.74 (.54) 3, 94 38.41***

526

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

Table 2 Regression results for the relationship between attitudes toward the video, attitudes toward the issue, and viral behavioral intentions, and civic behavioral intentions, and by commenting behavior and arousal level. All messages b (SE)

Low arousal b (SE)

High arousal b (SE)

Commenting Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.08 (.13) .01 (.09) .70 (.12)***

.14 (.15) .08 (.10) .61 (.14)***

.002 (.12) .07 (.09) .75 (.11)***

Model statistics R (R2adj) df F

.72 (.49) 3, 44 15.97***

.68 (.42) 3, 44 12.31***

.75 (.54) 3, 44 19.17***

No commenting Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.07 (.14) .03 (.08) .80 (.11)***

.03 (.14) .10 (.08) .79 (.11)***

.08 (.13) .05 (.08) .76 (.12)***

Model statistics R (R2adj) df F

.78 (.58) 3, 46 23.69***

.76 (.55) 3, 46 21.29***

.75 (.53) 3, 46 19.62***

Notes.  p 6 .10, ⁄p 6 .05, *** p 6 .001.

⁄⁄

p 6 .01.

Table 3 Regression results for the relationship between attitudes toward the video, attitudes toward the issue, and viral behavioral intentions, and civic behavioral intentions, and by virality and arousal level. All messages b (SE)

Low arousal b (SE)

High arousal b (SE)

Low views Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.06 (.17) .13 (.11) .70 (.12)***

.18 (.17) .25 (.11)  .64 (.12)***

.04 (.15) .03 (.10) .71 (.13)***

Model statistics R (R2adj) df F

.71 (.47) 3, 44 15.02***

.70 (.46) 3, 44 14.41***

.71 (.47) 3, 44 14.96***

High views Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.005 (.13) .05 (.08) .77 (.12)***

.03 (.14) .002 (.09) .76 (.13)***

.03 (.12) .10 (.08) .74 (.11)***

Model statistics R (R2adj) df F

.78 (.59) 3, 46 24.24***

.75 (.54) 3, 46 19.94***

.77 (.57) 3, 46 22.75***

Notes. ⁄p 6 .05,   p 6 .10. *** p 6 .001.

⁄⁄

p 6 .01.

4.3.4. Number of views The regression model for the high views condition had a greater explanatory power (59%) compared to the low views condition (47%); whereas, the VBI–CBI association was stronger in the high views (b = .77, p < .001) compared to the low views (b = .71, p < .001) condition. The VBI–CBI association was stronger for highly arousing videos (b = .71, p < .001) than lowly arousing videos (b = .64, p < .001) in the low views conditions, where it was slightly higher for lowly arousing videos (b = .76, p < .001) than highly arousing videos (b = .74, p < .001) in the high views condition (see Table 3). 4.3.5. Commenting behavior  virality Table 4 summarizes the regression models for each of the four conditions in relation to all messages as lowly and highly arousing

videos. Results showed that the model with highest explanatory power was for the no commenting and high views condition (66%), followed by commenting with low views (53%), commenting with high views (44%), and no commenting with low views (40%), respectively. The VBI–CBI association followed a similar trend in terms of beta weights. Regarding the arousal level of the video, results showed when participants commented on the videos, the VBI–CBI relationship was stronger for highly arousing videos in both conditions of low (b = .84, p < .001) and high (b = .69, p < .001) views compared to lowly arousing videos (low view: b = .64, p < .001; high views: b = .64, p < .001). In the two commenting conditions (with low and high views), the model explained greater variance for highly arousing videos (low views: 60%; high views: 45%) compared to the models for lowly arousing videos. The combination of

527

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531 Table 4 Regression results for the relationship between attitudes toward the video, attitudes toward the issue, and viral behavioral intentions, and civic behavioral intentions, and by commenting behavior, virality, and arousal level. All messages b (SE)

Low arousal b (SE)

High arousal b (SE)

Commenting and low views Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.13 (.24) .07 (.18) .75 (.18)***

.25 (.24) .24 (.18) .64 (.19)***

.06 (.20) .133 (.17) .84 (.18)***

Model statistics R (R2adj) df F

.77 (.53) 3, 19 9.30**

.75 (.49) 3, 19 7.91***

.81 (.60) 3, 19 12.09***

Commenting and high views Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.06 (.18) .11 (.12) .67 (.18)***

.06 (.20) .12 (.13) .64 (.20)**

.03 (.17) .122 (.12) .69 (.17)***

Model statistics R (R2adj) df F

.72 (.44) 3, 21 7.33**

.69 (.40) 3, 21 6.31**

.72 (.45) 3, 21 7.50***

No commenting and low views Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.11 (.27) .02 (.15) .70 (.19)***

.06 (.25) .15 (.15) .66 (.18)***

.05 (.22) .01 (.15) .66 (.21) ***

Model statistics R (R2adj) df F

.69 (.40) 3, 21 6.33**

.68 (.39) 3, 21 6.02**

.64 (.33) 3, 21 4.89**

No commenting and high views Attitudes toward the video (AV) Attitudes toward the issue (AI) Viral behavioral intentions (VBI)

.06 (.20) .06 (.13) .88 (.17)***

.13 (.22) .17 (.13) .93 (.17)***

.07 (.18) .07 (.122) .80 (.18)***

Model statistics R (R2adj) df F

.84 (.66) 3, 21 16.52***

.83 (.64) 3, 21 15.27***

.82 (.62) 3, 21 13.86***

Notes.  p 6 .10, ⁄p 6 .05. ** p 6 .01. *** p 6 .001.

commenting and low views yielded 49% of explained variance, and the combination of commenting and high views yielded 40% of explained variance. As for the two conditions when participants did not comment on the video (with low and high views), greater explanatory power was for lowly than highly arousing videos. While the explanatory power for the condition of no commenting and low views was higher for lowly arousing (39%) than highly arousing (33%) videos, the strength of VBI–CBI association was equal (b = .66, p < .001) for both types of videos. The difference between low and high arousing videos was more apparent in the condition when participants did not comment and the videos had high views. In this condition, the regression model for lowly arousing videos explained 64%, while it explained 62% for highly arousing videos. The beta weight for the VBI–CBI relationship was larger for lowly arousing (b = .80, p < .001) than highly arousing (b = .75, p < .001) videos.

5. Discussion 5.1. Summary of findings The current study investigated the effects of commenting behavior, emotional intensity, and virality on the expression of Civic Behavioral Intentions (CBI). Additionally, the study explored the ways in which message and issue evaluations, as well as viral

behavioral intentions predicted CBI. The following section highlights the study’s most intriguing findings. It is clear from the results of this study that emotional engagement in viral content has the potential to drive civic engagement in important societal discourse, such as the growing conversation about cyberbullying. On its own, arousingness of the cyberbullying videos affected CBI. Participants reported higher intentions to engage in offline civic behaviors related to cyberbullying when the videos were highly arousing, while commenting behavior and virality did not have significant main effects on CBI. This suggests that highly arousing videos facilitated greater excitation transfer that could be channeled to the offline environment. The interaction among the arousingness and virality of the video and whether or not participants were instructed to comment on the video speaks to an interesting and important finding. It would seem that two phenomena could occur relative to viral videos on societal issues. When participants view a highly arousing video on an issue they perceive as important, or normative, based on the virality it has already achieved, and they have some personal familiarity by commenting on the video, they are more likely to report intentions to engage in the issue offline. Our findings showed that for videos with high arousal, virality did not matter when participants were instructed to comment on the videos, while videos with low views garnered greater CBI when highly arousing videos had low compared to high views. Conversely, we see that virality matters when the videos had low arousal and

528

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

participants were instructed to comment on the videos with greater CBI reported for videos with low views. Both findings point to a counter-intuitive thought. One would think that greater virality should drive positive persuasive outcomes. However, our findings showed that under certain circumstances, videos with low views could be better at persuading individuals to take up civic action than those with high views. It is plausible that with the abundance of online content, participants are motivated to engage with novel content that is not yet viral. In other words, participants might have been motivated to exercise civic participation for an issue that has not become mainstream. Another plausible explanation is that participants who were exposed to a video with a low number of views might have felt that this important issue – cyberbullying – has not been given the attention it deserves and, consequently, were more inclined to express greater CBI. However, we should note that the discrepancy between the low and high views conditions was rather large; therefore, future research should investigate the effects of virality with variable ranges of numerical indicators of virality (e.g., the number of views, likes, shares, etc.). The results may also be interpreted in a different way. As examined by Thelwall et al. (2012), people tend to be more engaged in topics or comments (e.g., news or politics) that have negative sentiments. Thus, cyberbullying videos that are highly arousing may have a better chance of triggering both negative and positive comments and drive people to be more engaged with content, leading to a greater likelihood of offline behaviors. Coupled with our understanding of cognition and the nature of online communication, the selectivity of online users might make it hard for content with low emotional intensity to float to the top and excite individuals to a level where their motivational system activation drives offline actions. Future research should investigate these relationships relative to individual (e.g., gender, ethnicity, executive functioning) and cultural differences that might guide the ways in which individuals engage in public discourse and public discussion of social issues. Even more interesting, however, is the fact that when participants were told not to comment on highly-arousing videos, they were more likely to look for a way to engage in the issue offline when the video was less viral. This seems to further validate the assumptions of excitation transfer theory: participants who could comment on highly-arousing videos and who had more assurance they were in good company given the virality of the video were more ready to engage in civic actions offline geared toward decreasing the prevalence of offline bullying and cyberbullying. But when participants were not given the opportunity to release their residual excitement, the presence of ‘‘good company’’ was a less important factor in determining their intentions to engage in the issue offline. This finding held true in the regression model exploring viral behavioral intentions. When viewers of highly-arousing videos were not given a route to excitation residual release through commenting, they were more likely to report intentions to share the video with their own social networks. Some of the most important findings in this study, however, seem to challenge some of the assumptions of the theory of planned behavior and its older brother, the theory of reasoned action. Both theories posit that attitudes and subjective norms are key predictors of behavioral intentions. Our study, however, alludes to a potential difference that may have come about as people’s offline lives have become more entangled with and dependent on their online lives. Our study found that the strongest predictor of offline civic engagement was viral behavioral intentions, regardless of attitudes toward the video or issue. Viral behavioral intentions were also stronger than virality, or what could be perceived norms in this study.

5.2. Theoretical and practical considerations This study has significant implications for both future research and for those involved in creating online messages addressing high-stakes societal issues such as cyberbullying. From a theoretical standpoint, this study suggests that there is an opportunity for media researchers to better understand excitation transfer within the online environment. More research into the nature of excitation transfer as it relates to hot-button issues could be of vital importance to practitioners as well. Current online awareness-building and engagement strategies often encourage online discourse through social commenting. However, practitioners may want to consider the implications of taking the proverbial edge off for viewers of highly-arousing videos through offering immediate access to offline civic engagement opportunities, rather than creating online discussion through social commenting. It is also important from a theoretical and practical standpoint that this and other recent studies, such as Hagerstrom et al. (2014) study, have found that highly arousing videos yield greater likelihood toward important behavioral intentions and outcomes. These findings add to the growing LC4MP literature and serve as more evidence that content arousingness could lead to more information processing and a tendency to approach rather than avoid an issue. The most provocative finding in this study is the possibility that scholarly researchers may need to revise the theories of reasoned action and planned behavior when it comes to understanding how online media induces persuasive outcomes. There seem to be important differences in the roles that both attitudes and subjective norms play in people’s online behavior and the conversion of that behavior to offline behavioral intentions. From a practical standpoint, this may also be important because the traditional advertising model that assumes a need to change viewers’ attitudes may not hold when it comes to certain types of societal issues online. Instead, video creators may be better served by focusing on creating content that incites their audiences to a need for some type of immediate action instead of longer-term attitudinal change. On a practical note, our findings also point to the need and potential of online and offline interventions attempting to alleviate the prevalence of cyberbullying and curb its negative effects on college students. Our study used user-generated content and showed that structural message features (i.e., arousal level and virality) as well as behavioral outcomes (i.e., commenting) can play a role in motivating college students to take civic action against cyberbullying. Kraft and Wang (2010) maintained the majority of college students who are cyberbullied cope with cyberbullying alone, without any utility of community resources. First, college administrators as well as social media designers should attempt to offer resources for cyberbullying victims. Second, it is important to engage college students in civic actions to try to change the social norms surrounding the acceptability of cyberbullying. 5.3. Limitations and future research There are a few limitations worth noting. First, the study used a college student sample to investigate the effect of cyberbullying videos on the development of civic behavioral intentions. Future studies should replicate the study’s procedure with different age groups (i.e. adolescents, youth, and older adults). An obvious limitation of this research is the generalizability of our results to a larger population. Despite this constraint, our study related how people respond to highly-arousing online video by moving to action; future research should explore this subject with a variety of subject-types, including those representing different

529

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

demographics, regions, or even cultures. Additionally, the issue chosen for this study is one that continues to receive considerable attention by not only the media, but also through U.S. public schools. Exploring the more general findings and their applicability to a broader array of issues, including those that have not received as much mainstream attention, is important to furthering our understanding of the effects of online videos. In our study, we used user-generated content found on YouTube that addressed the issue of cyberbullying. In an attempt to enhance the study’s ecological validity, we kept the videos’ original titles and sources as they appeared on YouTube. However, it is possible that these elements of the video could have affected the participants’ responses on the attitude and behavioral intention measures, despite the fact that the titles were comparable for all four videos to which the participants were exposed. Future research should examine the effects of video titles and sources on evaluative and behavioral outcomes. Finally, future research should continue to tease apart the effects of excitation transfer in the online environment. It would

seem that the runway between stimulation and excitation residual reduction could be shorter online, so understanding what activities serve to reduce excitement and how they accomplish this is important from both a theoretical and applied standpoint. Future studies should employ physiological measures of cognition and emotion to investigate this phenomenon. 5.4. Conclusion This study investigated the effects of commenting behavior, emotional arousal, and virality on the development of civic behavioral intentions related to offline bullying and cyberbullying. The study found that arousal level intensifies civic action. Additionally, the study found that viral behavioral intentions were the strongest predictors of offline civic behavioral intentions. Appendix A. Items, factor analysis, and reliability analysis results for predictors and dependent variable

Eigenvalue M (SD)

% of Var. explained M (SD)

a

Cronbach’s

Attitudes toward the video (AV) Negative/positive Bad/good Unfavorable/favorable

2.13 (.15)

70.73 (4.51)

.84 (.02)

Attitudes toward the issue (AI) Negative/positive Bad/good Unfavorable/favorable

2.73 (.10)

90.98 (3.30)

.97 (.01)

Viral behavioral intentions (VBI)a This YouTube video is worth sharing with others I would recommend this YouTube video to others I would ‘‘LIKE’’ this video on YouTube I would ‘‘COMMENT’’ on this video on YouTube I would ‘‘SHARE’’ this video on my social media pages (e.g., Facebook, Twitter, Pinterest, etc.) If I saw this video on Facebook, I would ‘‘LIKE’’ it If I saw this video on Facebook, I would ‘‘COMMENT’’ on it If I saw this video on Facebook, I would ‘‘SHARE’’ on my Facebook page If I saw this video on Twitter, I would ‘‘RETWEET’’ it If I saw this video on Twitter, I would ‘‘REPLY’’ to it If I saw this video on Twitter, I would make it a ‘‘FAVORITE’’

6.79 (.43)

61.94 (3.65)

.94 (.01)

Civic Behavioral Intentions (CBI) This video makes me want to participate in a meeting that discusses cyberbullying and bullying This video makes me want to volunteer for an anti-cyberbullying/bullying organization This video makes me want to engage in a community project to reduce cyberbullying/bullying This video makes want to attend a community or neighborhood meeting dealing with the issue of cyberbullying/bullying This video makes me want to participate in an anti-cyberbullying/bullying protest This video makes me want to sign a petition to push for more laws in schools and university to protect individuals from cyberbullying/bullying This video makes me want to sign a petition to push for stricter laws to penalize cyberbullies and bullies

5.83 (.09)

83.35 (1.31)

.97 (<.01)

M (SD)

a Notes. An additional item ‘‘I would ‘‘DISLIKE’’ this video on YouTube’’ was presented to participants, yet due to unsatisfactory factor loading of this item and its negative contribution to the reliability coefficient, we decided to remove it from the averaged index of VBI.

530

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531

Appendix B. Video descriptions The four videos included in the study entailed references to both offline bullying and cyberbullying, with greater emphasis on cyberbullying. All videos were downloaded from YouTube and then uploaded to Qualtrics. Upon later inspection, we discovered that the links to each of the videos are no longer working, hence, in the following, we provide a description of the video content. Please contact the corresponding author for more information about the stimuli.

B.1. Highly-arousing videos 1. A Cyberbullying Suicide Story (4:33 min): The video focuses on the bullying and cyberbullying story of Ryan Halligan, a 13-year-old, who was bullied at school and then committed suicide. 2. Bullycide: Bullied to Death Memory (4:21 min): The video introduces a number of youth suicides as a result of being bullied and cyberbullied. The video is comprised of youth pictures alongside their stories of how they were bullied and then how they committed suicide.

B.2. Lowly-arousing videos 1. Words do hurt (2:57 min): A teenager tells her story of being bullied and cyberbullied using hand-written notes. 2. Teen Bullying (3:06 min): A user-generated video including statistics and other information about bullying and cyberbullying in the United States.

References AdAge (2015, April 21) Hyundai’s father–daughter space story goes viral: Sylvester Stallone promotes Warburtons with a mini action movie. Retrieved from: . Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50(2), 179–211. Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall. Alexa.com (2013). Youtube.com. Retrieved from . Alhabash, S., McAlister, A. R., Hagerstrom, A., Quilliam, E. T., Rifon, N. J., & Richards, J. I. (2013). Between likes and shares: Effects of emotional appeal and virality on the persuasiveness of anticyberbullying messages on Facebook. Cyberpsychology, Behavior, and Social Networking, 16(3), 175–182. Alhabash, S., & McAlister, A. R. (2014). Redefining virality in less broad strokes: Predicting viral behavioral intentions from motivations and uses of Facebook and Twitter. New Media & Society. 1461444814523726. Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs, NJ: Prentice-Hall. Berger, J., & Milkman, K. L. (2011). What makes online content viral? Journal of Marketing Research, 49(2), 192–205. Bolls, P. D. (2010). Understanding emotion from a superordinate dimensional perspective: A productive way forward for communication processes and effects studies. Communication Monographs, 77(2), 146–152. Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: The self-assessment manikin and the semantic differential. Journal of Behavior Therapy and Experimental Psychiatry, 25(1), 49–59. BullyingStatistics.com (2013). Cyber bullying statistics. Retrieved from . Cacioppo, J. T., Gardner, W. L., & Berntson, G. G. (1999). The affect system has parallel and integrative processing components: Form follows function. Journal of Personality and Social Psychology, 76(5), 839. Cantor, J. R., & Zillmann, D. (1973). The effect of affective state and emotional arousal on music appreciation. Journal of General Psychology, 89, 97–108. Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39, 752–756. Chaiken, S., & Troppe, Y. (1999). Dual-process theories in social psychology. New York: Guilford Press.

Cheng, X., Liu, J., & Dale, C. (2013). Understanding the characteristics of internet short video sharing: A YouTube-based measurement study. IEEE Transactions on Multimedia, 15(5), 1184–1194. Choi, Y. K., Miracle, G. E., & Biocca, F. (2001). The effects of anthropomorphic agents on advertising effectiveness and the mediating role of presence. Journal of Interactive Advertising, 2(1). Cialdini, R. B., Reno, R. R., & Kallgren, C. A. (1990). A focus theory of normative conduct: Recycling the concept of norms to reduce littering in public places. Journal of Personality and Social Psychology, 58(6), 1015–1026. Coulter, K., & Punj, G. (2007). Understanding the role of idiosyncratic thinking in brand attitude formation. Journal of Advertising, 36(1), 7–20. Eckler, Petya., & Bolls, Paul. (2011). Spreading the virus: Emotional tone of viral advertising and its effect on forwarding. Journal of Interactive Advertising, 11(Spring), 1–12. Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7(2), 117–140. Fiske, S. T., & Taylor, S. (1984). Social cognition (1st ed.). New York: McGraw-Hill. Glenn, D. (2013). YouTube reaches more U.S. adults than cable networks [infographic]. Social times. Retrieved from: . Hagerstrom, A., Alhabash, S., & Kononova, A. (2014). Emotional dimensionality and online ad virality: Investigating the effects of affective valence and content arousingness on processing and effectiveness of viral ads. In: J., Huh, (Ed.), Proceedings of the 2014 Conference of the American Academy of Advertising, (p. 109). Hansen, C. H., & Hansen, R. D. (1990). The influence of sex and violence on the appeal of rock music videos. Communication Research, 17, 212–234. Hinduja, S., & Patchin, J. W. (2010). Bullying, cyberbullying, and suicide. Archives of Suicide Research, 14(3), 206–221. i-SAFE (2004). Cyber bullying: Statistics and tips. Retrieved from: . Kirby, J. (2004). Getting the bug. Brand Strategy, 184, 33. Kiriakidis, S. P., & Kavoura, A. (2010). Cyberbullying: A review of the literature on harassment through the internet and other electronic means. Family & Community Health: The Journal of Health Promotion & Maintenance, 33(2), 82–93. Kowalski, R. M., Giumetti, G. W., Schroeder, A. N., & Reese, H. H. (2012). Cyberbullying among college students: Evidence from multiple domains of college life. In L. A. Wankel & C. Wankel (Eds.), Misbehavior online in higher education: Cutting-edge technologies in higher education (pp. 293–321). Bingley, UK: Emerald Group Publishing Limited. Kowalski, R. M., & Limber, S. P. (2013). Psychological, physical, and academic correlates of cyberbullying and traditional bullying. Journal of Adolescent Health, 53(1 Supplement), S13–S20. Kraft, E., & Wang, N. (2010). An exploratory study of the cyberbullying and cyberstalking experiences and factors related to victimization of students at a public liberal arts college. International Journal of Technoethics, 1(4), 74–91. Lang, A. (2000). The limited capacity model of mediated message processing. Journal of Communication, 50(Winter), 46–70. Lang, A. (2006). Using the limited capacity model of motivated mediated message processing to design effective cancer communication messages. Journal of Communication, 56, S57–S80. Langos, C. (2012). Cyberbullying: The challenge to define. Cyberpsychology, Behavior, and Social Networking, 15(6), 285–289. http://dx.doi.org/10.1089/cyber. 2011.0588. Lenhart, A. (2007). Data memo: Cyberbullying and online teens. Pew Internet and American Life Project. Retrieved from: . Lenhart, A., Madden, M. (2007). Teens, privacy & online social networks. Pew Internet & American Life Project. . MacDonald, C. D., & Roberts-Pittman, B. (2010). Cyberbullying among college students: Prevalence and demographic differences. Procedia – Social and Behavioral Sciences, 9, 2003–2009. MacKenzie, S. B., & Lutz, R. J. (1989). An empirical examination of the structural antecedents of attitude toward the ad in an advertising prestesting context. Journal of Marketing, 53(2), 48–65. Maio, G. R., & Haddock, G. (2007). Attitude change, In: A.W., Kruglanski, E.T., Higgins, (Eds.), Social psychology: Handbook of basic principles, (pp. 565–586). Muehling, D. D. (1987). An Investigation of factors underlying attitudes toward advertising-in-general. Journal of Advertising, 16(1), 32–40. National Crime Prevention Council (2010). Cyberbullying. Retrieved from: . Nudd, T. (2014, November 20). The 20 most viral ads of 2014: World Cup spots led the way, with pranks, love and empowerment not far behind. AdWeek. Retrieved from: . Nunnally, J. C. (1967). Psychometric theory. New York, US: McGraw-Hill. Patchin, J. W. (2010). 2010. Cyberbullying Research Center. Retrieved from . Patchin, J. W., & Hinduja, S. (2006). Bullies move beyond the schoolyard: A preliminary look at cyberbullying. Youth Violence and Juvenile Justice, 4(2), 148–169. http://dx.doi.org/10.1177/1541204006286288. Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. Advances in Experimental Social Psychology, 19, 123–205. Phelps, J., Lewis, R., & Mobilio, L. (2004). Viral marketing or electronic word-ofmouth advertising: Examining consumer responses and motivations to pass

S. Alhabash et al. / Computers in Human Behavior 51 (2015) 520–531 along email. Journal of Advertising Research, 333–348. http://dx.doi.org/10.1017/ S0021849904040371. Purcell, K. (2013). Online video 2013. Pew Internet and American Life Project. Retrieved from: . Raskauskas, J., & Stoltz, A. D. (2007). Involvement in traditional and electronic bullying among adolescents. Developmental Psychology, 43(3), 564–575. http:// dx.doi.org/10.1037/00121649.43.3.564. Schenk, A. M., & Fremouw, W. J. (2012). Prevalence, psychological impact, and coping of cyberbully victims among college students. Journal of School Violence, 11, 21–37. StatisticsBrain.com (2013). Cyber/bullying statistics. Retrieved from . Strom, P. S., & Strom, R. D. (2005). Cyberbullying by adolescents: A preliminary assessment. Educational Forum, 70(1), 21–36. Thelwall, M., Sud, P., & Vis, F. (2012). Commenting on YouTube videos: From Guatemalan rock to El Big Bang. Journal of the American Society for Information Science and Technology, 63(3), 616–629.

531

Wade, A., & Beran, T. (2011). Cyberbullying: The new era of bullying. Canadian Journal of School Psychology, 26(1), 44–61. http://dx.doi.org/10.1177/ 0829573510396318. Webster, R. (2010, January 29). From cyber bullying to sexting (stats and videos): What’s on your kids’ cell? Examiner.com. Retrieved from: . YouTube (2013). Statistics. Retrieved from: . Zalaquett, C. P., & Chatters, S. J. (2014). Cyberbullying in college: Frequency, characteristics, and practical implications. SAGE Open, 4(1), 1–8. Zillmann, D. (1971). Excitation transfer in communication-mediated aggressive behavior. Journal of Experimental Social Psychology, 7, 419–434. Zillmann, D., Mody, B., & Cantor, J. R. (1974). Empathetic perception of emotional displays in films as a function of hedonic and excitatory state prior to exposure. Journal of Research in Personality, 8, 335–349.