Psychology of Popular Media Culture

When Enemies Go Viral (or Not)—A Real-Time Experiment During the “Stop Kony” Campaign Daniel Sullivan, Mark J. Landau, and Aaron C. Kay Online First Publication, April 14, 2014. http://dx.doi.org/10.1037/ppm0000031

CITATION Sullivan, D., Landau, M. J., & Kay, A. C. (2014, April 14). When Enemies Go Viral (or Not)—A Real-Time Experiment During the “Stop Kony” Campaign. Psychology of Popular Media Culture. Advance online publication. http://dx.doi.org/10.1037/ppm0000031 Psychology of Popular Media Culture © 2014 American Psychological Association 2014, Vol. 3, No. 2, 000 2160-4134/14/$12.00 DOI: 10.1037/ppm0000031 When Enemies Go Viral (or Not)—A Real-Time Experiment During the “Stop Kony” Campaign

Daniel Sullivan Mark J. Landau University of Arizona University of Kansas

Aaron C. Kay Duke University

In March–April 2012, using 2 online videos, nonprofit organization initiated a “Stop Kony” campaign to turn Ugandan warlord into an interna- tional enemy. Although the first video was the fastest of all time, interest in the campaign eventually faded away. Might individual-level psychological processes help explain why the campaign was initially successful, and ultimately failed? To test this possibility, we used a combination of experimental manipulations and real-time data tracking responses to the “Stop Kony” videos as they appeared. Integrating and advancing beyond prior theory on enemyship and idea contagion, our findings suggest that when a complex adverse situation is reduced to the actions of a clear enemy, this inspires moral outrage against the enemy. However, if the complexity of the situation becomes clearer, the enemy inspires less moral outrage and determination to act.

Keywords: Joseph Kony, media, idea contagion, enemyship, moral outrage

On March 5, 2012, the nonprofit organization would place Kony’s image on every street cor- Invisible Children posted an online video called ner. The message of the video apparently ap- Kony 2012 about the abuses of pealed to viewers: Within 6 days of its posting, Joseph Kony, a Ugandan warlord and head of it had been viewed more than 100 million times, the terrorist group the Lord’s Resistance Army. the fastest spread of a “viral video” in history Focusing on his enslavement of children, the (Huffington Post, 2012). video portrayed Kony as an enemy of human Nearly as remarkable as the campaign’s me- ideals everywhere, and called out for participa- teoric rise to popularity was its subsequent fall tion in a global campaign to bring him to jus- from public consciousness. A mere month later tice. Director implored viewers to (April 5), when Invisible Children posted a sec- take part in a “Cover the Night” event on April ond video entitled Kony 2012—Part II: Beyond 20, 2012, when people in cities everywhere Famous, the response was markedly under- whelming (Falkenthal, 2012). Two months later, the date of the planned “Cover the Night” event passed with little fanfare (Carroll, 2012). Today, two years later, Kony’s name has virtu- Daniel Sullivan, Department of Psychology, University This document is copyrighted by the American Psychological Association or one of its allied publishers. ally vanished from the mainstream media, and of Arizona; Mark J. Landau, Department of Psychology, This article is intended solely for the personal use of the individualUniversity user and is not to be disseminated broadly. of Kansas; Aaron C. Kay, Fuqua School of he remains at large. Business and Department of Psychology and Neuroscience, The “Stop Kony” campaign may be consid- Duke University. ered a failed experiment in enemyship: An ul- The authors thank Leah L. Kapa, who contributed the timately ineffective attempt to transform a little- original idea for this study. This research was supported by NSF through a Graduate Research Fellowship awarded to known figure into a symbol of evil. Why was Daniel Sullivan, under Grant DGE-0907996. the campaign initially successful, and why did it Correspondence concerning this article should be ad- fail in the long term? These questions connect to dressed to Daniel Sullivan, Department of Psychology, Uni- the more general issue of why some ideas be- versity of Arizona, 1503 E University Boulevard, P.O. Box 210068, Tucson, AZ 85721. E-mail: [email protected] come “viral” in the public sphere and others do .edu not. We believe that the answer to these ques-

1 2 SULLIVAN, LANDAU, AND KAY

tions lies partly in the psychology of enemy- focused emotion that involves anger toward a ship, and specifically those factors that contrib- particular individual or group for an injustice ute to an enemy figure’s ability to inspire moral they have committed against a third party outrage in individuals, and thus drive them to (Rothschild, Landau, Molina, Branscombe, & social action. We therefore propose to situate Sullivan, 2013; van Zomeren et al., 2004). The these questions in the context of contemporary link between enemyship and moral outrage has theory on enemyship (Sullivan, Landau, & been suggested by many prior studies. A range Rothschild, 2010), moral outrage (van Zom- of theorists and researchers (e.g., Schmitt, 1932/ eren, Spears, Leach, & Fischer, 2004), and idea 2007; Simon & Klandermans, 2001; Sullivan, contagion (Berger & Milkman, 2012; Heath & Landau, Rothschild, & Keefer, 2014) have pro- Heath, 2008). posed that people’s attachment to their political Building on this prior work, we conducted a identities is largely reinforced by the outrage real-time experiment examining changing atti- they feel against enemy figures who represent tudes toward and effects of Invisible Children’s conflicting value systems. For example, Haidt videos in vivo, collecting data at intervals start- and Algoe (2004) proposed that people may ing the first week Kony 2012 appeared online, artificially amplify the immorality of collective until shortly after the “Cover the Night” date. enemy figures to attain a clear sense of their Our study represents the first investigation of in-group’s virtuous nature and identity. This the potential for an enemy to inspire moral process of moral amplification typically invokes outrage in the context of an unfolding real-time feelings of moral outrage against a clear enemy event. The findings suggest that although Kony target. 2012 may have achieved initial popularity ow- Related theorizing (e.g., Simon & Klander- ing to its depiction of a clear enemy figure who mans, 2001; Thomas, McGarty, & Mavor, elicited moral outrage, Kony 2012—Part II ac- 2009) suggests that outrage toward an enemy knowledged too much of the situational com- can transform apathy into feelings of inspiration plexity surrounding this issue, and therefore and determination to restore moral justice by failed to present Kony in simple terms as the defeating the enemy. Supporting this claim, Pa- root of evil. gano and Huo (2007) found that feelings of moral outrage directed specifically against Sad- Enemyship and Moral Outrage dam Hussein and his regime for injustices per- petrated against the Iraqi people predicted pos- An existential perspective on enemyship itive attitudes toward political actions intended (Landau, Sullivan, Rothschild, & Keefer, 2012; to punish Hussein and prevent future human Sullivan et al., 2010) posits that perceiving the rights abuses in Iraq. These findings from the self as having powerful political and personal moral outrage literature, combined with the enemies, although superficially disagreeable, aforementioned findings of Sullivan et al. actually enables the self to maintain a sense of (2010), suggest that the identification of a clear personal efficacy. To specify, most people real- enemy figure increases feelings of moral out- ize that multifarious potential threats to their rage and determination to defeat the enemy. well-being are randomly distributed throughout When Invisible Children created the Kony their environment in the form of nonagentic 2012 video, they were attempting to portray factors that are difficult to understand and con- complex problems—civil war, widespread in- This document is copyrighted by the American Psychological Association or one of its allied publishers. trol. One strategy for sustaining perceived con- justice, and structural imbalance in central Af- This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. trol is tracing all likely misfortunes to the mach- rica, as well as between Africa and the rest of inations of one or a few clearly defined agentic the world—in simple terms as the product of enemies. one man’s misdeeds. As Russell, the video’s How does reducing a complex situation to the director, explicitly stated in an interview: “Be- actions of an identifiable enemy enhance per- cause of the zeitgeist of the culture and the sonal control? One possibility that has yet to be world, we need an enemy . . . We need to know explored in the literature is that identifying a what the worst is” (CNN Wire Staff, 2012). A clear enemy induces feelings of moral outrage, qualitative analysis of Kony 2012 (Kouveld, inspiring a desire to take efficacious action 2012) suggested that the most common themes against the enemy. Moral outrage is an other- in the video were (1) the in-group (U.S. view- STOP KONY 3

ers); (2) the possibility of the in-group taking tions—such as moral outrage—are negative in collective action; and (3) the identification of valence (Berger & Milkman, 2012). A large Kony as an enemy. This lends strong support to field study of sharing patterns for 800 videos our prediction that watching Kony 2012 should found that videos that elicited high emotional elicit an increase in moral outrage, an emotion arousal were twice as likely to be shared as which combines hatred of an enemy with desire low-arousal videos (Nelson-Field, Riebe, & to take collective action. Newstead, 2013). Although videos with posi- tive emotional content are more likely to be Idea Contagion and the Problem of shared than those with negative content, the Complexity effect is much smaller than that for overall arousal. Given that moral outrage is a high- Almost immediately after the appearance of arousal emotion combining both negative and Kony 2012, criticism of the campaign emerged positive elements (a hatred of an enemy, but from various quarters, suggesting that the issues also a positive motivation to take action; involved were more complex than the video had Thomas et al., 2009), it stands to reason that acknowledged, and that Kony was not as pow- videos that elicit moral outrage should be more erful as the filmmakers had claimed. In their popular than those that do not. In the current second video, Invisible Children acknowledged study, therefore, we tracked feelings of moral the criticism and the complexity that it implied. outrage in response to the Kony videos as they The video’s message was more nuanced, sug- appeared. gesting that apprehending Kony was only one Our guiding analysis further suggests that the (however important) step in improving the lives decline of the “Stop Kony” campaign was due of people in the central African regions where in large part to the second video’s acknowledg- he has operated. Other steps would include, for ment of the complexity of the issues in the example, rehabilitating both the region and con- region, as well as the possibility that Kony was flict survivors by supplying educational and fi- not as powerful or central a force as he seemed nancial resources. in the first video, which undermined the likeli- Why did the second Kony video fail to go hood that Kony could serve as a psychologically viral? According to contemporary theory and satisfying enemy. By focusing less on a con- research on the psychology of social transmis- crete enemy, the second Kony video lost the sion and idea contagion, some of the most im- ability to inspire emotional outrage. portant characteristics of ideas that “stick” (i.e., that become popular through social transmis- A Real-Time Study Tracking Responses to sion) are simplicity, concreteness, and ability to “Stop Kony” inspire emotions (Heath & Heath, 2008). Al- though the second Kony video may have been To track the course of an international exper- more comprehensive and accurate than the first iment in enemyship processes, we collected Kony video in its presentation of the facts, its data on reactions to “Stop Kony” beginning the relative complexity may have rendered it less week of Kony 2012’s release. In the first month memorable, vivid, and immediately appealing of data collection, we asked participants to to the public. watch the video online. Shortly after the release This perspective fits our existential view of of the second Kony video, we modified the This document is copyrighted by the American Psychological Association or one of its allied publishers. enemyship, which suggests that people prefer design such that participants watched this video This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. the simple message that apparently chaotic haz- instead of the first. In addition to this real-time ards in their environment stem from a single tracking data, we included an experimental ma- concrete, powerful enemy. On this view, the nipulation to better determine how participants first Kony video went viral because it simplified were responding to the videos and the ideas a complex issue by presenting people with a about them circulating in contemporary media. focal enemy who inspired feelings of moral Specifically, after watching one of the two vid- outrage and determination to act. eos, participants were randomly assigned to Also supporting this view is evidence that read one of two actual articles representing the online content is more likely to go viral if it diverging responses to “Stop Kony” proliferat- elicits strong emotions, even if those emo- ing in the media at that time. One article was 4 SULLIVAN, LANDAU, AND KAY

supportive of the idea that Kony was a powerful several filler items were five items assessing enemy; the other criticized this notion and por- feelings of moral outrage and determination: “I trayed the situation as far more complex. We feel morally outraged about something or to- assessed reactions to these articles as well as ward someone; I feel angry about something or feelings of moral outrage and determination someone; I feel hostile about something or to- prior to and after exposure to the videos and our ward someone; I feel inspired to do something; article manipulation. I feel determined to do something.” These items We initially designed the experiment to test were designed based on theory and research whether exposure to criticism of “Stop Kony” (Thomas et al., 2009) suggesting that feelings of would affect participants’ feelings of moral out- moral outrage toward an enemy generally co- rage and determination to act. Although theory occur with determination to take action. We suggests that reducing a complex issue to the collected responses on these items at the begin- actions of a clear enemy should facilitate feel- ning of the study to provide a baseline measure ings of outrage and determination to act, this (“Pre Measure”); responses to the same items hypothesis has not received empirical scrutiny. were collected again after the video induction The current study fills this gap in the context of and article manipulation to test for change in an unfolding real-time event. The introduction this construct (“Post Measure”). Importantly, of the second Kony video—which acknowl- we designed the items to not refer to a specific edged more of the complexity regarding the target of moral outrage (Joseph Kony being the situation in —was a fortuitous implied target in this case). We did this to happenstance, because it gave us the opportu- measure baseline feelings of outrage/determina- nity to modify our design to test additional tion—collected prior to the administration of untested hypotheses about eneymyship’s effects our experimental manipulations and thus prior on emotions and behavioral intentions. On the to introducing “Stop Kony” as a context—and basis of idea contagion research, we hypothe- sized that the second video would be less com- pelling to participants than the first due its rel- 1 Different samples were recruited not for any a priori ative level of complexity, and that participants reason, but simply to increase power and collect as much data as possible in the midst of a real-world, unfolding would therefore be more likely to endorse crit- event. As expected the samples differed in age, such that icism of the “Stop Kony” campaign after watch- MTurk users were older overall (M ϭ 32 years) than un- ing the second video than the first. On the basis dergraduates (M ϭ 20 years), t(213) ϭ 6.70, p Ͻ .001. It is of enemyship research, we hypothesized that impossible to draw definitive conclusions regarding the possible influence of sample characteristics on our out- the second video’s acknowledgment of complex comes of interest, however, because sample was con- factors beyond Kony himself would render it founded with the video variable (not by a priori design, all less effective than the first video at inspiring undergraduates participated later in the study, and therefore feelings of moral outrage and determination to all saw Kony 2012—Part II). In addition (once more not by act. design), the disproportionate number of MTurk users rela- tive to undergraduates makes it difficult to interpret any group differences. However, what analyses we can perform Method suggest that sample characteristics did not have an impor- tant influence on our outcomes of interest. First, although undergraduates were more likely to have seen Kony 2012 Two hundred seventeen U.S. citizens (98 fe- prior to the study, we controlled for this variable in our

This document is copyrighted by the American Psychological Association or one of its allied publishers. male) participated. Of these, 179 were recruited analyses, so the sample difference in this aspect is not likely

This article is intended solely for the personal use of the individualthrough user and is not to be disseminated broadly. Amazon Mechanical Turk and partici- to account for our findings. Further, the only significant pated in exchange for US$7.00. The remaining interaction or main effect involving sample was a main effect of the variable on Pre Measure feelings of moral 38 were undergraduates recruited through the outrage and determination, t(215) ϭ 2.06, p ϭ .04. Some- University of Kansas participant pool who par- what interestingly, undergraduates showed higher baseline 1 ticipated in exchange for course credit. levels of moral outrage and determination to act (M ϭ 3.40, SD ϭ 1.08) compared with MTurk users (M ϭ 3.00, SD ϭ Moral Outrage and Determination to Act 1.10). Again, however, this finding is difficult to interpret; it might simply be the result of a more accurate estimate of the population value among MTurk users resulting from the Participants rated the extent to which they larger sample. Furthermore, all of our effects of interest were currently feeling each of several emotions remained significant when controlling for sample in our (1 ϭ Not at all,7ϭ Extremely). Embedded in models. STOP KONY 5

to compare these baseline rates with postma- prologue emphasizing the importance of the nipulation feelings. Interitem reliability for all Internet and youth culture in the modern world. five items was acceptably high at both the be- The majority of the first part, however, presents ginning and end of the study session (Pre Mea- the suffering of a Ugandan boy named Jacob sure: ␣ϭ.67; Post Measure: ␣ϭ.87).2 Given whom Russell met in 2003. Jacob is afraid for prior theory and high interitem reliability, we his life and has seen his brother murdered; how- combined scores on all five items into a com- ever, very little context is given to explain his posite measure of felt moral outrage/determina- suffering. This context is given in the second tion assessed both “Pre” and “Post” exposure to part of the video, which also lasts about 10 min. our independent variables. In this section, Russell explains to his young The filler items in which the moral outrage son that Joseph Kony is direct cause of the measure was embedded were included to avoid suffering of Jacob and 30,000 other children in arousing suspicion and alerting participants to central Africa. The video portrays Kony as a the true nature of the study. The items were vicious and enigmatically powerful enemy, similar in wording to the moral outrage/ comparable with Hitler and , determination items, but assessed other unre- who has committed unforgivable crimes against lated emotions, including feeling “interested; children and who is capable of eluding foreign irritable; excited; ashamed; upset; guilty; sad; peacekeeping forces. It also proposes that cap- enthusiastic; regretful; and proud.” turing Kony is a simple solution for improving “Stop Kony” Videos life in . As a representative of the Inter- national Criminal Court says in the video: “The Participants were then told some basic infor- criminal here is Kony. Stop him . . .” before mation about the video Kony 2012. They were adding, as an aside, “. . . and then, solve other instructed that, regardless of whether they had problems.” The final 10 min of the video ex- already seen the video, they would be asked to plain how Kony will be stopped through a unit- watch the video in full before completing addi- ing of Democratic and Republican politicians tional measures. Prior to watching the video, against a common enemy, and through massive participants responded to the single item, “Have public participation in campaign events like you already seen Kony 2012?” (yes or no). “Cover the Night.” The viewer is encouraged to About half the undergraduates (58%) had seen take part in the “Stop Kony” campaign. Kony 2012 prior to the study, whereas only a minority of MTurk users had previously seen Kony 2012 Part II the video (22%; for further information on sam- ple differences, see Footnote 1). To account for Participants whose data were collected dur- any influence of prior exposure to the video, we ing the period of April 12 to May 4, 2012, controlled for this variable in our analyses. watched the second video, Kony 2012 Part II— Kony 2012 Beyond Famous (approx. 20 min). This video is also an edited compilation of news clips, inter- Participants whose data were collected dur- views, and other forms of footage accompanied ing the period of March 10 to April 11, 2012,

watched the first video posted by Invisible Chil- 2 This document is copyrighted by the American Psychological Association or one of its allied publishers. There is a noted increase in the reliability of our moral dren, Kony 2012 (approx. 30 min). Kony 2012 is

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. outrage and determination measure from Pre to Post. It is an edited series of news and Internet clips, likely that this is a test–retest effect, with interitem reliabil- footage shot by its director Jason Russell in ity of the measure increasing as a function of increased Africa and with his family in the United States, familiarity with the items. However, it should also be kept in mind that, prior to the administration of our independent and other material specially prepared or previ- variables (the “Stop Kony” videos and related articles), ously recorded by Invisible Children. The im- participants did not have a particular target on which to ages are accompanied by Russell’s narration. focus feelings of moral outrage, nor a particular goal or The video is divided into roughly three parts, issue about which to feel determination. It is therefore also likely that exposure to a concrete target (Joseph Kony) on which each seem strategically planned to induce which to focus these emotions not only elevated their in- feelings of moral outrage and determination to tensity, but also reduced variability between items assessing act. The first 10 min of the video include a short these emotions. 6 SULLIVAN, LANDAU, AND KAY

by narration. It begins with a 7-min section Kony was not in fact a serious threat (see Ap- primarily about Kony 2012 and the response to pendix for full text of each article). the video. Although some of the criticism of the After reading one of these two articles, par- video is acknowledged, this section primarily ticipants completed a single manipulation check focuses on the role of the video in raising item: “The information in the article suggests awareness and generating attempts to stop that Joseph Kony is a serious threat” (1 ϭ Kony. This section is followed by a short 3-min Definitely no,7ϭ Definitely yes). Participants section arguing that Kony and his army remain then responded to three items assessing their a threat in the Democratic Republic of the agreement with the critical or supportive article Congo and central Africa. In the second half of (1 ϭ Definitely no,7ϭ Definitely yes): “I agree the video, however, the narrative of Kony 2012 with the perspective presented in the article; I Part II becomes more complex. Invisible Chil- think the information presented in the article is dren presents a “Comprehensive Approach” to accurate; I think the author of this article is resolving issues in the regions where Kony has mistaken in their view on the issue.” We re- operated, including four steps: (1) protecting verse-scored the last item and formed composite ␣ϭ civilians by establishing a radio network to alert article agreement scores ( .91). them of raids by hostile forces; (2) ensuring that Finally, participants responded to the same members of the Lord’s Resistance Army who five-item measure used to assess feelings of want to surrender can do so peacefully; (3) moral outrage and determination at the begin- rehabilitating and reconstructing the regions by ning of the session to provide a “Post Measure.” repairing infrastructure (e.g., building schools); and (4) capturing Kony and those closest to Results him. It should be noted that the first video focused almost exclusively on step 4. As a Manipulation Check member of Invisible Children interviewed in the second video states, “Everyone across the board To test whether the article manipulation ef- will agree [the problem] is a complex one, and fectively created diverging perceptions of the it requires a multipronged approach.” The video threat posed by Kony, we submitted manipula- concludes with a short section encouraging tion check responses to a 2 (Article) ϫ 2 viewers to participate in the “Cover the Night” (Video) ANOVA. We found only the expected event on April 20th, and instructing them how main effect of article, F(1, 213) ϭ 216.52, p Ͻ to do so. As of writing, both Kony 2012 and .001. A t test indicated that participants in the Kony 2012 Part II were still available on You- supporting article condition agreed that the ar- ϭ Tube; we refer readers to the videos for further ticle portrayed Kony as more of a threat (M ϭ information. 5.72, SD 1.38) compared with participants in the critical article condition (M ϭ 2.60, SD ϭ ϭ Ͻ ϭ Article Manipulation 1.72), t(215) 14.68, p .001, d 2.00.

After watching one of the two videos, partic- Agreement With the Article ipants were randomly assigned to read an ex- Analysis by condition. To test whether the cerpt from one of two articles that were posted second Kony video was less effective than the This document is copyrighted by the American Psychological Association or one of its allied publishers. (or modified in the case of a Wikipedia article) first at reducing complex international problems This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. online in the week that Kony 2012 was released. to a single, simple source, we submitted article Participants in the supporting article condition agreement scores to a similar ANOVA. There read an excerpt from the contemporary Wikipe- was a main effect of article, F(1, 213) ϭ 42.26, dia (2012) article on Kony, which provided p Ͻ .001, such that participants tended to agree information corroborating the potential hazard more with the supporting article (M ϭ 5.13, he posed and the strength of his forces. Partic- SD ϭ 1.21) overall compared with the critical ipants in the critical article condition read an article, M ϭ 3.75, SD ϭ 1.85. This effect was excerpt from an article posted on NPR.org qualified by a significant interaction, F(1, (Memmott, 2012), which criticized Kony 213) ϭ 6.24, p ϭ .01. Pairwise comparisons 2012’s oversimplifications and suggested that (Fisher’s LSD) revealed that participants were STOP KONY 7

no more likely to agree with the supporting of timepoint on reactions to the article separately article after watching Kony 2012 (M ϭ 5.30, within each article condition. In the critical article SD ϭ 1.24) than after watching Kony 2012— condition, timepoint was positively correlated ϭ ϭ ϭ ϭ Ͻ Part II (M 4.94, SD 1.16), p .23. with article agreement, rs .24, p .01, and However, consistent with predictions, partici- polynomial contrasts yielded significant evidence pants were more likely to agree with the critical of a linear trend, 95% CI (0.08, 1.72), p ϭ .03. article after watching Kony 2012—Part II (M ϭ The effect of time on reactions to the critical 4.09, SD ϭ 1.71) than after watching Kony article is presented in Figure 1. Our findings 2012 (M ϭ 3.40, SD ϭ 1.93), p ϭ .02, d ϭ .38. suggest that, as time passed since the first post- That is, after watching the second video, partic- ing of Kony 2012, participants became more ipants were more likely to acknowledge that the likely to agree that the situation in central Africa problems identified in the initial video were is more complex than the sole actions of Joseph complex and could not be traced to Kony alone. Kony. Analysis by timepoint. To take advantage Although there were no significant trend ef- of the longitudinal nature of our real-time data, fects in the supporting article condition, reac- we also analyzed reactions to the articles as a tions to this article were negatively correlated ϭ function of time of data collection. We classi- with time at a marginal significance level, rs fied participants into six groups based on the Ϫ.18, p ϭ .06, suggesting that reactions to the timepoint at which their data were collected: article endorsing Kony 2012’s message became 1 ϭ March 10, 2 ϭ March 17 to March 25, 3 ϭ less positive as time passed. April 11 (the first three arbitrarily chosen peri- Somewhat expectedly, timepoint played a ods of data collection), 4 ϭ April 12 to April 17 role in the likelihood of participants having seen (the first timepoint at which participants were Kony 2012. At Timepoint 1, only 22% of par- shown Kony 2012—Part II), 5 ϭ April 19 (the ticipants had already seen the video; Timepoint eve of the “Cover the Night” event), 6 ϭ April 2: 19%; Timepoint 3: 10%; Timepoint 4: 38%; 23 to May 4 (the last period of data collection, Timepoint 5: 28%. While these numbers were after “Cover the Night”). We analyzed the effect fairly consistent, with a general increase for

5

4.5

4

3.5

3 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

2.5

2 March 10 March 17 - April 11 April 12 - April 19 April 23 - March 2 5 April 17 May 4

Figure 1. Agreement with the critical article as a function of timepoint. Note. Higher scores indicate greater agreement with the critical article. Scale ranged from 1 to 7. 8 SULLIVAN, LANDAU, AND KAY

Timepoints 4 and 5, by Timepoint 6 50% of Analysis by timepoint. Additional analy- participants had seen Kony 2012. However, ses for effects of time revealed a significant controlling for this variable in our timepoint quartic trend in feelings of moral outrage/ analyses did not change the pattern of results. determination Post Measure, 95% CI (Ϫ1.05, Ϫ.005), p ϭ .05. This trend is depicted in Fig- Feelings of Moral Outrage and ure 2. These data suggest that Kony 2012 was Determination to Act most able to inspire moral outrage in mid- March (Timepoint 2); that feelings of outrage Analysis by condition. We submitted reached a nadir at Timepoint 4 with the intro- moral outrage/determination scores at the be- duction of the second video; and that outrage ginning and end of the session to a mixed-model ϫ ϫ briefly increased once more at Timepoint 5 on 2 (Article) 2 (Video) 2 (Measure: Pre vs. the eve of the “Cover the Night” event. There Post) ANOVA. This analysis yielded a main ϭ Ͻ were no effects of timepoint on Pre Measure effect of measure, F(1, 213) 113.87, p scores. .001, such that Post Measure moral outrage and ϭ We hypothesized that, at least among partic- determination scores were higher overall (M ipants in the critical article condition, declining SD ϭ 4.25, 1.57) compared with Pre Measure feelings of moral outrage as a function of time scores (M ϭ 3.07, SD ϭ 1.10). This was not may have occurred owing to increased acknowl- unexpected, given that at the beginning of the edgment of complexity regarding the situation study participants had no particular target in Uganda (and a decreased sense that Kony against which they might direct feelings of alone could be blamed). To test this mediational moral outrage. However, the main effect of hypothesis, we used Preacher and Hayes’ measure was qualified by the Measure ϫ Video ϭ Ͻ (2008) bootstrapping approach (Footnote 3 de- interaction, F(1, 213) 10.61, p .01. No 3 other significant main effects or interactions scribes this approach in more detail). emerged, Fs Ͻ 1, ps Ͼ .30. For those participants who read the critical Pairwise comparisons (Fisher’s LSD) revealed article, we regressed moral outrage/determina- that moral outrage/determination increased Pre tion scores (Post Measure) onto timepoint, with (M ϭ 3.04, SD ϭ 1.14) to Post (M ϭ 4.57, SD ϭ agreement with the article (i.e., acknowledg- 1.54) among participants who watched Kony ment of complexity) entered as the proposed 2012, F(1, 213) ϭ 98.58, p Ͻ .001, d ϭ 1.13. mediator. Five thousand bootstrapping resa- Moral outrage/determination also increased Pre mples were performed. The 95% CI for the (M ϭ 3.10, SD ϭ 1.06) to Post (M ϭ 3.92, SD ϭ effect of timepoint on outrage/determination via 1.54) among participants who watched Kony the mediator of agreement with the critical ar- Ϫ Ϫ 2012—Part II, F(1, 213) ϭ 27.04, p Ͻ .001, d ϭ ticle did not contain zero ( .24, .03). Also, .62. Supporting predictions, the magnitude of the the significant effect of timepoint on outrage/ effect was considerably larger for the first video (a determination (␤ϭϪ.19, SE ϭ .09, t(110) ϭ large effect size) than for the second video (a Ϫ1.98, p ϭ .05) became nonsignificant when medium effect size; Cohen, 1992). Also support- controlling for reactions to the article, t(109) ϭ ing predictions, focusing only on the Post Mea- Ϫ.56, p ϭ .58. This suggests that, as time sure, participants who watched Kony 2012 re-

ported more moral outrage/determination than 3 This document is copyrighted by the American Psychological Association or one of its allied publishers. This approach to testing mediational hypotheses is recom- those who watched Kony 2012—Part II, F(1,

This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. mended over more traditional regression methods. Evidence 213) ϭ 9.82, p Ͻ .01, d ϭ .42. Consistent with for mediation can be obtained by testing the significance of the these findings, submitting a difference score (Post multiplicative parameter a ϫ b (where a represents the IV’s Measure scores minus Pre Measure scores) to the effect on the mediator and b represents the mediator’s effect on the DV controlling for the IV). However, this interactive pa- same analysis also yielded a main effect of Video, rameter is highly unlikely to have a normal distribution in most such that the amount of change in moral outrage/ cases (owing to multicollinearity and other issues). Therefore, determination was higher among participants who a bootstrapping approach is most appropriate for testing this watched Kony 2012 (M ϭ 1.53, SD ϭ 1.70) parameter, because it does not make assumptions about the shape of the distribution, but rather samples multiple times compared with those who watched Kony 2012— from the data (with replacement) to create an approximate Part II (M ϭ .83, SD ϭ 1.50), F(1, 220) ϭ 11.08, distribution based on the data (in order to test for the signifi- p Ͻ .01. cance of parameters in terms of this distribution). STOP KONY 9

6

5.5

5

4.5

4

3.5

3 March 10 March 17 - April 11 April 12 - April 19 April 23 - March 2 5 April 17 May 4

Figure 2. Feelings of moral outrage and determination to act (Post Measure) as a function of timepoint. Note. Higher scores indicate greater feelings of moral outrage and determination to act. Scale ranged from 1 to 7.

passed, participants became more convinced watched the second video showed a more fa- that suffering in Uganda is a complex issue that vorable response to an article criticizing the transcends Kony alone, which in turn led to notion that Kony is a threat. decreased feelings of moral outrage and deter- Finally, analyses of the data by timepoint mination to take action. further confirm that the release of Kony 2012— Part II, one month after the release of the first Discussion video, led to a major reduction in feelings of moral outrage against Kony. These feelings Our findings support the notion that Kony temporarily resurged in anticipation of the 2012—Part II was not as effective as the orig- planned “Cover the Night” event, but then de- inal Kony 2012 video in casting Kony as a clined once more, presumably as a response to convincing enemy figure. They also generally the failure of “Cover the Night” to materialize suggest that the effectiveness of the videos in into a worldwide phenomenon. Also, as time inspiring viewers and deflecting criticism de- passed and criticism of the “Stop Kony” cam- clined as a function of passing time. First, we paign became more widespread, participants be- This document is copyrighted by the American Psychological Association or one of its allied publishers. observed that although feelings of moral out- came more likely to endorse criticism of the This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. rage and determination to act increased relative videos, which directly contributed to the grad- to baseline when participants watched either ual decrease in the videos’ ability to inspire Kony 2012 or Kony 2012—Part II, the magni- feelings of moral outrage. tude of this effect was considerably larger A clear limitation of these data, resulting among participants who watched the first video, from their real-time nature, is that exposure to and Post Measure levels of moral outrage were the second video is confounded with time. We larger among these participants compared with cannot conclusively determine on the basis of those who watched the second video. Second, these data whether the declining ability of the we observed an interaction between video and “Stop Kony” campaign to inspire feelings of article, such that those participants who determination and moral outrage was owing to 10 SULLIVAN, LANDAU, AND KAY

the second video’s acknowledgment of prob- outrage. These findings are almost certainly lematic complexity surrounding the issue, or to complementary; it is likely that exposure to a increasing awareness of this complexity and simplified enemy target gives people an en- criticism of the project as a function of time. We hanced sense of control, facilitating their sense believe that both passing time and the cam- that they could take action against this target paign’s inability to strategically maintain an and hence galvanizing moral outrage (in line idealized and simplistic image of Kony as an with the analysis of Thomas & McGarty, 2009). enemy figure contributed to the rapid diminish- This heightened sense of moral outrage and deter- ment of interest in the issue. It is very unlikely mination, experienced in the imagined presence of that the passage of time alone could explain our an enemy, likely further enhances perceptions of observed effects. Consider that although many control and agency, a possibility suggested by news stories have a relatively short cycle, en- research showing that anger induces greater con- emy figures like Osama bin Laden have stayed fidence and certainty in information processing in the public eye for much longer periods of (Tiedens & Linton, 2001). The reciprocal interac- time. Indeed, our analyses suggest that the de- tion between control perceptions and feelings of crease in feelings of moral outrage toward Kony outrage in reaction to clear enemy figures is a over time was a result of increased awareness of fruitful topic for further research. and agreement with criticisms of the campaign. The present study makes an important contri- In short, it is unlikely that mere passage of time bution to the moral outrage literature by demon- caused our effects, but rather the increasing strating the key role of simplified presentations of awareness of complexity surrounding Kony that an enemy in eliciting moral outrage. The observed happened over time. differences in reactions to Kony 2012 versus Kony As a real-world test of enemyship processes, 2012—Part II support contemporary theory con- this study is the first to demonstrate in the cerning idea contagion (e.g., Heath & Heath, context of an unfolding political event that re- 2008). The first video presented a simple message ducing a complex issue to the actions of a single portraying Kony as a concrete enemy figure, and enemy can inspire moral outrage and inspiration dramatically increased moral outrage and determi- to take action. Clear enemies increase feelings nation relative to baseline levels, whereas the sec- of moral outrage, and this emotional impetus is ond video presented a more complex view of the probably why the first Kony video went viral at situation, was less successful at deflecting criti- such intense speed. However, acknowledging cism, and therefore produced a less substantial complexity in connection with an enemy figure increase in moral outrage. Although prior studies can diminish the potential for that enemy to on moral outrage suggest the importance of an galvanize outrage and determination to act. enemy figure or third-party scapegoat in eliciting The present findings contribute to the litera- the emotion (e.g., Rothschild et al., 2013; Thomas ture in multiple ways. First, they represent the et al., 2009), there has not yet been a demonstra- first real-time documentation of increases (and tion of the specific importance of simplified, clear decreases) in moral outrage in response to an representations of the enemy to induce moral out- enemy figure as he rose to attention in the online rage. and news media. Accordingly, they also con- On the whole, our findings might have some- tribute to the growing literature on the transmis- what pessimistic implications. They suggest sion of ideas in an online society (e.g., Berger & that the most popular political media content This document is copyrighted by the American Psychological Association or one of its allied publishers. Milkman, 2012), by documenting the fluctuat- may be the sort that it is able to inspire strong This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ing ability of an online video to elicit moral emotional reactions, such as feelings of moral outrage at the same time that it was rising and outrage. Furthermore, when media content at- declining in popularity in the online community. tempts more nuanced or complex approaches to Furthermore, this study advances beyond the political issues, it seems that the ability to in- extant psychological research on enemyship. spire outrage and determination is significantly Although past studies have focused on the ca- lessened. “Stop Kony” largely failed as a col- pacity for enemy figures to boost perceptions of lective experiment in enemyship because the personal control (e.g., Sullivan et al., 2010), our leaders of the movement were unable to consis- data show that enemies—when simplistically tently defend a clear image of Joseph Kony as a portrayed—can also induce feelings of moral powerful enemy. Kony’s 15 min of infamy have STOP KONY 11

come and gone, and yet many people in central NPR.org.Retrievedfromhttp://www.npr.org/blogs/ Africa continue to suffer, unaided, whether they thetwo-way/2012/03/07/148146240/ugandan- are direct victims of Kony’s actions or for other warlord-joseph-kony-under-spotlight-thanks-to- reasons. This highlights the importance of find- viral-video. Retrieved March 8, 2012. ing ways to maintain public emotional invest- Nelson-Field, K., Riebe, E., & Newstead, K. (2013). ment in political issues, despite the often com- The emotions that drive viral video. Australasian Marketing Journal, 21, 205–211. doi:10.1016/j plex realities that surround these topics. .ausmj.2013.07.003 Pagano, S. J., & Huo, Y. J. (2007). The role of moral References emotions in predicting support for political actions in post-war Iraq. Political Psychology, 28, 227– Berger, J., & Milkman, K. L. (2012). What makes 255. doi:10.1111/j.1467–9221.2007.00563.x online content viral? Journal of Marketing Re- Preacher, K. J., & Hayes, A. F. (2008). Asymptotic search, 49, 192–205. doi:10.1509/jmr.10.0353 and resampling strategies for assessing and com- Carroll, R. (2012, April 21). Kony 2012 Cover the paring indirect effects in multiple mediator mod- Night fails to move from the internet to the streets. els. Behavior Research Methods, 40, 879–891. . Retrieved from http://www doi:10.3758/BRM.40.3.879 .guardian.co.uk/world/2012/apr/21/kony-2012- Rothschild, Z. K., Landau, M. J., Molina, L. E., campaign-uganda-warlord Branscombe, N. R., & Sullivan, D. (2013). Dis- CNN Wire Staff. (2012, March 12). Invisible children placing blame over the ingroup’s harming of a to address questions in new film Monday. Retrieved disadvantaged group can fuel moral outrage at a from http://www.cnn.com/2012/03/12/world/africa/ third-party scapegoat. Journal of Experimental So- uganda-viral-video/index.html?hptϭhp_c3 cial Psychology, 49, 898–906. doi:10.1016/j.jesp Cohen, J. (1992). A power primer. Psychological .2013.05.005 Bulletin, 112, 155–159. doi:10.1037/0033–2909 Schmitt, C. (1932/2007). The concept of the political .112.1.155 (G. Schwab, Trans.). Chicago, IL: University of Falkenthal, G. L. (2012, April 19). Can “Cover the Chicago Press. doi:10.7208/chicago/9780226738840 Night” save KONY 2012 on April 20? Communi- .001.0001 ties at WashingtonTimes.com. Retrieved from http:// Simon, B., & Klandermans, B. (2001). Politicized communities.washingtontimes.com/neighborhood/ collective identity: A social psychological analysis. media-migraine/2012/apr/19/can-cover-night-save- American Psychologist, 56, 319–331. doi:10.1037/ kony-2012-april-20/. Retrieved April 20, 2012. 0003-066X.56.4.319 Haidt, J., & Algoe, S. (2004). Moral amplification Sullivan, D., Landau, M. J., & Rothschild, Z. K. and the emotions that attach us to saints and de- (2010). An existential function of enemyship: Ev- mons. In J. Greenberg, S. L. Koole, & T. Pyszc- idence that people attribute influence to personal zynski (Eds.), The handbook of experimental exis- and political enemies to compensate for threats to tential psychology (pp. 322–341). New York, NY: control. Journal of Personality and Social Psy- Guilford Press. Heath, C., & Heath, D. (2008). Made to stick: Why chology, 98, 434–449. doi:10.1037/a0017457 some ideas survive and others die. New York, NY: Sullivan, D., Landau, M. J., Rothschild, Z. K., & Random House. Keefer, L. A. (2014). Searching for the root of all Huffington Post. (2012, April 12). How “Kony 2012” evil: An existential-sociological perspective on po- went viral. Retrieved from http://www.huffingtonpost litical enemyship and scapegoating. In J.-W. Van .com/2012/04/12/kony-2012-viral-infographic_n_ Prooijen & P. A. M. Van Lange (Eds.), Power, 1421812.html. Retrieved April 23, 2012. politics, and paranoia: Why people are suspicious Kouveld, T. (2012). Violence and victimhood: The of their leaders (pp. 292–311). Cambridge: Cam- This document is copyrighted by the American Psychological Association or one of its allied publishers. language and image of Kony 2012 (BA thesis). bridge University Press. This article is intended solely for the personal use of the individual user and is notUtrecht: to be disseminated broadly. Utrecht University. Thomas, E. F., & McGarty, C. A. (2009). The role of Landau, M. J., Sullivan, D., Rothschild, Z. K., & efficacy and moral outrage norms in creating the Keefer, L. A. (2012). Deriving solace from a nem- potential for international development activism esis: Having scapegoats and enemies buffers through group-based interaction. British Journal of against the threat of meaninglessness. In M. Mi- Social Psychology, 48, 115–134. doi:10.1348/ kulincer & P. Shaver (Eds.), Meaning, mortality, 014466608X313774 and choice: The social psychology of existential Thomas, E. F., McGarty, C. A., & Mavor, K. I. (2009). concerns (pp. 183–202). Washington, DC: APA. Turning “apathy into movement”: The role of proso- doi:10.1037/13748-010 cial emotions in motivating action for social change. Memmott, M. (2012, March 8). Ugandan warlord Personality and Social Psychology Review, 13, 310– Joseph Kony under spotlight thanks to viral video. 333. doi:10.1177/1088868309343290 12 SULLIVAN, LANDAU, AND KAY

Tiedens, L. Z., & Linton, S. (2001). Judgment under mouth is! Explaining collective action tendencies emotional certainty and uncertainty: The effects of through group-based anger and group efficacy. specific emotions on information processing. Jour- Journal of Personality and Social Psychology, 87, nal of Personality and Social Psychology, 81, 973– 649–664. doi:10.1037/0022–3514.87.5.649 988. doi:10.1037/0022–3514.81.6.973 Wikipedia. (2012, March 8). Joseph Kony. Retrieved van Zomeren, M., Spears, R., Leach, C. W., & Fi- from http://en.wikipedia.org/wiki/Joseph_Kony. scher, A. H. (2004). Put your money where your Retrieved March 8, 2012.

Appendix

Full Text of the Article Manipulation Supporting Article Condition Sacred Heart Secondary and St. Mary’s girls schools. Originally, Kony’s group was called the For a decade starting in the mid-1990s, the United Holy Salvation Army (UHSA) and was LRA was strengthened by military support from not perceived as a threat by the National Resis- the government of Sudan, which was retaliating tance Army (NRA). By 1988, it had become a against Ugandan government support for rebels major player in Ugandan affairs: an agreement in what would become . Sudan between the NRA and the Uganda People’s withdrew its support for the LRA shortly after Democratic Army left members of the latter the International Criminal Court (ICC) issued a group unsatisfied, and many joined the United warrant for Kony’s arrest, however. Holy Salvation Army as a form of rebellion. One such person was Commander Odong Critical Article Condition Latek, who convinced Kony to use standard military tactics instead of attacking in cross- Foreign Policy has weighed in with a story shaped formations and sprinkling holy water. headlined “Joseph Kony is not in Uganda (and The new tactics proved successful, and the other complicated things).” Among its points: UHSA delivered several small but stinging de- “It would be great to get rid of Kony. He and feats against the NRA. After these victories, the his forces have left a path of abductions and NRA responded by significantly weakening Ko- mass murder in their wake for over 20 years. ny’s group through political actions and a mil- But let’s get two things straight: (1) Joseph itary campaign named Operation North. The Kony is not in Uganda and hasn’t been for 6 operation was devastating to what would be- years; (2) the LRA now numbers at most in the come the Lord’s Resistance Army, and with hundreds, and while it is still causing immense their numbers reduced from thousands to hun- suffering, it is unclear how millions of well- dreds, they engaged in retaliatory attacks on meaning but misinformed people are going to civilians and NRA collaborators. The LRA say help deal with the more complicated reality.” that spirits have been sent to communicate this And Time Magazine’s Global Spin Blog mission directly to Kony. Adds that: The bulk of Kony’s foot soldiers were chil- “Analysts agree that after concerted cam- This document is copyrighted by the American Psychological Association or one of its allied publishers. paigns against the LRA, its numbers at this This article is intended solely for the personal use of the individualdren. user and is not to be disseminated broadly. While estimates of the number of children conscripted since 1986 vary, some put the figure point have diminished, perhaps amounting to as high as 104,000. When abducting the chil- 250 to 300 fighters at most. Kony, shadowy and dren, Kony and his army often killed their fam- illusive, is a faded warlord on the run, with no ily and neighbors, thus leaving the children with allies or foreign friends.” little choice but to fight for him. By 1992 Kony had renamed the group the Received May 16, 2013 United Democratic Christian Army and it was at Revision received January 30, 2014 this time that they kidnapped 44 girls from the Accepted February 4, 2014 Ⅲ