May 2020

The "Need for Chaos" and Motivations to Share Hostile Political Rumors

Michael Bang Petersena*, Mathias Osmundsena, and Kevin Arceneauxc

a Department of , Aarhus University, Aarhus, Denmark

c Department of Political Science, Temple University, Philadelphia, USA

* Corresponding author: Michael Bang Petersen, E-mail: [email protected]

Abstract

Why are some people motivated to share hostile political rumors, such as conspiracy theories and other derogatory news stories? Previous research mostly focuses on the thesis that people’s partisan identities motivate them to share hostile political rumors as a way to tarnish their political opponents. In this manuscript, we demonstrate disruptive psychological motivations also play an important, but often overlooked, role in the spread of hostile rumors. We argue that many individuals who feel socially and politically marginalized are motivated to circulate hostile rumors because they wish to unleash chaos to “burn down” the entire established political order in the hope they can gain status in the process. We conducted 8 large studies (including a nationally representative study) to show that some individuals are predisposed to have a Need for Chaos when facing social isolation and discontent, and that this need is the strongest predictor of explicit motivations to share hostile political rumors. Panel and experimental data show that chaotic motivations reflect an interplay between traits related to status-seeking and states of social exclusion. Consistent with the rising inequality in the United States and beyond, we find that these motivations are strikingly prevalent. To stem the tide of hostile political rumors on social media, the present findings suggest that real-world policy solutions are needed to address the growing social and political frustrations of democratic populations.

Keywords: Political rumors, partisanship, antisocial personality traits, social status, political violence

1 The current day expansiveness of social media provides citizens with unprecedented power to craft and share news with each other. Unfortunately, in the domain of politics, this technological transformation has made it easier than before to spread what we call hostile political rumors in a way that goes “farther, faster, deeper, and more broadly than truth” (Vosoughi, Roy and Aral, 2018, 1147).

Hostile political rumors simultaneously portray politicians and political groups negatively and possess low evidential value (Rosenstiel, Mitchell and Jurkowitz, 2012; Vosoughi, Roy and Aral, 2018). They encapsulate conspiracy theories (Uscinski and Parent, 2014), “fake news” (Allcott and Gentzkow,

2017), discussions of political scandals (Warner, McGowen and Hawthorne, 2012) and negative campaigns (Gross and Johnson, 2016). Although these different types of news vary in substance and form, they nonetheless share two crucial features: they seek to incite hostility toward a specific target and are difficult to verify or categorically disprove. Hostile political rumors can shape political outcomes in considerable ways, from sparking significant small-scale incidents — such as protests

(Tucker et al., 2017), derailing of police investigations (Starbird et al., 2014), and cyberbullying of political opponents (Munger, 2017) — to influencing large-scale outcomes — such as the 2016 US presidential election (Allcott and Gentzkow, 2017) and the Brexit referendum (Bastos and Mercea,

2017).

In this manuscript, we identify the psychological motivations underlying the willingness to share hostile political rumors in contemporary United States. In doing so, we outline the existence of a largely overlooked strategy for status-acquisition focused on destruction -- the incitement of chaos

-- as well as demonstrate the relevance of this strategy for contemporary political motivations. In doing so, we build on recent research on how the interplay between status-oriented personality traits and social rejection can lead to an escalation of aggressive motivations (Krizan & Johar, 2015;

Kurzban & Neuberg, 2005; Twenge & Campbell, 2003) and argue that such motivations, when sufficiently strong, will take the form of a general destructive mindset. We develop and validate a

2 novel scale that directly taps this mindset, the Need for Chaos scale, and apply the scale and framework to motivations to share hostile rumors in politics. Across eight large studies (including a representative study of the US population), we find evidence that the Need for Chaos emerges in an interplay between maladaptive traits and feelings of marginalization and is a strong predictor of willingness to share hostile political rumors, over and beyond traditional factors such as partisanship.

Overall, our findings imply that a challenge facing modern society is the existence of marginalized status-seekers who are willing to spread rumors in order to incite chaos in a society that they do not feel recognizes them.

Psychological Goals Underlying Hostile Rumor-Sharing

In understanding motivations to share hostile political rumors it is helpful to distinguish between epistemic and social goals (Douglas et al. 2017). Significant work on why people believe conspiracy theories (Van Prooijen & Jostmann, 2013) and why some people find it difficult to discern "fake news" from accurate news stories (Pennycook & Rand, 2019) have emphasized the role of epistemic motives

-- i.e., the need to build a "stable, accurate, and internally consistent understanding of the world"

(Douglas et al. 2017: 538). In this regard, some scholars argue that types of hostile political rumors, such as "fake news" and conspiracy theories, often appeal to people with less strict epistemic needs for critical thinking (Pennycook & Rand, 2019). Thus, from an epistemic perspective, people may be motivated to share hostile political rumors simply because they believe them to be true.

While some people may be duped into believing outlandish things, it is likely the case that sharing hostile political rumors serves social goals beyond promoting the perceived truth (Douglas et al. 2017). Hostile political rumors appeal almost by definition to individuals motivated by aggressive impulses. Accordingly, research on conspiracy beliefs has demonstrated that such beliefs relate to support for political violence (Uscinski & Parent, 2014), political extremism (Van Proojien et al.,

3 2015), and anti-social personality traits such as narcissism (Chicocka et al., 2016). Sharing hostile political rumors can buttress aggressive social goals in two complimentary ways. It can serve as direct verbal aggression against a target or as a way to indirectly facilitate aggression (Björkqvist et al. 1994).

Thus, sharing hostile rumors with ingroup audiences can mobilize like-minded individuals against an outgroup target (Horowitz, 2001).

In the present series of studies, we provide a detailed examination of the aggressive social goals underlying motivations to share hostile political rumors. While it is likely that motivations to share hostile political rumors reflect, at least in part, aggression-oriented social goals, we set out to answer a key question that remains open: To whom or what are these goals directed toward?

Helping Your Party or Destroying the System?

A standard explanation in , cognitive science and political science is that people's current motivations to share hostile political rumors reflect partisan motivations (Allcott and

Gentzkow, 2017; Van Bavel & Pereira, 2018; Spohr, 2017). Most individuals form partisan identities in adolescence and early adulthood in the form of strong emotional attachments to a political party

(Campbell et al., 1960), which influence a range of political and social behaviors (Arceneaux and

Vander Wielen, 2017; Cohen, 2003). Deepening political conflict has caused partisans in the U.S. and beyond to develop increasingly hostile feelings toward each other (Iyengar, Sood and Lelkes, 2012;

Mason, 2018), making it natural to presume that deeply committed partisans are willing to strategically share hostile political rumors to target members of the opposing party and, thus, mobilize audiences on behalf of their own party. Consistent with this expectation, a number of studies have shown that partisanship plays a major role in predicting political conspiracy beliefs (Uscinski and

Parent, 2014; Miller, Saunders and Farhart, 2016) and beliefs in "fake news" stories (Allcott and

Gentzkow, 2017; Van Bavel and Pereira, 2018). Thus, in the United States, Republican voters tend to

4 believe stories that denigrate Democrats and Democratic voters tend to believe stories that denigrate

Republicans.

The United States is currently experiencing high levels of partisan polarization (Iyengar et al.,

2019). Accordingly, it may be unsurprising that US partisans are more likely to believe partisan- aligned rumors. However, it does not necessarily follow that partisans should also be more motivated to share hostile political rumors that often feature outrageous claims and conspiracy theories (Duran,

Nicholson and Dale, 2017). Believing is a private act, while sharing is a public act. Consequently, sharing entails potential social repercussions including aggressive reactions, which believing does not

(Altay et al., 2019). Because feelings of hostility facilitate hostile rumor sharing, hostile feelings need to reach a significant level to outweigh the potential costs. At present, it is unclear whether a commitment to one mainstream party over another entails that level of hostility, despite polarization.

Furthermore, many current day political rumors (such as Birtherism, Deep State or Pizzagate conspiracy theories) are indictments of the entire political system as rigged (Uscinski and Parent,

2014), and sharing could be damaging to the broader political system that includes the leaders of one’s party. For a partisan, many hostile political rumors are double-edged swords. Consequently, it is possible that the upsurge in hostile political rumors is about more than just partisan cheerleading.

Alongside growing partisan polarization, advanced democracies have also experienced rising levels of income inequality and stagnation in real wages (Turchin, 2016). As these objective indicators of social wellbeing have worsened, citizens living in these countries have increasingly expressed a feeling of “losing out” and have become increasing discontent with the political establishment

(Kitschelt, 2002). According to prior work, these feelings of status-loss and marginalization have shaped recent political events, including the election of Donald Trump as president of the United

States (Mutz 2018) and the rise of populism in Europe (Kriesi & Schulte-Cloos, 2020).

Here, we argue that feelings of marginalization may relate to motivations to engage in

5 disruptive behavior such as sharing hostile political rumors. Thus, while some research finds that individuals exposed to conspiracy theories are more likely to favor the status-quo (Jolley et al., 2018), other research finds that indicators of marginalization such as lack of trust and alienation predict conspiratorial beliefs beyond partisanship (Miller, Saunders and Farhart, 2016; Balmas, 2014). More generally, for a select number of individuals, feelings of discontent have been found to lead to radicalized behavior. Turchin (2016), for example, links rising inequality to an upswing of violent political events in the United States since the 1970s. Psychological research also shows that uncertainty about one’s own social standing predicts identification with radicalized groups (Hogg,

Kruglanski & van den Bos, 2013 and Farquharson, 2010), support for political violence (Gøtzsche-

Astrup, 2019) and enhanced derogation of outgroups (Van den Bos et al., 2007). Essentially, feelings of marginalization may drive a syndrome of highly disruptive sentiments that include motivations to share hostile political rumors that target the mainstream political system. Perhaps paradoxically, these feelings seem to be particularly widespread among groups traditionally believed to be majority or dominant groups (e.g., white males) (Bai & Federico, 2019; Kaufmann, 2018). Taken together, this constellation of findings implies that disruptive sentiments in current day politics may be tied to disgruntled members of majority groups who direct their frustration toward a system that they believe has abandoned them.

In sum, it is possible to identify, at least, two potential targets when individuals engage in sharing of hostile political rumors. First, prior work has argued that individuals may seek to help one party within the political system over another via verbal aggression and/or audience mobilization, but has mainly provided tests focusing on belief in hostile political rumors rather than sharing motivations.

Second, given the psychological difference between believing and sharing and the existence of growing sociostructural inequality that goes beyond party polarization, our claim is that some individuals may also seek to target the entire political system. Individuals motivated to strategically

6 help one actor within the political system should be highly selective in the rumors shared and invest mainly in sharing those that denigrate the adversary of that actor. In contrast, individuals motivated to target the entire system should instead share hostile rumors in a promiscuous, indiscriminate fashion.

The Causes of Extreme Political Dissatisfaction

We hypothesize that some individuals are so disaffected with the current structure of society and their status in it, that they engage in indiscriminate sharing of hostile political rumors to disrupt the entire established democratic “cosmos” and start anew. Considering individual differences in how people react to marginalization is important, because just as lashing out against the system could be a predictable response to perceived marginalization, research on system justification shows that individuals in marginalized societal positions sometimes tend "to accept and justify status hierarchies"

(Overbeck et al., 2004: 35). Furthermore, recent work on radicalization suggests that feelings of marginalization, even if extreme, only lead to highly disruptive sentiments for some individuals

(Gøtzsche-Astrup, 2019). Taken together we anticipate that radicalized mindsets require both particular situational triggers and particular personality dispositions.

Dispositional traits related to perceptions of status competition and the personal importance of status may be particularly relevant for the emergence of disruptive sentiments. These traits include individual differences in status-driven risk-taking (Ashton et al., 2010), social dominance orientation

(Pratto et al., 1994), competitive jungle worldviews (Duckitt et al., 2002) and the dark triad of psychopathy, machiavellianism and narcissism (Paulhus et al., 2002). Such status-oriented traits have been found to predict a number of antisocial behaviors such as engagement in interpersonal violence

(Wilson and Daly, 1985), support for the use of force in politics (Bartusevicius et al., 2020) and, of particular relevance to the sharing of hostile political rumors, engagement in cyberbullying (Buckels,

7 Trapnell and Paulhus, 2014).

While the relationship between status-oriented traits and antisocial behavior in politics and beyond is well-established, we know significantly less about the relationship between such traits and motivations to disrupt established hierarchies, especially in politics. Status-oriented individuals often support existing social hierarchies, especially if these hierarchies benefit them directly. In politics, significant research has shown that social dominance orientation is linked to support for established hierarchies between races and genders, especially among white males (Pratto et al., 1994). More counterintuitively, research on system-justification has also suggested that minority group members high in social dominance are particularly likely to engage in justifying status differences between minority and majority groups (Overbeck et al., 2004). According to past research, status-oriented traits could potentially lead to more rather than less support for the political system.

To understand the relationship between status-oriented motivations and dissatisfaction with the current system, we contend that it is important to take situational triggers into account and, in particular, feelings of marginalization. Work on system-justification has shown that a key driver of dissatisfaction is discrepancies between people's actual social standing and the social standing to which they feel entitled (Zimmerman & Reyna, 2013). Individuals with status-oriented traits, especially those from majority groups, will feel greater entitlement and will be more motivated to obtain a superior position within the societal hierarchy. Among individuals with status-oriented traits, feelings of marginalization should activate greater feelings of dissatisfaction with the current system.

Thus, we predict that when status-oriented traits are coupled with feelings of marginalization, it creates an escalation of dissatisfaction.

Destruction as Status-Seeking

Research on status-seeking in humans demonstrates the existence of two distinct pathways to social

8 status: Prestige and dominance (Cheng et al., 2013). While prestige refers to attempts to receive recognition for your ability to convey benefits to others due to your talents and skills, dominance

"refers to the induction of fear, through intimidation and coercion, to attain social rank" (Cheng et al.,

2013: 105). The traits discussed above, and those of focus in the present work, all capture dominance- related strategies to status. Furthermore, in present environments, societal marginalization might in and of itself strengthen a dominance-based strategy. For individuals marginalized from the current knowledge economy, for instance, possibilities of accruing status through their skills are limited, offering them only one possible path to recognition: Dominance.

Past research consistently finds that individuals seek to assert dominance via the use or threat of aggression (Wilson & Daly, 1985). However, dominance strategies have mostly been analyzed as ways to protect the present status of individuals in leader positions or high-status groups (Cheng &

Tracy, 2014). Such individuals have been found to react with targeted reactions to, for example, insults and challengers (Cohen et al., 1996). In contrast,we are interested in understanding how dominance strategies operate among individuals at the periphery of society.

While dominance strategies often lead to targeted aggression, some evidence suggests that they may escalate into a tendency of generalized aggression and destructiveness if dominance-oriented individuals are placed in marginalized rather than central positions. For example, the cross-cultural phenomenon of "running amok" is argued to emerge in the face of severe status-loss (Hempel et al.,

2000). Similarly, in Western societies, shooting rampages have been analyzed as a response to chronic feelings of marginalization among individuals with antisocial dispositions (Kurzban & Neuberg,

2005; Twenge & Campbell, 2003). In the words of one of the shooters from the Columbine school massacre, “The lonely man strikes with absolute rage!” (Krizan & Johar, 2015: 797).

At the less extreme end, psychological studies of exclusion and aggression have shown that the combination of situational exclusion and dispositional sensitivity to rejection can lead to

9 generalized aggressive tendencies including against innocent bystanders (DeWall et al., 20016).

Similarly, reviews on the frustration-aggression hypothesis, finds that displaced aggression is a reliable phenomenon (Marcus-Newhall et al., 2000) and that it emerges in individuals with antisocial personalities when they face the continuous presence of multiple small provocations (Miller et al.,

2003; Twenge & Campbell, 2003), as would be experienced in a marginalized position.

While generalized destructive tendencies such as displaced aggression are often interpreted as a maladaptive process (Miller et al., 2003; McCullogh et al., 2011), there may be adaptive benefits to displays of destruction for the marginalized individual. First, displays of destructive tendencies may serve as hard-to-fake signals of the motivation to impose costs and, hence, operate as a general deterrence device (Frank, 1988). For example, Fessler et al. (2014) provide evidence in favor of what they term "the crazy bastard" hypothesis, which is that people who engage in extreme behaviors are perceived by others to be formidable.

Second, and relatedly, destructive tendencies may help address the adaptive problem referred to as "the Banker's paradox" by evolutionary psychologists (Tooby & Cosmides, 1996): When people are most in need, they are the least likely to be seen as valuable objects for social investments, in the same way that people who have a higher need to lend money are often, from a bank's perspective, the worst credit risks. Cultivating unique or irreplaceable skills or roles for others, which motivates these others to care when a need emerges, offers a solution to the Banker’s paradox (Tooby & Cosmides,

1996). Yet if marginalized individuals feel they have nothing to offer, destructive tendencies may serve as a radical alternative solution: a burn-down-the-house-strategy (see also Hagen, 2002). If one reacts with extreme intensity to situations of need or rejection, others may be pressured to invest immediately, if they cannot walk away. Essentially, this behavioral pattern is seen among vulnerable narcissists where minor rejections may trigger "narcissistic rage" involving disproportionate and displaced aggression (Krizan & Johar, 2015).

10 Finally, destruction may be thought of as a form of niche construction (Oldling-Smee et al.,

1996) among marginalized dominance-focused individuals: Attempts to cultivate a social ecology in which they are more likely to be successful. Antisocial individuals are better at navigating social conflict than others (Sell et al., 2009). If they can stir up chaos and conflict, antisocial individuals push status competition to a niche in which they have an advantage and, hence, may be able advance from their current social position. "Chaos isn't a pit. Chaos is a ladder", as was stated by several characters in the TV-series Game of Thrones.

In sum, when coupled with extreme feelings of marginalization, dominance-oriented strategies may flip from the use of measured and targeted aggression against specific status challengers to motivations to engage in generalized destruction in order to deter further marginalization and to advance socially by sowing chaos and conflict. Here, we contend that the indiscriminate sharing of hostile problem rumors, targeting the entire established political system, is one societal face of such motivations.

The Need for Chaos

We are interested in understanding the political and societal consequences of the psychological mindset that emerges in the interplay between situations of societal marginalization and antisocial predisposition. While extensive research has been done on the presumed situational and dispositional causes, few attempts have been made to directly distill and measure the mindset itself that emerges from the combination of situations and dispositions.1 Furthermore, we know next to nothing about

1 The literature on Competitive Jungle Worldviews comes close as this worldview has been "thought to result from the combination of a personality disposition high in tough-mindedness and exposure to social situations characterized by high levels of inequality and competition." (Perry & Sibley, 2010). Yet, here we are not interested in the effects of exposure to competition. Rather, we are interested in the effects of exposure to continuously losing competitions.

11 how motivations related to generalized destructiveness operate in politics outside of the closed circles of radicalized groups and movements (Kruglanski et al., 2014).2

We term this mindset, the Need for Chaos. We define Need for Chaos as a desire for a new beginning through the destruction of order and established structures. The term "chaos" is chosen to reflect both the modern meaning of disorder and the original Greek meaning where it refers to a state from which order is produced. Thus, the Need for Chaos is predicted to be a motivation to gain something after the disruption: status. We use the term “need” in the same way as it is used in multiple measures of other psychological individual differences (such as the need for or need for cognitive closure) in the sense of something that “...directs behavior towards a goal and cause tension when this goal is not attained” (Cacioppo and Petty, 1982, 117) Hence, it is not a biological need in the sense of thirst or hunger. Rather, the underlying evolutionary problem that the desire for chaos expresses is related to status acquisition (see Kenrick et al., 2010).

Within the vocabulary of personality psychology, we understand the Need for Chaos as a

"characteristic adaptation" (McAdams et al., 2006, 208-209), i.e., the adaptation of the behavioral tendencies inherent in dispositions to a specific cultural environment. It is not in itself a stable personality trait. Rather, it emerges in the interplay of environments related to marginalization and traits related to dominance. Accordingly, the stability of the Need for Chaos depends on the stability of its environmental triggers. A Need for Chaos may be triggered quickly in dominance-oriented individuals by situations of rejection, leading to sudden shifts, but it may also exhibit high degrees of temporal stability if dominance-oriented individuals are embedded in a situation of chronic

2 Popular culture, in contrast, has frequently examined both this mindset and its societal consequences in the form of, for example, the book (and later movie) Fight Club and the Joker character from the Batman universe. The crux of the mindset is clearly expressed in the movie The Dark Knight, where the Joker's motivations are explained with the observation that

"some men just want to watch the world burn".

12 marginalization.

Overview of Studies

In the present manuscript, we examine how the Need for Chaos is associated with motivations to share hostile political rumors. Furthermore, we use these motivations as a window into how Need for Chaos shapes political cognition in more fundamental ways. , Specifically, we conducted eight survey studies

(see Table S1 in the Supplemental Materials for a study overview). Study 1 initially constructed and validated our novel Need for Chaos scale and offered preliminary evidence that Need for Chaos predicted willingness to share hostile political rumors. Study 2 replicated Study 1 using a nationally representative, three-wave longitudinal design that further explored the construct validity and temporal stability of the Need for Chaos scale. Study 3 offered another replication of Study 1 with an alternative measure of Need for Chaos that accounted for methodological artifacts. Studies 4 and 5 tested the key argument that people high in Need for Chaos are motivated to share hostile rumors in order to disrupt the political order. Study 6 used cross-sectional survey data to show that the Need for

Chaos scale was empirically distinct from other prominent measures of antisocial traits and that an interplay of status-seeking traits and feelings of social marginalization predicted higher levels of Need for Chaos. Study 7 mounted experimental data to show that induced social marginalization increased the Need for Chaos, especially among individuals with status-seeking personality traits. Finally, in

Study 8, we examined if people who need chaos want to topple the established order and to start anew so that they can be on top of the social ladder.

13 Study 1

Study 1 had four goals. Our first goal was to use exploratory factor analysis to initially construct and validate a measure for the “Need for Chaos” characteristic adaptation, the NFCChaos scale. The second goal was to provide initial evidence that Need for Chaos is associated with destructive tendencies, using an existing scale that measures behavioral intentions of engaging in political violence. Our third goal was to test the key claim that people with a high level of Need for Chaos are motivated to share hostile rumors about politicians from both sides of the political spectrum. Finally, our fourth goal was to compare the effects of Need for Chaos to the alternative explanation that the strength of people’s partisan identity underlies their willingness to share hostile political rumors. We expect that the association between a Need for Chaos and the willingness to share hostile rumors targeting political elites operates over and beyond any effects of partisan affiliations.

Method

Participants and design. In February 2018, we recruited a convenience sample of 1,004 U.S. citizens through the TurkPrime platform to participate in our study. In our sample, 52% of the participants were female and the average age of the participant was 39.5 years old (SD = 12). For additional information on this sample and the ones that follow, see the study-specific subsections in the Supplemental Materials.

Procedure and materials. Participants completed an online survey questionnaire measuring

Need for Chaos, beliefs in and intentions to share a series of political rumors, motivations to engage in violent activism on behalf of one’s group, partisanship as well as demographic variables.

Unless otherwise noted, we scale all continuous variables in this and the following studies to range from 0 to 1, so that we can interpret the unstandardized regression coefficients as the change in

14 percentage points of the full scale of the dependent variable as we move from the low to the high extreme of the independent variable.

Measures. Need for Chaos. We fielded a pool of 11 items intended to measure a “Need for

Chaos” (see Table 1). We developed the initial item pool de novo by considering key aspects of

“chaotic” motivations, such as a desire for a new beginning (e.g., item 3), the destruction of established structures (e.g., item 4), and upsetting the established order (e.g., item 9). When formulating the item wordings, we took inspiration from popular culture such as the movies The Dark Knight (e.g., item 5) and Fight Club (e.g., item 10). For each item, we asked participants if they disagreed or agreed with what they had just read. (Response on seven-item scale: 0 = Strongly Disagree; 1 = Strongly Agree).

Hostile Political Rumors. We asked participants to read six political rumors that denigrated the political left (e.g., “Former President Obama has been creating a “shadow-government” through his non-governmental organization Organizing for Action to take down President Trump”) and six rumors that put the political right in a negative light (e.g., “Under the Republican tax bill passed in

December 2017, Medicare will stop covering cancer treatment”). For each rumor, we asked if participants disagreed or agreed (1) the rumor was true (rumor belief), and if (2) they wanted to share it on social media (rumor sharing). (Response on seven-item scale: 0 = Strongly Disagree; 1 =

Strongly Agree).

Partisan Identity. We used a modified form of the Partisan Identity Scale from Bankert,

Huddy, and Rosema (2017). We asked participants if they disagreed or agreed with six statements, like “When I speak about [Democrats/Republicans], I usually say ‘we’ instead of ‘they,’ and “When people criticize [Democrats/ Republicans], it feels like a personal insult.” (Response on seven-item scale: 0 = Strongly Disagree; 1 = Strongly Agree). They were asked to consider each statement in both the “Democratic” and “Republican” version, allowing us to create both a measure of Democratic

15 partisan identity and Republican partisan identity, scaled so higher values indicate more intense identities (αDem ID = .87, MDem ID = .37, SDDem ID = .27, αRep ID = .86, MRep ID = .29, SDRep ID = .27).

Political Violence Intentions. To measure intentions to engage in political violence, we utilized a revised version of the scale from Moskalenko and McCauley (2009) (Gøtsche-Astrup,

2019). We presented our participants with “a series of possible actions that you can carry out to promote your group’s political rights and interests. By ‘your group,’ we mean the political, religious, or social group that you identify with the most.” They were then asked if they disagreed or agreed with 15 statements. (Response on seven-item scale: 0 = Strongly Disagree; 1 = Strongly Agree). We averaged their responses to five of the statements (e.g., “I would attack police or security forces if I saw them beating members of my group”) in order to construct a measure of Political Violence

Intentions, where higher values express intentions to engage in political violence (α = .90, M = .22,

SD = .25).

Results

Is Need for Chaos a unidimensional construct? Yes. To construct our proposed measure of Need for

Chaos, we conducted a maximum likelihood exploratory factor analysis (results in Table 1). The analysis revealed one main factor with an Eigenvalue of 5.76 that explained 52.5% of the variance in the total model, and a second factor with an Eigenvalue of 1.49, explaining 13.6% of the variance.

Based on the item loadings, cross-loadings, and item clarity, we retained eight of the original items and used these to form our new unidimensional Need for Chaos-scale, labeled NFCChaos. The retained eight items appear in the last column of Table 1. The final NFCChaos scale had high internal consistency

(α = .91, M = .17, SD = .21). Perhaps not surprisingly, NFCChaos was right-skewed, suggesting that most people are reluctant to express views of burning down society.

16

Table 1. Exploratory Factor Analysis of Items for the “Need for Chaos” (NFCChaos) scale

Initial Final Item Statement Item Pool Items # Factor 1 Factor 2 Factor 1 1 ”I get a kick when natural disasters strike in foreign 0.75 -0.33 0.80 countries.” 2 “Our social institutions are rotten to the core.” 0.45 0.61 - 3 “I fantasize about a natural disaster wiping out most of humanity such that a small group of people can start all 0.81 -0.20 0.83 over.” 4 “I think society should be burned to the ground.” 0.84 -0.10 0.86 5 “When I think about our political and social institutions, I cannot help thinking "just let them all 0.81 0.15 0.78 burn". 6 “We cannot fix the problems in our social institutions, 0.76 0.41 0.69 we need to tear them down and start over. “ 7 “We need to tear down the current political institutions 0.65 0.57 - and start all over.” 8 “I prepare for a time when the military, police and the 0.63 0.29 - state no longer can protect me.” 9 “I need chaos around me - it is too boring if nothing is 0.77 -0.34 0.81 going on.” 10 “Sometimes I just feel like destroying beautiful 0.76 -0.39 0.82 things.” 11 “There is no right and wrong in the world.” 0.64 -0.31 0.67 Eigenvalue 5.76 1.49 4.92 Proportion of explained variance 52.4 % 13.6 % 64.7% Cronbach’s α .91 Notes. Principal Components Analysis used as extraction method. Unrotated factors. The final items are selected to achieve a unidimensional scale with high factor loadings and high face validity. A case could be made that item 11 should not have been included in the final set given its relatively low loading in the initial factor analysis. Considerations of construct validity, however, led us to retain item 11 (given that it formed part of a unidimensional scale). Hence, no other item captured the moral anomie specifically expressed in this item.

Does NFCChaos predict intentions to engage in political violence? Yes. To test the prediction, we regressed support for violent activism on NFCChaos while adjusting for gender, age, educational level and ethnicity. The regression model revealed that NFCChaos strongly predicted willingness to engage

17 in violent activism (β = .72, p < .001): Moving from the low to the high end on the Need for Chaos scale was associated with a 72 percentage points increase in the willingness to engage in violent activism These results provide initial evidence of our claim that a Need for Chaos as measured in a diverse sample of the normal population is strongly associated with a classical indicator of intentions to engage in highly disruptive activities.

Does NFCChaos predict motivations to share hostile political rumors? Yes. To make the best use of our data, we changed to a long data format with twelve observations per participant, one for each of the twelve hostile rumors. We then regressed motivations to share hostile rumors on NFCChaos and the same set of covariates as before. We included fixed effects for the hostile rumors and clustered standard errors by participants. We display the results in Figure 1s Panel A, which, for comparison, also shows estimated regression coefficients from the model that regressed the measure of believing that the rumors are true on NFCChaos.

Figure 1, Panel A, shows that as NFCChaos increases, people are more likely to share hostile political rumors (β = .55, p < .001). In substantive terms, for a participant with the lowest level of

Need for Chaos (i.e., NFCChaos = 0) and mean values on the covariates, the predicted motivation to share hostile rumors is exceptionally low at .11 on the 0-1 coded rumor sharing scale. Meanwhile, for people with the highest level of Need for Chaos (i.e., NFCChaos = 1) and mean values on the covariates, the motivation to share is very high at .65. Figure 1, Panel A, also reveals that NFCChaos positively correlates with believing hostile political rumors are true (β = .46, p < .001). While this correlation is high, a seemingly unrelated regression (SUR) run on the two models shows that the absolute size of the coefficient for NFCChaos is significantly larger for motivations to share hostile rumors relative to motivations to believe them (χ2 > 58.40, p < 0.001). We return to this point below.

18 Does the association between NFCChaos and motivations to share hostile political rumors depend on partisan identities? No, not in ways consistent with standard expectations from psychological theories of partisanship. To establish a benchmark, SM Section 1b examined how partisan identity predicted beliefs in hostile political rumors and intentions to share the rumors, without accounting for Need for

Chaos. To this end, SM Section 1b first regressed believing that the specific rumors were true on two measures of partisan identity: A measure of (1) identification with the rumor target (i.e.., a variable where high values indicate Republican [/Democratic] identification for rumors that target Republican

[/Democratic] elites), and (2) a measure of identification with the political opponents of the rumor target (e.g., a variable where high values indicate Democratic [/Republican] identification for rumors that target Republican [/Democratic] elites). These analyses showed, in line with the existing literature on partisanship and conspiracy theories, that partisans were less likely to believe hostile rumors targeting their “own” party (β = -.06, p = .006), and more likely to believe nefarious rumors about the out-party (β = .34, p < .001). However, as we expected given the psychological differences between believing and sharing, these findings did not extend to motivations to share rumors. Partisans were willing to share rumors about both the in-party (β = .11, p < .001) and the out-party (β = .33, p < .001).

A clear explanation for these results emerges once we account for NFCChaos and its interaction with partisan identity strength. We find positive, statistically significant interactions between NFCChaos and partisan affiliation with the rumor’s target (βNFC-Chaos X Identification w. Rumor Target = .53, p < .001) or against the rumor’s target (βNFC-Chaos X Identification w. Rumor Target Opponent = .28, p < .001). These findings, which we present in Figure 1s Panel B and Panel C, illustrate that as NFCChaos increases, participants want to share hostile political rumors irrespective of whether or not they identify politically with the rumor’s target.

19

Figure 1. Panel A: Estimated regression coefficients from models that regress (1) intentions to share and (2) believe hostile political rumors on Need for Chaos. Unstandardized OLS regressions coefficients with 95% confidence intervals (cluster-robust standard errors). Panel B and C: Marginal effect of identifying with rumor’s target and opponent of rumor’s target on intentions to share hostile rumors, conditional on Need for

Chaos. Red dots give marginal-effects from a binning estimator that discretizes Need for Chaos into tercile bins

(Hainmueller, Mummulo, Xu 2019). The density plots show the distribution of Need for Chaos. All variables in the panels range from 0 to 1. All models adjust for rumor stories, gender, age, education, and ethnicity.

Discussion

Study 1 provides evidence that our measure of Need for Chaos is a unidimensional construct and that it is associated with a desire to violently disrupt the established system. The study also shows that chaotic motivations change desires to share hostile political rumors. Individuals low in Need for Chaos

20 are only motivated to share rumors that target the opposing political party. Individuals high in Need for Chaos, in contrast, are highly motivated to share rumors, independently of whether they target their partisan team or the opposing team. The promiscuous sharing of rumors among people who possess high levels of NFCChaos supports our key claim that chaotic motivations lead people to target the entire political order rather than a particular side. Moreover, we find that chaotic motivations predict the willingness to share hostile political rumors, over and beyond believing that the rumors are true. This suggests believing hostile does not underlie both Need for Chaos and the willingness to share the rumors. Second, it suggests that people who possess high levels of NFCChaos are not sharing hostile rumors merely for epistemic reasons.

Study 2

The overarching aim of Study 2 was to replicate in a nationally representative sample the main findings from Study 1 that NFCChaos predicts intentions to share hostile political rumors and engage in political violence. In addition, we designed Study 2 to increase the confidence in our results in several additional ways. First, because of the skewed nature of responses to the NFCChaos scale, we sought to probe the robustness of the findings by introducing a progressively strict series of response set checks.

Second, Study 2 was conducted as a longitudinal study that included three survey waves, spanning ten months. This longitudinal design allows us to examine the cross-lagged relationships between

NFCChaos and intentions to share hostile political rumors, and in doing so, diminish concerns about endogeneity bias. Third, the longitudinal design also allows us to gauge the temporal stability of

NFCChaos and shed further theoretical light on the dynamics of the NFCChaos scale. Recall that our theoretical framework views Need for Chaos as a characteristic adaptation that depends on both underlying traits (e.g. status-orientated personality traits) and the social context (e.g., social marginalization). This view holds that while Need for Chaos is potentially malleable, it should be

21 relatively stable so long as the social context remains unchanged. Finally, Study 2 sought to provide an additional test of the dimensionality of the NFCChaos scale using confirmatory factor analysis to confirm the factor structure of the eight-item NFCChaos scale. Our theoretical argument is that all eight items tap the same underlying mindset, and the analyses in Study 1 supported this argument. Yet, one can reasonably object that three of the items in our NFCChaos scale (i.e., “I think society should be burned to the ground”, “When I think about our political and social institutions, I cannot help thinking

‘just let them all burn’”, and “We cannot fix the problems in our social institutions, we need to tear them down and start over”) reflect a Need for Chaos in the domain of politics whereas the other five items reflect a Need for Chaos outside of the political domain.

Method

Participants and design. In July and August 2018, we recruited 1,529 US citizens through the

YouGov survey agency. We chose the sample to be representative of the broader US population on a number of demographic characteristics, including gender, age and educational level. Within our sample, 53% of participants are women, and the average age is 45.7 years old (SD=14). In terms of education, 5% reported “Less than high school” as their highest completed degree, 26% selected

“High school graduate[s],” 27% said “Some college, but no degree,” 9% had a “2 year college degree,”

21% had a “4 year college degree,” and 11% belonged to the “Post Graduate” category. We re- contacted the participants twice, once in September 2018 and again in March 2019. With respect to attrition rates, 91% (n = 1,394) of the original sample participated in wave 2 and 89% (n = 1,367) of the original sample participated in wave 3.

Procedure and materials. We administered online questionnaires to participants in the three study waves. We measured Need for Chaos in all three waves, motivations to share hostile political rumors in the first two waves, and partisanship and violent activism only in the first survey wave. The

22 first survey also included a number of demographic variables and a number of measures intended to guard against “survey trolling” (Lopez and Hillygus, 2018) that we present below. In SM Section 2b, we show that responses to Need for Chaos in the wave 1 survey did not predict attrition rates in the subsequent survey waves.

Measures. Need for Chaos. We used the eight-item measure of NFCChaos from Study 1 in all three waves.

Hostile Political Rumors. Participants read three political rumors that denigrated the political left (e.g., “Hillary Clinton Approved Deal to Transfer US Uranium Deposits to Russian Company for

Donations to Clinton Foundation”) and three rumors that put the political right in a negative light (e.g.,

“Republican National Committee Paid People to Vote Illegally in US Senate Elections”). Like in

Study 1, we asked for each rumor if the participants disagreed or agreed (1) the rumor was true, and if (2) they wanted to share it on social media. Note that we included new rumors to rule out that the findings from Study 1 hinged on the use of particular rumors. Also note that we only included the rumor questions in survey wave 1 and 2 and, hence, our cross-lagged analyses only use data from these two waves.

Political Violence Intentions. We used the five-item measure of violent activism from Study

1. Participants only responded to the scale in survey wave 1 (α = .86, M = .26, SD = .22).

Partisan Identity. We used the same measure as in Study 1. Participants only responded to the scale in survey wave 1 (αDem ID = .86, MDem ID = .40, SDDem ID = .27, αRep ID = .86, MRep ID = .36, SDRep

ID = .26).

Results

Is Need for Chaos a unidimensional trait? Yes. A one-factor solution that allows covarying error terms between the three political items revealed a permissible fit for the factor structure (χ2 = 140,

23 Comparative Fit Index (CFI) = .99, Root Mean Squared Error of Approximation (RMSEA)= .07,

Tucker-Lewis Index (TLI) = .98). A two-factor solution with a latent political and a latent nonpolitical

NFCChaos factor revealed a poorer fit to the data (χ2 = 240, CFI = .98, RMSEA = .11, TLI = .96). And when we compared the relative fits of the two models we found that the one-factor model fitted the data significantly better than the two-factor model (|χ2one factor - χ2two factor|= 100, p < .001). The

Supplemental Materials show that a one-factor solution is also superior to a two-factor solution in the studies that follow. Together, these findings bolster our confidence that some people simultaneously prefer chaos in both political and nonpolitical domains.

Was Need for Chaos stable across the three survey waves? Yes. Panel A in Figure 2 shows participants’ levels of Need for Chaos across the three survey waves. Pearson’s r correlations shows that NFCChaos had a test-retest stability of .69 between wave 1 and 2, which matches the stability of shorter Big Five personality inventories measured at similar time intervals (Gosling, Rentfrow and

Swann Jr, 2003). Importantly, the test-retest stability is equally high between waves 1 and 3 (r = .74).

At the same time, it should be noted that individuals only temporarily maintain very high levels of

Need for Chaos (i.e., greater than top 10th percentile). These findings suggest that over this time period where the social and political context was relatively stable, Need for Chaos was also relatively stable.

24

Figure 2. Panel A: Levels of Need for Chaos among participants in the three survey waves. Grey lines give individual trajectories for panelists who completed all three survey waves. Black lines give average values across the three surveys for participants with the 1%/10%/75% highest values of Need for Chaos in Wave 1

(W1). Pearson’s RWave1-Wave2 = .69, Pearson’s RWave1-Wave3 = .74. Panel B: Estimated regression coefficients from progressively restrictive models that regress intentions to share hostile rumors on Need for Chaos and (1) adjust for covariates (baseline), (2) exclude survey trolls, (3) exclude straight-liners and (4) exclude 10% of participants with highest levels of Need for Chaos. Unstandardized OLS regressions coefficients with 95% confidence intervals (cluster-robust standard errors). All variables range 0-1.

Did Need for Chaos predict violent political intentions in survey wave 1? Yes. We followed the same procedure as in Study 1 and found that NFCChaos predicts violent political intentions (β = .62, p < .001).

25 Thus, the link between a Need for Chaos and inclinations to engage in violent acts replicated in a nationally representative sample of U.S. citizens.

Did Need for Chaos predict motivations to share hostile rumors in survey wave 1? Yes. We estimated four increasingly restrictive models that allowed us to probe the robustness of the findings. The first model followed the procedure from Study 1 and regressed motivations to share hostile rumors on

NFCChaos and a series of baseline covariates. The second model was similar, except here we followed

Lopez and Hillygus’ (2018) recommendations to guard against “survey trolling” by only retaining participants (n=944) responding “never” to the question: “We sometimes find people don’t always take surveys seriously, instead providing humorous, or insincere responses to questions. How often do you do this?” The third model additionally guarded against response set bias by excluding participants (n=39) who, on a five-point scale, gave the exact same favorability rating to Bernie

Sanders, Ted Cruz, Hilary Clinton, and Donald Trump (i.e., those who “straight-lined”, see again

Lopez and Hillygus 2018). Finally, the fourth model excluded the 10% of participants with the highest scores on the Need for Chaos scale to reduce further response set bias. We display the results in Figure

2s panel B The results reveal that the expected positive association between NFCChaos and intentions to share hostile political rumors persists as we move from the baseline model to the most restrictive model (βbaseline = .55, p < .001; β+ exclude 10% most extreme obs. = .31, p < .001). These findings suggest that the association between NFCChaos and the willingness to share hostile political rumors is not simply a function of methodological artifacts, survey trolling, or extreme responses (although the findings do suggest that individuals high in Need for Chaos, unsurprisingly, also enjoy trolling surveys).

We next used Seemingly Unrelated Regressions to replicate the finding from Study 1 that the absolute size of the coefficient for NFCChaos was significantly larger for motivations to share hostile rumors relative to motivations to believe them (χ2Seemingly Unrelated Regression > 104, p < 0.001). Finally, in

26 SM Section 2b, we replicated the positive, statistically significant interactions between NFCChaos and partisan identification with the rumor’s target (βNFC-Chaos X Identification w. Rumor Target = .23, p < .001) and with the opponent of the rumor’s target (βNFC-Chaos X Identification w. Rumor Target Opponent = .76, p < .001).. This again suggests that people who need chaos want to share hostile political rumors, irrespective of whether they identify politically with the rumor target.

Did Need for Chaos measured in wave 1 predict intentions to share hostile rumors in wave 2? Yes.

In our next analyses, we used structural equation modeling on our observed variables to examine the two-wave cross-lagged relationship between NFCChaos and motivations to share hostile political rumors in the first two waves that had data on both variables (Figure 3). While the cross-lagged effects

(controlling for other variables from wave 1) of both NFCChaos and sharing motivations were significant and positive (βNFC-Chaos = .20, p < .001; βSharing = .03, p = .007), the effect of NFCChaos is significantly stronger (p < .001). This suggests that Need for Chaos plays a stronger causal role in shaping motivations to share rumors than the willingness to share rumors play in shaping Need for

Chaos.

27

Figure 3. Cross-lagged regression analysis of the relationship between NFCChaos and motivations to share hostile political rumors at times 1 (T1) and 2 (T2). N = 1320, Observations = 7920. The unit of analysis is motivations to share a particular rumor. Analyses are performed using Structural Equation Modeling (based on observed variables). We use cluster robust standard errors with participant ID as the cluster variable (the number of unique participants in the analysis is 1,320). Consistent with the interpretation that the association between NFCChaos and motivations to share hostile political rumors is mainly driven by a causal effect from

NFCChaos on sharing motivations, we find that the cross-lagged path from NFCChaos to sharing motivations is significantly stronger than the cross-lagged path from sharing motivations to NFCChaos (difference = .16, p <

.001). The analysis, however, also suggests the existence of some endogeneity as there is a significant but small residual cross-lagged effect from sharing motivations to NFCChaos.

Discussion

Study 2 replicated the finding that Need for Chaos predicts intentions to share hostile rumors in a nationally representative sample of U.S. citizens. Further, in line with the argument that Need for

Chaos is a characteristic adaptation, the longitudinal data showed that Need for Chaos was relatively stable over time. Finally, results from the cross-lagged models revealed that Need for Chaos measured at wave 1 strongly predicted motivations to share rumor at wave 2. These results further support our

28 framework that Need for Chaos exerts a causal effect on intentions to share rumors, and that the causality does not flow in the opposite direction. Finally, in Study 2, we took a series of steps to guard against the role of response set or acquiescence bias. Nonetheless, one potential concern lingers: The fact that all the statements in the NFCChaos scale have the same valence. We sought to address this potential concern in Study 3.

Study 3

The goal of Study 3 was to replicate the finding from Study 1 and 2 that a Need for Chaos predicts sharing of hostile political rumors using an alternative NFCChaos scale that contained items with pro- and con-trait valences.

Method

Participants and design. In September 2018, we recruited a convenience sample of 1,004

U.S. citizens through the TurkPrime platform to participate in our study. Within the sample, 52% of the participants were female, and the average age of the participant was 37.9 years old (SD = 12).

Procedure and materials. Participants completed an online survey questionnaire measuring their Need for Chaos, their belief in, and willingness to share, a series of political rumors as well as demographic variables.

Measures. Need for Chaos. We included two “Need for Chaos” measures: The eight-item

NFCChaos scale from Study 1 and 2, and a ten-item NFCChaos-R version with two new reversed items in addition to the original items. We included NFCChaos-R in Study 3 to counter potential acquiescence bias induced by participants who provide affirmative answers to all questionnaire items they encounter, irrespective of the content. Thus, the two new items asked participants if they disagreed or agreed with statements where affirmative responses would indicate a “need for order”: “We need to

29 uphold order by doing what is right, not what is wrong” and “It’s better to live in a society where there is order and clear rules than one where anything goes.” See SM Section 3a for details on developing the NFCChaos-R scale. We randomly assigned participants to complete either the original NFCChaos scale

(α=.88, M = .16, SD = .19) or the NFCChaos-R scale (α= .86, M =.24, SD = .21).

Hostile Political Rumors. We measured intentions to share and believe in the six rumors from

Study 2.

Results and Discussion

Does Need for Chaos predict intentions to share hostile political rumors? Yes. We followed the procedure of Study 1 and changed our data to a long format before estimating two models, one where we regressed motivations to share hostile political rumors on our original NFCChaos scale, and one where we regressed sharing motivations on the NFCChaos-R scale. The two regression models revealed that both scales were equally strongly associated with motivations to share rumors (βNFC = .35, p <

.001 and βNFC-R = .42, p < 0.001; test of difference: βDifference = .08, p = .34). These results from Study

3 show that we can replicate the association between a Need for Chaos and motivations to share hostile political rumors in another sample, and that our findings in earlier studies are not driven by our measure of Need for Chaos.

Study 4

The three first studies established that NFCChaos predicts motivations to share hostile political rumors.

We now turn our attention to why people who need chaos are motivated to share hostile rumors. Our framework implies that people who possess high levels of NFCChaos wish to mobilize others against mainstream elites, and that sharing rumors that put these elites in a negative light is an effective tool

30 to this end. However, to make a convincing case for this argument, we must first rule out several alternative accounts for why people would want to share hostile political rumors.

In the Introduction, we discussed epistemic goals (i.e., goals related to accuracy and truth) and strategic social goals related to partisanship, which have played a significant role in research on, for example, conspiracy beliefs. If epistemic goals motivate people to share hostile political rumors, then chaos-seekers should be more likely to share hostile political rumors simply because they are more likely to believe them. The findings in Studies 1-3 offer evidence against this alternative explanation:

Need for Chaos is more strongly associated with sharing hostile political rumors than it is with believing in them. If partisan goals motivated people to share hostile political rumors, then it implies that chaos-seekers should only share rumors that target their partisan opponents. Again, the findings presented above are not consistent with this explanation: Need for Chaos predicts sharing hostile political rumors, irrespective of partisan affiliation with the rumor’s target.

Nonetheless, it is possible that chaos-seekers share rumors targeting their inparty and their outparty for different reasons. Perhaps they share rumors targeting the outparty to aggress against their opponents, and rumors targeting the inparty in order to mobilize fellow partisans to defend their team.

We address this alternative explanation in Study 4. In addition, the literature on hostile online behavior as pointed to the role of hedonic, non-instrumental goals. Thus, work on cybertrolling argues that

“trolls just want to have fun” (Buckels, Trapnell and Paulhus, 2014). We also consider this alternative explanation in this study.

Method

Participants and design. In March 2018, we recruited a diverse convenience sample of 1,105

U.S. citizens through the TurkPrime platform to participate in our study. With this sample, 59% of the participants were female, and the average age of the participant was 39.7 years old (SD = 13).

31 Procedure and materials. Participants completed an online survey questionnaire measuring

Need for Chaos, motivations for sharing particular rumors, party identification, as well as demographic variables.

Measures. Need for Chaos. We used the eight-item measure of NFCChaos from Study 1.

Motivations for sharing hostile political rumors. We presented participants with two lists of rumors, a list with the six rumors targeting Democrats from Study 1 and a list with the six rumors targeting Republicans. We presented the lists in random order. After having read the lists of hostile political rumors, we asked participants to choose the one rumor about Democrats, and the one about

Republicans they were “most motivated to share.” We then asked two sets of questions, designed to tap into the different goals outlined above. To assess the role of motivations to share for defensive purposes, we asked participants if they chose to share the particular rumors “To defend the groups or people that are negatively portrayed in the story” or “To oppose the groups or people that are negatively portrayed in the story.” (Response on seven-point item: 0= To oppose; 1 = To defend; M =

.33, SD = .30). This question tapped participants’ intentions to share the rumor to defend or attack the rumor target and, hence, allow us to examine the alternative explanation for the previous findings posed by the partisanship account. We next assessed the other alternative explanations for rumor sharing by obtaining participants’ level of agreement on 7-point scales with a number statements.

First, we sought to tap motivations directly related to mobilization intentions, as predicted by our account: “This story would be the most useful story for mobilizing people against political groups that

I oppose” (M = .55, SD = .29). We also assessed two potential epistemic reasons: “This story was the story that I believe comes closest to the truth of the different stories” (M = .68, SD = .26), and “This story contains the most important consequences, if it happens to be true” (M = .66, SD = .28). Finally, we assessed the role of hedonic goals: “This story was the story that I believe my network would find most amusing” (M = .53, SD = .31),

32 Partisan Identity. To measure partisan identity, we used a standard question from the

American National Election Studies (ANES) and asked “Generally speaking, do you usually think of yourself as a Republican, Democrat, Independent, Other?” Participants who answered either

“Republican” or “Democrat” were then asked if they saw themselves as “Strong” or “Not very strong” supporters of their party, while those answering “Independent” or “Other” were asked if they saw themselves as closer to one of the two parties. Within our sample, 367 participants identified to varying degrees as Republicans (RepStrong = 129, RepWeak = 156, RepLeaning = 82). 571 identified as

Democrats (DemStrong = 245, DemWeak = 205, DemLeaning = 121) and 167 saw themselves as pure

Independents.

Results

Does Need for Chaos lead people to share rumors to defend the rumor targets? No. To examine this, we tested how the two-way interaction between NFCChaos and partisan identification with the rumor target shapes defensive motivations. We present the findings in Figure 4s Panel A. Panel A reveals that for low levels of Need for Chaos, participants who identify politically with the political groups that are impugned by the rumor are more likely to share to defend them. This result aligns with a standard partisan account: If people share a hostile rumor that impugns the party they identify with, they do this to mobilize people on behalf of the party. However, this association is moderated by a significant, negative interaction with NFCChaos (βNFC X Identification w. Rumor Target = -0.246, p < 0.001). The association between partisanship and defensive motivations for sharing hostile political rumors evaporates and becomes statistically insignificant as NFCChaos increases (b = -.01, p = .77). For people who crave chaos, political sympathies toward the rumor targets matter little for intentions to help the target.

33

Figure 4. Panel A: Marginal Effect of Identification with Target of Hostile Rumor on Defense Motivations,

Conditional on Need for Chaos. Panel B: Estimated regression coefficients from models that regress four motivations to share hostile rumors on Need for Chaos. Unstandardized OLS regressions coefficients with 95% confidence intervals. All variables in the panels range from 0 to 1. All models adjust for rumor stories, gender, age, education, and ethnicity.

If people who need chaos share hostile rumors for other reasons than to defend the rumor target, then what motivates them? To examine this question, we regressed responses to the four questions about alternative motives for sharing hostile political rumors on NFCChaos. Figure 4 Panel B displays the results. Consistent with the cyberbullying literature, we found a marginally significant association between Need for Chaos and intentions to share hostile rumors to amuse their friends (βAmuse = 0.11,

34 p = 0.058). As we expected, intentions to share hostile rumors also reflects instrumental efforts to mobilize. Thus, people with a high level of NFCChaos also reported sharing hostile political rumors to mobilize against disliked groups, Democrats and Republicans alike (βMobilize = 0.16, p < 0.01). As we also expected, we found no evidence that these individuals share hostile rumors either because they believe them to be true — in fact, the coefficient is marginally significant and negative (βTruth = -0.09, p = .07) — or because they believe the rumors have important consequences if they turn out to be true

(βConsequence = 0.05, p = .33).

Discussion

In sum, we do not find compelling evidence that people who need chaos share hostile rumors to help defend their party’s elites or because they believe the rumors are true or important. Rather, as Need for Chaos increases, people are motivated to spread damaging information in an attempt to mobilize others against politicians in general. In doing so, they also expect to amuse their friends, but they are not just doing it “to have fun.” For individuals with high levels of NFCChaos, rumor sharing is also an instrumental act of mobilization. They seem to view sowing chaos as a means to an end.

Still, Study 4 leaves some questions open, especially related to the role of partisanship. First, while chaos-seekers do not share rumors that target their inparty as a way to defend their inparty, they may be motivated to share them as a sophisticated form of attack: To expose and mock the ludicrous outpartisans who believe these rumors. Second, we are left with the question about how we should understand the fact that people who explicitly support statements about “burning down society” at the same time express some identification with mainstream political parties. Some earlier research suggests that in contemporary society negative political sentiments are primarily directed against the political elites rather than rank-and-file voters (Druckman & Levendusky 2019). Hence, it may be that people high in Need for Chaos specifically feel anger towards partisan elites across the political

35 parties. At the same time, our theoretical account of Need for Chaos, suggests that this mindset emerges from chronic feelings of marginalization. Such feelings are likely to not only emerge from the operations of the established elites but also from larger societal dynamics. If so, individuals high in Need for Chaos may direct their anger at elites and lay people alike, as they feel abandoned by everyone, including their fellow partisans.

Study 5

To examine the two open questions from Study 4, we conducted Study 5 as a preregistered data collection (see http://aspredicted.org/blind.php?x=eq8q6r).3 Specifically, Study 5 leverages three analyses. First, utilizing a similar setup as in Study 4, we ask people directly whether they intend to share hostile rumors to hurt the rumor target or, alternatively, to ridicule the likely sender of the rumor.

Second, we examine how Need for Chaos is associated with expressions of anger towards voters and elites from the two major US political parties. In order to accomplish this goal, Study 5 fielded a novel measure of feelings of partisan abandonment.

Method

Participants and design. In February 2020, we recruited a convenience sample of 1,508 U.S. citizens through the TurkPrime platform to participate in our study. Within this sample, 49% of the participants were female and the average age of the participant was 39.3 years old (SD = 12). We

3 Specifically, we preregistered the (1) study hypotheses, (2) descriptions of the key dependent variables, (3) the statistical estimation technique, including which covariates to adjust for, (4) the precise rules for excluding participants who failed our attention checks, (5) the sample size (before applying exclusion criteria), as well as (6) a set of exploratory analyses for which we had not preregistered any hypotheses.

36 excluded 140 participants who failed our preregistered attention checks and 415 participants who identified as political “Independents” (see below).

Procedure and materials. Participants completed a questionnaire measuring their Need for

Chaos, motivations for sharing hostile political rumors, feelings towards partisan voters and elites, feelings of abandonment, and participants’ own partisanship.

Measures. Need for Chaos. We used the ten-item measure of NFCChaos-R from Study 3 (α=

.88, M =.21, SD = .19).

Motivations for sharing hostile political rumors. Consistent with Study 4, participants read a list with five hostile political rumors about Democrats and a list with five rumors targeting

Republicans, and were asked to choose the rumor from each list they were “most motivated to share.”

In Study 5, however, they were then asked to rate how accurate each of two statements reflected their motives for sharing those particular rumors about Democrats [/Republicans]: “This story would be the most useful story for mobilizing people against the corrupt Democratic [/Republican] elites” and

“This story would be the most useful story for mobilizing against lying Republicans [/Democrats]. It shows what ridiculous stuff they post” (response on seven-point scale: 0 = Very inaccurate; 1 = Very accurate).

Feelings towards partisan voters and elites. We asked participants how strongly they felt a series of emotions (anger, disgust, anxiety, contempt, enthusiasm, happiness, response on a six-point scale: 0: Not at all; 1: Very strongly) when thinking about voters of [/candidates and elected officials from] the Democratic [/Republican] party. We created four scales based on participants responses, where higher values indicate stronger negative emotions towards (1) Democratic voters (α = .79, M =

.41, SD = .24), (2) Republican voters (α = .79, M = .53, SD = .27), (3) Democratic elites (α = .79, M

= .45, SD = .24), Republican elites (α = .79, M = .58, SD = .28).

37 Partisanship. We deployed a traditional measure of partisanship from the American National

Election Survey (ANES) to construct a dummy variable indicating whether participants identified as

Republicans or Democrats. Consistent with our preregistration, we excluded non-partisans (i.e., participants who identified as Independents), leaving 590 participants identified as Democrats and

349 identified as Republicans.

Abandonment. Participants who identified as Republicans [/Democrats] were asked two sets of questions. First, they completed a standard set of questions that tapped the strength of their social identity as Republicans[/Democrats] (see Study 1). Second, they completed a novel scale with four questions about whether they felt recognized by other members of their own party (e.g., “I feel I am recognized as an equal by other Republican [/Democratic] voters,” “I feel that Republican

[/Democratic] politicians try to address the problems I have,” (Response on seven-point scale: 0 =

Strongly Disagree, 1 = Strongly Agree, Democratic Abandonment: α = .80, M = .43, SD = 16;

Republican Abandonment: α = 78, M = .49, SD = .14 ).

Results

Do partisans high in need for chaos share rumors about the in-party to expose corrupt in-party elites?

Yes. We regressed motivations to share rumors on Need for Chaos, partisan identification and the interaction between the two variables. We estimated four models: Two models examining if participants wanted to share hostile rumors about Democrats to “hurt them” or to “expose lying”

Republicans, and two models examining if people share rumors about Republicans to hurt them or to expose Democrats. The models adjusted for gender, age, ethnicity and education. In all four models, we found a statistically significant interaction between partisanship and Need for Chaos (all p’s < .03).

Figure 5 displays the substantive results by plotting the predicted values of motivations to share rumors across NFCChaos-R for Democratic and Republican identifiers, respectively. The figure shows

38 that for low levels of Need for Chaos, people behave like ardent partisans: They intend to share rumors about their in-party to expose their “lying political opponents” and to share rumors about the out-party to “hurt them.” As expected, however, partisans who score high on NFCChaos-R have only one goal in mind when sharing hostile political rumors: They want to hurt elites, irrespective of whether they belong to the in- or out-party.

Figure 5. Predicted values with 95% confidence intervals for motivations to share rumors about Democrats and Republicans across NFCChaos-R. Blue lines: Democratic identifiers. Red Lines: Republican identifiers.

Predicted values are based on OLS regression models. The vertical and horizontal lines to the left and at the bottom present the jittered distribution of the dependent variables and Need for Chaos, respectively. All variables range 0-1. All models adjust for gender, age, ethnicity, education.

39 Do partisans high in need for chaos dislike political elites and voters from both political parties? Yes.

We estimated four models that regressed strength of negative feelings towards Democratic and

Republican elites and voters, respectively, on Need for Chaos, partisan identification and the interaction between the two variables. In each of the four models, we found a statistically significant interaction between partisanship and Need for Chaos (all p’s < .03). Figure 6 displays the substantive results by plotting the predicted feelings towards the four groups across NFCChaos-R for Democratic and Republican identifiers, respectively. We find that partisans with a low need for chaos hold a positive view of in-party actors, both voters and elites, and a negative view of actors from the out- party. These results are similar to those from the analysis above. Yet, again, we find that partisans with strong chaotic motivations display a very different mixture of views on party elites and voters.

“Chaotic” partisans evaluate in-party actors, both voters and elites, just as negatively as they evaluate out-party actors. This finding supports our argument that people who are high in Need for Chaos want to target all political elites, but it also goes further than that. It seems that people high in need for chaos simply dislike everyone, both elites and voters and from both parties. Next, we examine whether these findings reflect feeling abandoned by their own party.

40

Figure 6. Predicted values with 95% confidence intervals for negative emotions towards partisan elites and voters across NFCChaos-R. Blue lines: Democratic identifiers. Red Lines: Republican identifiers. Predicted values are based on OLS regression models. The vertical and horizontal lines to the left and at the bottom present the jittered distribution of the dependent variables and Need for Chaos, respectively. All variables range 0-1. All models adjust for gender, age, ethnicity, education.

Do people high in need for chaos feel abandoned by their political party? Yes. We conducted two exploratory analyses. First, we examined whether partisans higher in Need for Chaos also had higher levels of social identification with their respective political party. Regression results where we pool participants who identify as Democrats and Republicans reveal that Need for Chaos does indeed

41 positively predict the strength of the partisan identification (β = .15, p < .001). Next, we conducted similar analyses for our abandonment scale. Again, regression results show that Need for Chaos positively predicts feelings of abandonment by fellow partisans (β = .10, p < .001). Individuals high in Need for Chaos are motivated to invest in their party but also feel that their fellow partisans turn their backs to them.

Discussion

Study 5 demonstrated how individuals high in Need for Chaos can, on the one hand, see themselves as partisans and, on the other hand, be motivated to indiscriminately share hostile political rumors.

Taken together, the evidence from Study 4 and 5 shows that indiscriminate sharing does not reflect a complex strategy for helping their inparty by spurring defensive action from fellow partisans or by mocking members of the outparty for believing silly things. Rather, partisans who are high in Need for Chaos feel abandoned by their fellow partisans. Thus, they share hostile political rumors about their inparty, because they are fed up with their party, elites and voters alike.

These findings provide the first piece of evidence of how a Need for Chaos is connected to feelings of marginalization — in this case, the feeling of being rejected by one's party. According to our theoretical argument, though, among individuals high in Need for Chaos, feelings of marginalization should go deeper than simply being abandoned by members of their political party.

These feelings should reflect a general, pervasive sense of being rejected by society. Coupling this deep sense of marginalization with a disposition for social dominance triggers Need for Chaos. We test this expectation in Study 6.

42 Study 6

The overall aim of Study 6 was to empirically test our claim that NFCChaos, as a characteristic adaptation, is the product of dominance-oriented, status-seeking traits and feelings of social marginalization. In doing so, we must first assess the divergent validity of NFCChaos and demonstrate that it is distinct from existing measures of dominance-related traits that have a prominent place in the psychological literature. Specifically, we focus on the most antisocial of the Dark Triad subfacets, i.e.,

Psychopathy (Jonason et al. 2009), which measures impulsivity, selfishness and lack of empathetic concern. We also employ Status-Driven Risk-Taking (Ashton et al. 2010), which is designed to measure the pursuit of social gains when facing risk. We utilize two dominance-oriented measures that have a prominent role in the literature: Competitive Jungle Worldview

(Duckitt 2001) and Social Dominance Orientation (Ho et al. 2015), which tap perceptions of society as a highly competitive place and preferences for group-based hierarchies, respectively.4 Finally, we provide evidence that while NFCChaos relates to both status-seeking and marginalization, NFCChaos is the proximate psychological cause of motivations to share hostile political rumors.

Method

Participants and design. In April 2018, we recruited a convenience sample of 1,506 U.S. citizens through the TurkPrime platform to participate in our study. Within the sample, 51% of the participants were female, and the average age of the participant was 37.4 years old (SD = 12).

Procedure and materials. Participants completed a questionnaire measuring their Need for

Chaos, motivations for sharing hostile political rumors, four measures of status-seeking and two measures of social marginalization.

4 Below we provide formal tests of divergent validity using confirmatory factor analyses. In addition, Figure S1 in the Supplementary Materials contain a correlation matrix showing the correlations between Need for Chaos and a larger range of psychological constructs measured in Study 6.

43 Measures. Need for Chaos. We used the eight-item measure of NFCChaos from Study 1 (α =

.94, M = .22, SD = .25).

Hostile Political Rumors. We used the six hostile political rumors from Study 2.

Measures of Status-Seeking. We measured psychopathy using Jonason et al.’s (2009) nine- item scale (e.g., “People who mess with me always regret it,” response on seven-point scale: 0 =

Strongly Disagree; 1 = Strongly Agree. α = .86, M = .31, SD = .24). We measured competitive jungle worldview using Duckitt’s (2001) 13-item scale (e.g., “Winning is not the first thing; it's the only thing,” response on seven-point scale: 0 = Strongly Disagree; 1 = Strongly Agree. α = .86, M = .34,

SD = 18). We measured social dominance orientation using Ho et al.’s (2015) eight-item scale (e.g.,

“-Some groups of people are simply inferior to other groups,” response on seven-point scale: 0 =

Strongly Disagree; 1 = Strongly Agree. α = .90, M = .29, SD = .23). Finally, we measured Status-

Driven Risk-Taking using Ashton et al.’s (2010) a fourteen-item scale (e.g., “I would like to live in a country where people who take huge risks have the chance to gain superior social status,” response on seven-point scale: 0 = Strongly Disagree, 1 = Strongly Agree. α = .91, M = .30, SD = .21).

Measures of Social Marginalization. We measured loneliness using Hays and DiMatteo’s

(1987) seven-item scale (e.g., “I feel isolated from others,” response on four-point scale: 0 = Never, 1

= Always. α = .90, M = .34, SD = .26). We measured perceived social status by asking participants to place themselves on a ten-step ladder, where people at the top of the ladder (step 10) are those in society that are best off in terms of job security and respect, and where people at the bottom of the ladder (step 1) are those that are worst off (M = .57, SD = .18). See Adler et al. (2010).

44 Results

Is Need for Chaos distinct from measures of dominance-oriented status-seeking? Yes. A one-factor solution that allowed all items from NFCChaos and the psychopathy scale to load on the same latent trait revealed a poorer fit for the factor structure (χ2 = 1848, CFI = .91, RMSEA = .10, TLI = .90) than a two-factor solution with a latent NFCChaos factor and a latent psychopathy factor (χ2 = 708, CFI

= .97, RMSEA = .06, TLI = .97, |χ2one factor - χ2two factor|= 1140, p < .001). In addition, we emphasize that the two-factor solution reveals only a moderately good fit to the data primarily because the items reflecting psychopathy do not load well on the proposed underlying psychopathy trait. Still, the results reveal that NFCChaos displays good divergent validity and clearly differs from a related maladaptive trait. Similarly, a one-factor solution that allows all items from NFCChaos and Social Dominance

Orientation (SDO) to load on the same latent trait revealed a poorer fit for the factor structure (χ2 =

4573, CFI = .76, RMSEA = .17, TLI = .71) than a two-factor solution with a latent NFCChaos factor and a latent SDO factor (χ2 = 1192, CFI = .94, RMSEA = .09, TLI = .93, χ2one factor - χ2two factor = 3380, p < .001). Also, a one-factor solution that allows all items from NFCChaos and the Competitive Jungle

Worldview (CJW) to load on the same latent trait revealed a poor fit for the factor structure (χ2 =

5594, CFI = .74, RMSEA = .15, TLI = .70) compared to a two-factor solution with a latent NFCChaos factor and a latent CJW factor(χ2 = 3548, CFI = .84, RMSEA = .12, TLI = .81, |χ21 factor - χ2|2 factor =

2046, p < .001). Finally, a one-factor solution that allowed all items from NFCChaos and the status- driven risk-taking (SDRT) scale to load on the same latent trait revealed a poor fit for the factor structure (χ2 = 6240, CFI = .74, RMSEA = .105 TLI = .70) relative to the two-factor solution with a latent NFCChaos factor and a latent SDRT factor(χ2 = 2796, CFI = .88, RMSEA = .10, TLI = .87, |χ21 factor - χ2|2 factor = 3444, p < .001).

45 Is Need for Chaos associated with status-seeking traits? Yes. We estimated a model where we regressed NFCChaos on four measures of status-seeking – Competitive Jungle Worldview (CJW),

Social Dominance Orientation (SDO), Psychopathy, Status-Driven Risk-Taking (SDRT) – while adjusting for gender, age, and education. Three of the four measures – SDO, Psychopathy, SDRT – had positive and statistically significant associations with NFCChaos (βPsychopathy = .69, p < .001, βSDRT=

.13, p < .001, βSDO = .07, p = .002) while CJW had a negative and statistically insignificant association with NFCChaos (βCJW = -.05, p = .155). Taken together, these findings provide strong evidence that people who seek status are also more likely to need chaos around them.

Is Need for Chaos associated with social marginalization? Yes. We regressed NFCChaos on perceived social status and loneliness, while adjusting for gender, age and education. We found that higher levels of loneliness are positively associated with a need for chaos (βLoneliness = .43, p < .001), while perceived social status is negatively associated with a need for chaos (βPerceived Social Status = -.40, p < .001). Thus, while people high in need for chaos strive for status, they feel lonely and they perceive themselves to be part of society’s lower echelons.

Does the interaction between status-seeking and social marginalization predict Need for Chaos? Yes.

Analyses in the SM Section 6b examine how the four status-oriented traits interact with the two measures of social marginalization in predicting need for chaos. For illustrative purposes, we focus on the most antisocial of the traits, psychopathy. Figure 7 Panels A and B displays the interaction between psychopathy, on the one hand, and perceived social status and loneliness, on the other hand.

We find a strong positive interaction between psychopathy and loneliness (βLoneliness X Psychopathy = .99, p < .001), suggesting that loneliness only promotes a Need for Chaos for those who strive for status and who have the cognitive skills to navigate anti-social situations (e.g., a lack of empathy). Similarly,

46 this implies that psychopathy only leads to a Need for Chaos if individuals feel marginalized. The same pattern holds for the interaction between psychopathy and perceived social status. The strongly negative and statistically significant association (βLoneliness X Psychopathy = -.91, p < .001) suggests that perceived social marginalization influences a Need for Chaos only to the extent it is coupled with a particular strive for status.

Figure 7. Panel A and B: Marginal Effect of loneliness and perceived social status on Need for Chaos, conditional on Psychopathy. Panel C: Estimated regression coefficients with 95% confidence intervals from models that regress motivations to share hostile political rumors on Need for Chaos. The models adjust for different sets of variables. The baseline model adjust for gender, age, education, base + marginalization additionally adjust for loneliness and perceived social status, base + status-seeking additionally adjusts for Social Dominance Orientation, Status-Driven Risk-Taking, Competitive Jungle Worldview and Psychopathy, while base + status + margin additionally adjust for both sets of marginalization and status-seeking variables. All variables in the panels range from 0 to 1.

47 Is Need for Chaos the proximate psychological cause of intentions to share hostile political rumors?

Yes. To test this claim we estimate four models that regressed motivations to share hostile political rumors on NFCChaos. The first model adjusts for baseline covariates (gender, age, education), the second model additionally adjusts for the social marginalization variables, the third model adjusts for the status-seeking variables, while the fourth model adjusts for both the social marginalization and the status-seeking variables. We display the results in Panel C of Figure 7. The panel shows that while the strength of the association between Need for Chaos and intentions to share hostile rumors understandably drops as we progressively adjust for status-seeking and social marginalization

(βBaseline= .76, p < .001; βbase + status + margin= .53, p < .001), Need for Chaos remains a strong and robust predictor of motivations of sharing rumors.

Discussion

Study 6 provides evidence in favor of our theoretical model that conceptualizes Need for Chaos as a characteristic adaptation that emerges from the interplay of a disposition towards status-seeking and situational feelings of status loss. Thus, Need for Chaos is distinct from established measures of antisocial, status-seeking and it better captures the proximate motivations to share hostile political rumors. At the same time, Study 6 was observational and, hence, we cannot rule out the possibility that the measures of marginalization that we used are dispositional rather than situational. To remedy this potential concern, Study 7 employed an experimental design that directly manipulated situational feelings of social marginalization.

Study 7

Study 7 was designed to test whether experimentally induced social marginalization influences participants’ Need for Chaos as well as their intentions to share hostile political rumors. To this end,

48 we rely on the prominent psychological Cyberball program. In addition, this provides a further opportunity to test if status-seeking interacts with situational exclusion on a Need for Chaos. To measure status-seeking, we focus on the most antisocial of the existing measure, specifically the Dark

Triad subfacet of Psychopathy.

Method

Participants and design. In November 2018, we recruited a convenience sample of 1,506

U.S. citizens through the TurkPrime platform to participate in our study. Within the sample, 55% of the participants were female, and the average age of the participant was 37.4 years old (SD = 12).

Procedure and materials. Participants first completed a pre-treatment questionnaire including sociodemographic questions and a psychopathy scale before completing a three-minute version of Cyberball. Cyberball is an “[…] online ball-tossing game that participants believe they are playing with two or three others” (Williams and Jarvis, 2006, 174), but where the “other players” are controlled by the programmer to either socially exclude (i.e., avoid throwing the ball to the human player) or socially include (i.e., throw the ball to the human player) the human participant. We randomly assigned participants to the “social inclusion”- or “social exclusion”-condition. After completing the game, participants reported their Need for Chaos and their intentions to share hostile political rumors

Measures. Need for Chaos. We used the eight-item measure of NFCChaos from Study 1.

Participants responded to the questions after participating in the Cyberball game.

Hostile Political Rumors. We measured intentions to share and believe in the rumors from

Study 2. Participants responded to the questions after participating in the Cyberball game.

Psychopathy. We used the nine-item measure of Psychopathy from Study 6. This measure was obtained before participants participated in the Cyberball game.

49

Results and Discussion

Did experimentally induced social marginalization increase a need for chaos and motivations to share hostile political rumors? Yes. Need for Chaos was significantly higher among participants who were socially excluded in the Cyberball game compared to participants who played the socially inclusive version (β = 0.014, p = .028). Similarly, we found higher motivations to share hostile rumors in the exclusion condition, although the effect was only marginally significant (β = 0.011, p = .066). Further, we emphasize that social exclusion had only substantively small effects on both dependent variables.

This finding likely reflects the mild treatment.

Did the interaction between experimentally induced social marginalization and psychopathy increase a need for chaos? Yes. Consistent with findings from Study 6, we found a positive, statistically interaction between social marginalization and psychopathy on Need for Chaos (βMarginalization X

Psychopathy = .09, p = .009). These findings highlight the importance of both perceived status loss and status aspirations in fueling a Need for Chaos.

Study 8

Studies 6 and 7 provided evidence of the causes of Need for Chaos. Our final, preregistered study (see http://aspredicted.org/blind.php?x=z4w3dj)5 was designed to focus on the effects of Need for Chaos:

5 Specifically, we preregistered the (1) study hypotheses, (2) descriptions of the key dependent variables, (3) the statistical estimation technique, including which covariates to adjust for, (4) the precise rules for excluding participants who failed our attention checks, (5) the sample size (before applying exclusion criteria), as well as (6) a set of exploratory analyses for which had not preregistered any hypotheses.

50 What exactly are people high in Need for Chaos seeking to politically achieve when they, for example, indiscriminately share hostile political rumors? In essence, what are their political ideals? While many status-oriented traits lead people to engage in system-justification, we expected individuals high in

Need for Chaos to reject the existing system and, hence, be low in system-justification. Furthermore, we expected them to politically support political changes that would elevate their status. In the studies above, we have assessed support for radical solutions such as political violence and rumor-sharing.

Here, we assessed support for a range of softer hierarchy-attenuating policies (Zimmerman & Reyna

2013). More importantly, we also directly assessed their political ideals. Specifically, we developed novel measures that tapped the extent to which they support political changes that would benefit only them or would benefit everyone. We expected individuals high in Need for Chaos to be oriented towards their own standing and be less oriented towards the standing of others. In this sense, individuals high in Need for Chaos differ from other disruptive (yet utopian) ideologies, such as anarchism, that seek to make a better world for everyone. Individuals high in Need for Chaos are nihilists who are out for themselves. To highlight this aspect of the chaos-seeker's psychology, we also obtained measures of anomic meaninglessness, i.e., the feeling that life is without any purpose

(Yang, 2015).

Method

Participants and design. In February 2020, we recruited a convenience sample of 1,508 U.S. citizens through the TurkPrime platform to participate in our study. Within this sample, 49% of the participants were female, and the average age of the participant was 39.3 years old (SD = 12). We excluded 265 participants who failed our preregistered attention checks.

Measures. Need for Chaos. We used the eight-item measure of NFCChaos from Study 1 (α =

.92, M = .18, SD = .22).

51 Psychopathy. For comparative purposes, we also obtained a measure of an antisocial, status- seeking trait, namely Psychopathy. We used the nine-item scale from Study 6 (α = .84, M = .26, SD

= .22).

System Justification. We measured system justification using Key and Jost’ (2003) eight-item scale (e.g., “Society is set up so that people usually get what they deserve”; response on seven-point scale: 0 = Strongly Disagree; 1 = Strongly Agree. α = .88, M = .57, SD = .21).

Hierarchy-attenuating policies. We measured preferences for hierarchy-attenuating policies using Zimmerman & Reyna’s (2013) eleven-item scale (e.g., “The government should increase support for people receiving food stamps”; response on seven-point scale: 0 = Strongly Disagree; 1 =

Strongly Agree. α = .92, M = .65, SD = .22).

Self- versus other-idealism. We measured self- versus other-idealism with two five-item batteries that only differed in whether one’s ideal society benefits oneself or everyone (e.g., self- idealism [other-idealism]: “My ideal society is a society where people like me [everyone] finally get the rewards we deserve”, response on seven-point scale: 0 = Strongly Disagree, 1 = Strongly Agree.

Self-idealism: α = .94, M = .60, SD = .24; Other-idealism: α = .94, M = .80, SD = .20).

Meaninglessness. We measured meaninglessness using Yang’s (2015) seventeen-item scale

(e.g., “No matter how hard I try in life, it does not make any difference,” response on five-point scale:

0 = Strongly Disagree, 1 = Strongly Agree. α = .94, M = .33, SD = .23)

Results

Are individuals high in Need for Chaos lower in system-justification relative to individuals high in psychopathy? Yes. We regressed system justification attitudes on Need for Chaos, psychopathy, and a series of covariates. In this way, we are able to assess any effects of psychopathy that exist over and beyond Need for Chaos. The main finding was that Need for Chaos and psychopathy had opposite

52 associations with system justification beliefs. When adjusting for psychopathy, need for chaos was associated with lower levels of system justification (β = -.10, p = .002), while psychopathy, when adjusting for Need for Chaos, was associated with higher levels of system justification (β = .12, p =

.001). This difference in the associations of Need for Chaos and psychopathy, respectively, is statistically significant (p < .001).

Are individuals high in Need for Chaos higher in support for hierarchy-attenuating policies relative to individuals high in psychopathy? Partly. We found that Need for Chaos was not statistically significantly associated with hierarchy-attenuating policies (β = .02, p = .503), while psychopathy had a clear, negative association (β = -.14, p < .001). This difference in the associations of Need for Chaos and psychopathy, respectively, is statistically significant (p = .004) and suggests that individuals high in Need for Chaos are more oriented towards breaking down existing hierarchies compared to those who are high in psychopathy but that soft policy solutions do not strongly cater to the motivations of individuals high in Need for Chaos.

Is Need for Chaos positively associated with ideals of improving society to benefit oneself, but negatively associated with ideals of improving society to benefit others? Yes. We estimated two models where we regressed our self-idealism and other-idealism scales on Need for Chaos and a series of covariates. Figure 8s Panel A and B display the results. The panels reveal that people high in Need for Chaos were more likely to support a society that benefited themselves than were people low in

Need for Chaos (Self-Idealism: β = .14, p < .001). In contrast, people with a high Need for Chaos were less likely to prefer a society that benefited others (Other-Idealism: β = -.17, p < .001).

53 Is Need for Chaos positively associated with meaninglessness? Yes. We regressed the meaninglessness scale on Need for Chaos and a series of covariates. Figure 8s Panel C displays the results. The panel shows that people with a high Need for Chaos were much more likely to report a high level of meaninglessness (β = .58, p < .001).

Figure 8. Predicted levels of other-idealism (Panel A), self-idealism (Panel B) and meaninglessness

(Panel C) across levels of Need for Chaos. Results from models that regressed the three dependent variables on Need for Chaos and that adjust for gender, age, race, education. Shades areas indicate

95% intervals. Red dots give predicted values from a binning estimator. All variables scaled to range from 0 to 1.

54 Discussion

Study 8 provided evidence that individuals who are high in Need for Chaos are less likely to engage in system-justification compared to status-seeking individuals without this mindset. Study 8 also provided suggestive evidence that individuals high in Need for Chaos are somewhat more likely to support hierarchy-attenuating policies than status-seeking individuals without this mindset.

Importantly, however, the desire of chaos-seekers to disrupt present status-hierarchies do not reflect idealistic ideological motivations. In contrast to many other radicalized individuals (Kruglanski et al.,

2014), individuals high in Need for Chaos do not see much meaning in the world. When they say they want to watch the world burn, they are expressing a desire to smash the established hierarchy so that they have a chance of ending up on top. The soft hierarchy-attenuating policies that we assessed level of support for as part of this study are seemingly not disruptive enough to cater to this desire.

Summary and Concluding Discussion

We have provided a comprehensive and in-depth psychological assessment of motivations to share hostile political rumors. Consistent with past conjectures (Allcott and Gentzkow, 2017; Spohr, 2017), we do find evidence that motivations to share hostile political rumors operate via a partisan logic in which partisans seek to aid their party against their mainstream opponents. Nonetheless, sharing hostile political rumors is not exclusively motivated by a desire to aid co-partisans. An element of the

US public indiscriminately shares hostile political rumors because they want to tear down the established system. Among this element, the normal partisan logic of sharing hostile political rumors does not apply. We show that these individuals are motivated by an underlying characteristic adaptation, which we call Need for Chaos, that emerges from the interplay between dominance- oriented traits and marginalized states (Studies 6-7). We find that Need for Chaos is associated with support for political violence (Studies 1-2), anger towards elites and people of all political allegiances

55 (Study 5), dissatisfaction with the political system, and a craving for personal status (Study 8). These individuals are not idealists seeking to tear down the established order so that they can build a better society for everyone. Rather, they indiscriminately share hostile political rumors as a way to unleash chaos and mobilize individuals against the established order that fails to accord them the respect that they feel they deserve.

Limitations

Two caveats are in order. First, most of our findings are observational. As a result, we cannot make firm conclusions about the direction of causality. At the same time, we have taken steps to increase the internal validity of our findings by (a) employing a panel design in Study 2, (b) using an experimental design in Study 7, (c) controlling for multiple potential confounds related to both political dispositions, psychological needs and malevolent personality traits, and (d) verifying the robustness of the results when accounting for survey trolls, response set bias, and extreme responses.

Second, we offer a study on self-reported psychological motivations, which is a methodological limitation insofar as we do not know whether these self-reports translate into actual behavior. For example, on the basis of these surveys, we cannot — and we do not — claim that substantial numbers of American citizens are ready to physically clash with the police or commit other forms of political violence. While this is a methodological limitation common in psychological research, we also see it as a substantive strength. Our study provides insights into the kinds of thoughts and behaviors that people are motivated to entertain when they sit alone (and, perhaps, lonely) in front of the computer, answering surveys or surfing social media platforms. In an age of hostile political rumors, behavior aimed at disrupting the system does not require much more than that. A few chaotic thoughts that lead to a few clicks to retweet or share is enough. When the echoes of similar processes across multiple individuals reinforce each other, it can add up to cascades of hostile political rumors.

56

How Widespread is the Need for Chaos?

Every society contains discontented radicals. In the age of social media, however, these radicalized individuals can more easily find like-minded others and can more easily share their views.

Furthermore, societies of today are facing specific sociopolitical developments that might spur a desire for chaos to a greater extent than in previous decades (Turchin, 2016). While massive sociopolitical progress has occurred over the past centuries (Pinker, 2012), Western democracies are currently facing an upsurge in inequality and perceived marginalization. We expect that these phenomena serve as catalysts for a Need for Chaos, because they simultaneously activate and frustrate status-striving.

Importantly, this dynamic does not necessarily imply that Need for Chaos only impels individuals of lower economic classes to revolt against the system. Increased inequality intensifies status- competitions across the entire status-hierarchy (Turchin, 2016), and it can induce even those who are objectively well off to feel that they are losing ground as other surge pass them.

Figure 9. Negative cumulative distributions for three indicators of Need for Chaos in Study 2 (a nationally representative sample of the U.S. weighted by gender, age, education and geography). N = 1,529. The scale in the x-axis ranges from Strongly Agree (1) to Strongly Disagree (5).

57

In Figure 9, we display the negative cumulative distributions (i.e., how much of the distribution is more or equally in agreement with the statement than the given response) of the three most political

NFCChaos indicators in the US. The statements are strong, expressing a support for social and political institutions to “burn”, “be burned to the ground”, and “[torn] down.” This type of distribution shows what percentage of the American public does not reject (i.e., disagree or disagree strongly with) the three statements. From our perspective, these percentages are considerable: 24%, 40% and 40%, respectively.

Implications for Understanding and Countering the Spread of Hostile Political Rumors

It appears that a substantial minority of Americans are so discontent with society that they cannot be bothered to oppose statements about the destruction of its institutions and, accordingly, are motivated to share hostile political rumors. This finding challenges three common assumptions about the nature of hostile political rumors in today’s political environment. First, hostile online behavior does not simply reflect a motivation to “have fun” (Buckels, Trapnell and Paulhus, 2014) — some people spread vile lies and rumors as an explicitly instrumental act that sows chaos as a means to an end.

Second, because some people spread hostile political rumors as a way to create chaos, as opposed to aiding a particular partisan team, these types of rumors may circulate even in the absence of a deeply partisan polity (Spohr, 2017). Third, because chaos-motivated sharers are not interested in the truth of the rumors they help spread, fact-checking will provide an incomplete solution to stemming the tide of "fake news" and conspiracy theories (Nyhan and Reifler, 2010). Essentially, our findings suggest that policies aimed at reducing the circulation of conspiracy theories and the like must go beyond nudges, such as fact-checking and the regulation of algorithms, and also alleviate the underlying feelings of discontent that seems to have become more prevalent over the past few decades.

58

Directions for Future Research

The present set of studies are meant as a first identification of the tendencies associated with a Need for Chaos and, in particular, as an examination of its association with motivations to share hostile political rumors. Accordingly, there are several research questions that are outstanding and which will be important to address in future research.

First, the present analyses suggest that individuals high in Need for Chaos do not hold idealistic political visions in the typical sense of the term. Rather, they focus on achieving status and recognition for themselves. In this regard, the measure of Need for Chaos seems to operate in different ways than other established measures of anti-social tendencies such as competitive jungle worldviews and social dominance orientation, which have been consistently shown to be associated with higher degrees of system-justification and a right-wing ideology. In contrast, Need for Chaos is associated with lower degrees of system-justification. Furthermore, additional analyses presented in the Supplementary

Materials for Study 6 (Table S18) show that a Need for Chaos is unrelated to political ideology relative to other dominance-oriented traits (competitive worldviews, social dominance orientation and psychopathy). Nonetheless, even though most individuals high in Need for Chaos do not share a single political vision, each individual chaos-seeker may still justify his or her actions through adherence to a disruptive, ideological vision placed at either the far-right or the far-left, depending on individual circumstances. Hence, future analyses should probe deeper into the political visions of individuals of

Need for Chaos.

Second, empirically and theoretically, we have focused on people’s assessment of their own social status in the form of feelings of loneliness and personal social exclusion. A key question, however, is whether and how these feelings of personal marginalization are influenced by group-level dynamics in contemporary society. A prominent hypothesis in social psychology is that people’s

59 personal self-esteem is directly related to the social standing of the groups that these people identify with (Rubin & Hewstone, 1998). In the present studies, there are several strands of research suggesting that individuals high in Need for Chaos are group-oriented. They express higher willingness to engage in political violence for “their group” (Studies 1 and 2) and higher levels of social identification with political parties (Study 5) than people low in Need for Chaos. Thus, a key focus for future research is to look into whether feelings of threat to gender- or ethnicity-based groups can also operate as an elicitor of a Need for Chaos among dominance-oriented individuals. Potentially, this can explain some of the present-day polarization among groups traditionally believed to be dominant such as white males (Bai & Federico, 2019; Kaufmann, 2018).

Finally, we leave open the question of whether a Need for Chaos may be an adaptive, potentially evolved, strategy for extremely marginalized individuals. To begin to illuminate this, there are a range of hard questions that need to be addressed thoroughly: How culturally widespread is the phenomenon of chaotic sentiments; to what extent does it emerge in small-scale societies; do people immediately recognize it in others; how do signals of chaotic sentiments influence impression formation; and to what exact does it’s elicitation, in fact, lead to an elevated status? The destructiveness inherent in the elicitation of a Need for Chaos suggests that it may be a maladaptive byproduct of an overactivated psychology of dominance-striving. At the same time, the seemingly universality of destructive rampages (captured in “running amok”, shooting sprees etc.) raises the possibility that a Need for Chaos may reveal the existence of an previously overlooked adaptation for status-seeking under extreme circumstances.

References

Allcott, Hunt and Matthew Gentzkow. 2017. “Social media and fake news in the 2016 election.”

Journal of Economic Perspectives 31(2):211–36.

60 Arceneaux, Kevin and Ryan J Vander Wielen. 2017. Taming Intuition: How Reflection Minimizes

Partisan Reasoning and Promotes Democratic Accountability. Cambridge: Cambridge

University Press.

Altay, S., Hacquin, A., & Mercier, H. (2019, October 1). Why do so Few People Share Fake News?

It Hurts Their Reputation. https://doi.org/10.31234/osf.io/82r6q

Ashton, M. C., Lee, K., Pozzebon, J. A., Visser, B. A., & Worth, N. C. (2010). Status-driven risk

taking and the major dimensions of personality. Journal of Research in Personality, 44(6),

734-737.

Balmas, Meital. 2014. “When fake news becomes real: Combined exposure to multiple news sources

and political attitudes of inefficacy, alienation, and cynicism.” Communication Research

41(3):430–454.

Bai, Hui, and Christopher M. Federico. 2019. "Collective existential threat mediates White population

decline’s effect on defensive reactions." Group Processes & Intergroup Relations:

1368430219839763.

Bankert, Alexa, Leonie Huddy and Martin Rosema. 2017. “Measuring partisanship as a social identity

in multi-party systems.” Political behavior 39(1):103–132.

Bartusevicius, H., van Leeuwen, F., & Petersen, M. (2020). Dominance-driven autocratic political

orientations predict political violence in non-WEIRD and WEIRD samples. Psychological

Science, forthcoming.

Bastos, Marco T and Dan Mercea. 2017. “The Brexit botnet and user-generated hyperpartisan news.”

Social Science Computer Review p. 0894439317734157.

Björkqvist, K., Österman, K., & Lagerspetz, K. M. (1994). Sex differences in covert aggression

among adults. Aggressive behavior, 20(1), 27-33.

Buckels, Trapnell and Paulhus, 2014

61 Cacioppo, John T. and Richard E. Petty. 1982. “The Need for Cognition.” Journal of Personality and

Social Psychology 42:116–131.

Campbell, Angus, Philip E Converse, Warren E Miller and Donald E Stokes. 1960. The American

Voter. Chicago: University of Chicago Press.

Cheng, J. T., & Tracy, J. L. (2014). Toward a unified science of hierarchy: Dominance and prestige

are two fundamental pathways to human social rank. In The psychology of social status (pp.

3-27). Springer, New York, NY.

Cheng, J. T., Tracy, J. L., Foulsham, T., Kingstone, A., & Henrich, J. (2013). Two ways to the top:

Evidence that dominance and prestige are distinct yet viable avenues to social rank and

influence. Journal of personality and social psychology, 104(1), 103.

Cichocka, A., Marchlewska, M., & De Zavala, A. G. (2016). Does self-love or self-hate predict

conspiracy beliefs? Narcissism, self-esteem, and the endorsement of conspiracy theories.

Social Psychological and Personality Science, 7(2), 157-166.

Cohen, D., Nisbett, R. E., Bowdle, B. F., & Schwarz, N. (1996). Insult, aggression, and the southern

culture of honor: An" experimental ethnography.". Journal of personality and social

psychology, 70(5), 945.

Cohen, G. L. (2003). Party over policy: The dominating impact of group influence on political

beliefs. Journal of personality and social psychology, 85(5), 808.

DeWall, C. N., Enjaian, B., & Bell, S. B. (2016). ONLY THE LONELY. Ostracism, Exclusion, and

Rejection, 95.

Douglas, K. M., Sutton, R. M., & Cichocka, A. (2017). The psychology of conspiracy theories.

Current directions in psychological science, 26(6), 538-542.

Druckman, J. N., & Levendusky, M. S. (2019). What do we measure when we measure affective

polarization?. Quarterly, 83(1), 114-122.

62 Duckitt, John. 2001. A dual-process cognitive-motivational theory of ideology and prejudice. In

Advances in experimental social psychology. Vol. 33 Elsevier pp. 41–113.

Duckitt, J., Wagner, C., Du Plessis, I., & Birum, I. (2002). The psychological bases of ideology and

prejudice: Testing a dual process model. Journal of personality and social psychology, 83(1),

75.

Duran, Nicholas D, Stephen P Nicholson and Rick Dale. 2017. “The hidden appeal and aversion to

political conspiracies as revealed in the response dynamics of partisans.” Journal of

Experimental Social Psychology 73:268–278.

Fessler, D. M., Tiokhin, L. B., Holbrook, C., Gervais, M. M., & Snyder, J. K. (2014). Foundations

of the Crazy Bastard Hypothesis: Nonviolent physical risk-taking enhances conceptualized

formidability. Evolution and human behavior, 35(1), 26-33.

Frank, R. H. (1988). Passions within reason: The strategic role of the emotions. WW Norton & Co.

Gross, Justin H and Kaylee T Johnson. 2016. “Twitter taunts and tirades: negative campaigning in

the age of Trump.” PS: Political Science & Politics 49(4):748–754.

Gøtzsche-Astrup, O. (2019). Personality moderates the relationship between uncertainty and

political violence: Evidence from two large US samples. Personality and individual

differences, 139, 102-109.

Hagen, E. H. (2002). Depression as bargaining: The case postpartum. Evolution and Human

Behavior, 23(5), 323-336.

Hempel, A. G., Levine, R. E., Meloy, J. R., & Westermeyer, J. (2000). A cross-cultural review of

sudden mass assault by a single individual in the oriental and occidental cultures. Journal of

Forensic Science, 45(3), 582-588.

Ho, Arnold K, Jim Sidanius, Nour Kteily, Jennifer Sheehy-Skeffington, Felicia Pratto, Kristin E

Henkel, Rob Foels and Andrew L Stewart. 2015. “The nature of social dominance orientation:

63 Theorizing and measuring preferences for intergroup inequality using the new SDO7 scale.”

Journal of Personality and Social Psychology 109(6):1003.

Hogg, M. A., Kruglanski, A., & van den Bos, K. (2013). Uncertainty and the roots of extremism.

Journal of Social Issues, 69(3), 407-418.

Horowitz, D. L. (2001). The deadly ethnic riot. Univ of California Press.

Iyengar, Shanto, Gaurav Sood and Yphtach Lelkes. 2012. “Affect, not ideology a social identity

perspective on polarization.” Public Opinion Quarterly 76(3):405–431.

Iyengar, S., Lelkes, Y., Levendusky, M., Malhotra, N., & Westwood, S. J. (2019). The origins and

consequences of affective polarization in the United States. Annual Review of Political

Science, 22, 129-146.

Johnson, R. T., Burk, J. A., & Kirkpatrick, L. A. (2007). Dominance and prestige as differential

predictors of aggression and testosterone levels in men. Evolution and Human Behavior,

28(5), 345-351.

Jonason, Peter K, Norman P Li, Gregory D Webster and David P Schmitt. 2009. “The dark triad:

Facilitating a short-term mating strategy in men.” European Journal of Personality: Published

for the European Association of Personality Psychology 23(1):5–18.

Kaufmann, E. (2018). Whiteshift: Populism, immigration and the future of white majorities.

Penguin UK.

Kenrick, D. T., Griskevicius, V., Neuberg, S. L., & Schaller, M. (2010). Renovating the pyramid of

needs: Contemporary extensions built upon ancient foundations. Perspectives on

psychological science, 5(3), 292-314.

Kitschelt, Herbert. 2002. Popular dissatisfaction with democracy: populism and party systems. In

Democracies and the populist challenge. Springer pp. 179–196.

64 Kriesi, H., & Schulte-Cloos, J. (2020). Support for radical parties in Western Europe: Structural

conflicts and political dynamics. Electoral Studies, 65, 102138.

Krizan, Z., & Johar, O. (2015). Narcissistic rage revisited. Journal of Personality and Social

Psychology, 108(5), 784.

Kruglanski, A. W., Gelfand, M. J., Bélanger, J. J., Sheveland, A., Hetiarachchi, M., & Gunaratna,

R. (2014). The psychology of radicalization and deradicalization: How significance quest

impacts violent extremism. Political Psychology, 35, 69-93.

Kurzban, R., & Neuberg, S. (2005). Managing Ingroup and Outgroup Relationships. In D. M. Buss

(Ed.), The handbook of evolutionary psychology (p. 653–675). John Wiley & Sons Inc.

Lopez, Jesse and Hillygus, D. Sunshine, Why So Serious?: Survey Trolls and Misinformation

(March 14, 2018). Available at SSRN: https://ssrn.com/abstract=3131087 or

http://dx.doi.org/10.2139/ssrn.3131087

Marcus-Newhall, A., Pedersen, W. C., Carlson, M., & Miller, N. (2000). Displaced aggression is

alive and well: a meta-analytic review. Journal of personality and social psychology, 78(4),

670.

Mason, L. (2018). Uncivil agreement: How politics became our identity. University of Chicago

Press.

McAdams, Dan P, and Jennifer L Pals. 2006. “A New Big Five: Fundamental Principles for an

Integrative Science of Personality..” American psychologist 61(3): 204–17.

McCullough, M. E., Kurzban, R., & Tabak, B. A. (2011). Evolved mechanisms for revenge and

forgiveness. In Human aggression and violence: Causes, manifestations, and consequences.

(pp. 221-239). American Psychological Association.

Miller, N., Pedersen, W. C., Earleywine, M., & Pollock, V. E. (2003). A theoretical model of

triggered displaced aggression. Personality and Social Psychology Review, 7(1), 75-97.

65 Miller, Joanne M, Kyle L Saunders and Christina E Farhart. 2016. “Conspiracy endorsement as

motivated reasoning: The moderating roles of political knowledge and trust.” American

Journal of Political Science 60(4):824–844.

Moskalenko, Sophia and Clark McCauley. 2009. “Measuring political mobilization: The distinction

between activism and radicalism.” Terrorism and political violence 21(2):239–260.

Munger, Kevin. 2017. “Experimentally Reducing Partisan Incivility on Twitter.” Working Paper.

http://cess.nyu.edu/wp-content/uploads/2017/02/Dont--Me.pdf .

Mutz, D. C. (2018). Status threat, not economic hardship, explains the 2016 presidential vote.

Proceedings of the National Academy of Sciences, 115(19), E4330-E4339.

Nyhan, Brendan and Jason Reifler. 2010. “When corrections fail: The persistence of political

misperceptions.” Political Behavior 32(2):303–330.

Odling-Smee, F. J., Laland, K. N., & Feldman, M. W. (1996). Niche construction. The American

Naturalist, 147(4), 641-648.

Overbeck, J. R., Jost, J. T., Mosso, C. O., & Flizik, A. (2004). Resistant versus acquiescent

responses to ingroup inferiority as a function of social dominance orientation in the USA and

Italy. Group processes & intergroup relations, 7(1), 35-54.

Paulhus, D. L., & Williams, K. M. (2002). The dark triad of personality: Narcissism,

Machiavellianism, and psychopathy. Journal of research in personality, 36(6), 556-563.

Pennycook, G., & Rand, D. G. (2019). Lazy, not biased: Susceptibility to partisan fake news is

better explained by lack of reasoning than by motivated reasoning. Cognition, 188, 39-50.

Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social dominance orientation: A

personality variable predicting social and political attitudes. Journal of personality and

social psychology, 67(4), 741.

66 Rosenstiel, TR, A Mitchell and M Jurkowitz. 2012. “Winning the media campaign 2012: Both

candidates received more negative than positive coverage in mainstream news, but social

media was even harsher." Pew Research Center.

Rubin, M., & Hewstone, M. (1998). Social identity theory's self-esteem hypothesis: A review and

some suggestions for clarification. Personality and Social Psychology Review, 2(1), 40-62.

Sell, A., Tooby, J., & Cosmides, L. (2009). Formidability and the logic of human anger.

Proceedings of the National Academy of Sciences, 106(35), 15073-15078.

Spohr, Dominic. 2017. “Fake news and ideological polarization: Filter bubbles and selective expo-

sure on social media.” Business Information Review 34(3):150–160.

Starbird, Kate, Jim Maddock, Mania Orand, Peg Achterman and Robert M Mason. 2014. “Rumors,

false flags, and digital vigilantes: Misinformation on twitter after the 2013 boston marathon

bombing.” iConference 2014 Proceedings .

Tooby, J., & Cosmides, L. (1996). Friendship and the Banker's paradox: Other pathways to the

evolution of adaptations for altruism. In Proceedings of the British Academy(Vol. 88, pp.

119-143).

Tucker, Joshua A, Yannis Theocharis, Margaret E. Roberts and Pablo Barbera. 2017. “From liberation

to turmoil: social media and democracy.” Journal of democracy 28(4):46–59.

Turchin, Peter. 2016. Ages of discord: A structural-demographic analysis of American history.

Beresta Books.

Twenge, J. M., & Campbell, W. K. (2003). “Isn’t it fun to get the respect that we’re going to

deserve?” Narcissism, social rejection, and aggression. Personality and Social Psychology

Bulletin, 29(2), 261-272.

Uscinski, Joseph E and Joseph M Parent. 2014. American conspiracy theories. Oxford University

Press.

67 Van Bavel, J. J., & Pereira, A. (2018). The partisan brain: An identity-based model of political

belief. Trends in cognitive sciences, 22(3), 213-224.

Van Den Bos, K., Euwema, M. C., Poortvliet, P. M., & Maas, M. (2007). Uncertainty Management

and Social Issues: Uncertainty as an Important Determinant of Reactions to Socially

Deviating People 1. Journal of Applied Social Psychology, 37(8), 1726-1756.

Van Prooijen, J. W., & Jostmann, N. B. (2013). Belief in conspiracy theories: The influence of

uncertainty and perceived morality. European Journal of Social Psychology, 43(1), 109-115.

Van Prooijen, J. W., Krouwel, A. P., & Pollet, T. V. (2015). Political extremism predicts belief in

conspiracy theories. Social Psychological and Personality Science, 6(5), 570-578.

Vosoughi, Soroush, Deb Roy and Sinan Aral. 2018. “The spread of true and false news online.”

Science 359(6380):1146–1151.

Warner, Benjamin R, Sarah Turner McGowen and Joshua Hawthorne. 2012. “Limbaugh’s social

media nightmare: Facebook and Twitter as spaces for political action.” Journal of Radio &

Audio Media 19(2):257–275.

Wilson, M., & Daly, M. (1985). Competitiveness, risk taking, and violence: The young male

syndrome. Ethology and sociobiology, 6(1), 59-73.

Yang, A. (2015). Quantifying anomia: development of a scale(Doctoral dissertation).

Zimmerman, J. L., & Reyna, C. (2013). The meaning and role of ideology in system justification

and resistance for high-and low-status people. Journal of Personality and Social Psychology,

105(1), 1.

68