Science Journals — AAAS REVIEW False equivalencies: Online activism from left to right Deen Freelon1,2*, Alice Marwick2,3, Daniel Kreiss1,2 Digital media are critical for contemporary activism—even low-effort “clicktivism” is politically consequential and contributes to offline participation. We argue that in the United States and throughout the industrialized West, left- and right-wing activists use digital and legacy media differently to achieve political goals. Although left-wing actors operate primarily through “hashtag activism” and offline protest, right-wing activists manipulate legacy media, migrate to alternative platforms, and work strategically with partisan media to spread their messages. Although scholarship suggests that the right has embraced strategic disinformation and conspiracy theories more than the left, more research is needed to reveal the magnitude and character of left-wing disinformation. Such ideological asymmetries between left- and right-wing activism hold critical implications for democratic practice, social media governance, and the interdisciplinary study of digital politics. A ctivism is a fixture of contemporary politics, both democratic and otherwise. At its core is the drive to enact or pre- vent political, cultural, and/or social changes by a range of means. Although nonelite citizens have advanced activist claims against the powers that be for millennia (1), in the 21st century, digital media offer un- precedented tools for activists around the world to help realize their sociopolitical visions. In this review, which focuses on the United States but also incorporates evidence from other countries, we argue that both the ideological left and right use the ad- ditional channels and low-cost parti- cipation afforded by digital media to reach potentially sympathetic publics. However, despite some similarities, re- cent research indicates that left and right differ sharply in how they use di- gital media. Whereas the left generally combines on- and offline protest ac- tions with transmedia branding, an ap- proach known as “hashtag activism” (2), the right tends to eschew offline protest (notwithstanding a few prominent excep- tions), preferring instead a combination of “trolling” or manipulating mainstream me- dia, protest against and even strategic exit from platforms owned by “Big Tech,” and cooperation with ideologically friendly me- dia outlets. Moreover, available evidence suggests that the right has invested far more than the left in disinformation and conspiracy theories as core components of its activist repertoire, although a lack of sim- ilar research on the left makes comparisons difficult. These asymmetric trends hold im- portant implications both for scholarship and for democratic practice. Low cost, high benefit: Clicktivism and political participation Since the start of social media’s diffusion throughout Western societies, concerns have been raised about its efficacy for political par- ticipation. One prominent early objection was that “slacktivism” or “clicktivism,” low-cost sym- bolic actions such as sharing, “liking,” changing one’s profile image, and generally posting ac- tivist content on social media, projects an impression of efficacy without actually being effective (3). The two assumptions underlying this objection are, first, that such digitally mediated symbolic behaviors are generally not consequential in and of themselves and, sec- ond, that they substitute for more impactful actions such as voting or offline protest. Later, we will turn to recent research on how digital activism can be highly impactful on its own, contributing to phenomena such as disinfor- mation. Meanwhile, empirical research has consistently failed to support the proposition that digital action substitutes for offline action (4–6). That is, people who are strongly inte- rested in politics tend to express that interest through both online and offline behaviors. Dig- ital political activities–including low-cost ones– are a complement to, not a substitute for, their offline counterparts. Inversely, those who are uninterested in politics tend to avoid it both online and offline. Specifically, Lane et al. found that sharing information about politics on social media predicted offline political ac- tivities such as attending political meetings, contacting public officials, and donating money to political campaigns (4). de Zúñiga et al. (5) found that the use of social media to address community problems, which they call “social media social capital,” predicted the propensity to engage in similar activities offline. And a meta- analysis of 106 survey studies of young people's civic and political use of digital media in >35 countries found thatthe use of digital media for political purposes was positively correlated with offline political and civic engagement (6). The unanimity of the literature on this point has led some to declare that the clicktivism debate is conclusively settled (7). However, this conclusion is premature given several im- portant questions that lack solid empirical answers. One of the most pressing begins with the observation that political engage- ment is issue specific: An individual can be engaged with one or more issues and disen- gaged from others. The clicktivism question then evolves from whether low-cost digital ac- tivities exhaust one’s engagement with politics in general to whether such activities may do so for specific issues that lie beyond the person’s usual interests. For example, whereas liking, sharing, and posting memes about environ- mental topics may be just one of many ways an environmentalist engages with her pet issue, it may be the only way she does so for, say, Black Lives Matter when that movement is trending nationally on Twitter. The pat- tern of punctuated equilibrium that typ- ifies social movement activity on social media implies that some variant of this will be true at least some of the time. To continue with the Black Lives Matter example, a study that tracked related tweets over a 1-year period overlapping the movement’s birth showed a few sharp peaks of interest (most prominently in August, November, and December of 2014 and in April and May of 2015) separated by lengthy periods of much lower activity (Fig. 1) (8). This is typical of such movements’ social media activity and indeed of social me- dia in general (9). Logically, the bursts of attention that create such peaks must be provided by people (or bots, a non-negligible possibility) who engage for a short time and then depart, leaving a com- mitted core of activists to sustain the baseline conversation. Whether such participation is considered clicktivism is more a question of philosophy than empiricism. On the one hand, the degree of individual commitment is un- doubtedly low, but on the other, the aggregate crests of attention generated by thousands or millions of such actions can catapult a protest movement from obscurity to international pro- minence (10). As Freelon et al. document (8), grassroots attention on social media played a substantial role in spreading the initial pub- lic awareness of Black Lives Matter’s existence and goals, which was an essential precursor to its widespread acceptance by the American DEMOCRACY IN THE BALANCE Freelon et al., Science 369, 1197–1201 (2020) 4 September 2020 1 of 5 “Digital political activities–including low-cost ones–are a complement to, not a substitute for, their offline counterparts.” 1Hussman School of Journalism and Media, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 2Center for Information, Technology, and Public Life, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. 3Department of Communication, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA. *Corresponding author. Email: freelon@email.unc.edu o n A p ril 5 , 2 0 2 1 h ttp ://scie n ce .scie n ce m a g .o rg / D o w n lo a d e d fro m http://science.sciencemag.org/ public in mid-2020 (11). Our hypothetical en- vironmentalist may not have engaged with Black Lives Matter at all if low-cost online actions were unavailable; thus, rather than substituting for higher-cost street-level activ- ism, online actions broaden symbolic support for movements (12). Our main arguments on clicktivism can be summarized thusly: There is a continuum of online activist participation ranging from post- ing and liking content to high-level decision- making as a full-time activist. Even more, as the remainder of this review clearly reveals through the lens of recent empirical research, low-cost digital activities can sum to sub- stantial effects ranging from publicizing movements for mass audiences to circulating disinformation that undermines democratic deliberation and processes. A number of American activist movements have substan- tially furthered their goals through digital means over the past decade, including Oc- cupy Wall Street, Black Lives Matter, #MeToo/ #TimesUp, far-right anti-immigration advo- cates, and the mens’ rights movement. Sim- ilar results have been observed outside of the United States (10, 12, 13). To add a right-wing example to the Black Lives Matter case de- tailed above, Benkler et al. explain how far- right media, activists, and social media users successfully introduced the term “globalist,” an anti-Semitic dog whistle, into the journal- istic mainstream (14). This effort began with white nationalist sites such as VDARE and continued through Breitbart (a far-right site that avoids explicit white nationalism), Fox News, and the Trump administration after the 2016 election, finally ending up as a synonym for “neoconservative” in The New York Times. The online-only media outlets at the begin- ning of this chain rely heavily on social media sharing to boost their messages (15). In the United States, this is the main way they attract the attention of Fox News, which is more di- rectly networked with more traditional media outlets and the Trump administration. Over- all, this example demonstrates how far-right actors can insert their preferred terminology and ideas into more “respectable” outlets that would otherwise try to avoid such associa- tions. Other studies have demonstrated that sites such as Breitbart (and their European counterparts) serve similar “bridging” func- tions between far-right and legacy media (16, 17). In these and other ways, slacktivism has been a consequential component of con- temporary social movements and will likely continue to be so in the future. The empirical record has very little to say on the question of ideological asymmetries in slacktivism, mostly because left-wing protests have been studied far more than right-wing protests (18). Based on what we know about how most areas of life typically work online, we might expect that right-wing actors would use online and offline means to pursue their interests similarly to the way that those on the left do. One survey-based study found that for American respondents with low political interest, “easy political behaviors [such as liking and commenting on social media] can be gateway behaviors to more significant po- litical activities,” but that ideology was not a significant predictor of this tendency (19). Left- and right-wing digital strategies and ecosystems One of digital media’s most important contrib- utions to activism is how they have opened new pathways to reach target audiences. Be- fore the digital age, protesters who wished to project their messages nationally or interna- tionally had only one viable option: attracting the news media’s attention, which they usually did through street protests. Mailing lists and alternative media extended their reach only moderately. Today, digital media afford activ- ists across the political spectrum two general methods of promoting their causes. The first is to circumvent the news media entirely and appeal directly to digital platform users. This method offers the advantage of placing mes- sage control mostly in the hands of activists and sympathetic partiesbut by definitionmostly reaches people who are already platform users. Second, activists use digital platforms to attract journalists’ attention (because most use social media extensively as a gauge of public opin- ion and as a source of stories) (20) in the hopes that they will cover their movement. The ad- vantage here is that news outlets can reach in- dividuals outside of the digital spheres within which activists operate, as well as those who are not digitally active at all, but may also alter activist messages in ways that are not always favorabletomovements(21). These twomethods are not mutually exclusive; many of the best- known activist movements in recent years have used both (2, 8, 22). Although activists on both sides use digital media to reach audiences directly and indi- rectly through the news media, the left and the right have each evolved their own dis- tinct style of doing so. The dominant style on the left has been labeled hashtag activism (2, 23, 24) and bears three main distinguish- ing characteristics. The first and foremost of these is the creation of a declarative hashtag to serve as the movement’s unifying slogan; e.g., #BlackLivesMatter, #MeToo, and #Fightfor15 became shorthand for a host of demands and priorities. The limited amount of attention that most people decide to allocate to news in general and activist appeals in particular gua- rantees that only a few protest hashtags will attain national or international prominence. Such hashtags often come to the public’s at- tention through news coverage of shocking and disruptive events, such as Michael Brown’s death at the hands of police officer Darren Wilson in Ferguson, MO (#BlackLivesMatter), the disclosure of Harvey Weinstein’s decades- long history of sexual predation (#MeToo), and a series of American fast-food worker strikes in 2012–2013 (#Fightfor15). Second, such hash- tags are buoyed by the widespread engage- ment of nonelites, ordinary citizens who relate to the hashtag’s core message or simply want to declare their support. This is what causes them to “trend” on social media and thereby trigger the third element: attention and support from elite third parties. Most prominent among these are mainstream news outlets, which are often thefirstelitestopublicizeactivisthashtags.Others include celebrities, businesses, and politicians, all of whom hold disproportionate power to direct attention to movements. Examples in- clude hip-hop artists Talib Kweli and Common (#BlackLivesMatter), ice cream company Ben & Jerry’s (#BlackLivesMatter), actress Alyssa Milano (#MeToo), and Senator Bernie Sanders (#Fightfor15). Although much hashtag activism research is U.S. focused, the phenomenon has alsobeenobservedincountriessuchasArgentina (25), Bangladesh (26), France (27), and India (27). The right engages with these dual pathways very differently. Several fundamental differ- ences with the left explain this. First, American conservatives’ mistrust of the mainstream news media has been intensifying for decades (28, 29), a pattern that seems to be common on the right across Europe and India as well (30–32). The sense that traditional news out- lets are irredeemably biased against con- servatives is one of the driving factors in the establishment of right-wing media ecosystems, the roots of which in the United States reach back at least to the 1930s (33). Second, con- servatives have more recently developed an analogous belief that “Big Tech,” a pejorative term for the companies that produce and main- tain the internet’s most widely used commu- nication platforms and hardware, including Facebook, Google, Twitter, Apple, and Amazon, is also biased against them (34). These two beliefs have led the right to interact with the news media and tech platforms in more radi- cally oppositional ways than the left despite the latter’s critiques of those institutions. Dis- taste for (and being deplatformed from) Big Tech has prompted some far-right users to de- camp to platforms more accepting of their politics, including Telegram, Gab, and Voat (35). Third, since 2016, the center-right’s pres- ence on social media has diminished substan- tially (14, 36, 37), leaving the far right as the dominant conservative presence. Together, these short- and long-term trends have shifted the right into a world apart from the left and center, and its activist tactics reflect that re- ality. Figure 2 quantifies this phenomenon by depicting the percentages of “fragmented” Freelon et al., Science 369, 1197–1201 (2020) 4 September 2020 2 of 5 o n A p ril 5 , 2 0 2 1 h ttp ://scie n ce .scie n ce m a g .o rg / D o w n lo a d e d fro m http://science.sciencemag.org/ users that retweeted media outlets along the ideological spectrum in 2017. Outlets with pre- dominantly far-right audiences attracted nearly four times more fragmented users (those that disproportionately retweeted within one parti- tion) than the second most fragmented partition. Conservative mistrust of the mainstream media has inspired two distinctive tactics for interacting with two kinds of media outlets. Those that lack an explicitly conservative out- look often find themselves targeted by media manipulation, an umbrella term that refers to a repertoire of bad-faith tactics intended to attract journalistic attention (22). One of the most prominent of these, known as “trading up the chain,” involves planting a sensation- alistic hoax, conspiracy theory, or extreme viewpoint in a small or local news outlet that may not fact-check it (22, 38). This story may then be repeated by larger outlets, either be- cause of its content or because an elite (such as Donald Trump) has endorsed it. Whether the underlying claim is presented as true or debunked, the goal of spreading it further is fulfilled. By contrast, right-wing activists’ in- teractions with ideologically friendly outlets are understandably far less contentious. What Benkler et al. have called the American “right- wing media ecosystem” is a densely interlinked region of the media network that stands far apart from other media in terms of digital, professional, and ideological connections (14, 16). Its reach on social media platforms is extensive, in most cases larger than its left- wing equivalent (14). The ostensibly more journalistic outlets in this network, such as Fox News and the Daily Caller, regularly le- gitimize content surfaced by the more radical outlets, which include Infowars, Gateway Pundit, and Breitbart. The right-wing media ecosystem’s favored topics during the Trump administration have prominently included un- compromising opposition to non-Western im- migration, the evils of the so-called “deep state,” and attacks on the legitimacy of the Mueller investigation (14). Two other tactics used disproportionately by right-wing actors are specific to social plat- forms. The first is the strategic manipulation of platform algorithms to increase attention to desired messages. Much as the gatekeep- ing function of legacy journalism shaped the norms, practices, and patterns of news cov- erage of social movements, social platforms’ emphasis on user engagement affects what information is displayed to individual users, for example, by giving greater reach to emo- tionally charged content, videos, and visual graphics over text (39). Thus, successful online activists must understand how social plat- forms algorithmically sort content to ensure that their own is given priority. Although both left- and right-wing actors engage in such tactics, preliminary evidence suggests that the right has been more successful. For instance, platforms such as YouTube have recommen- ded increasingly extreme far-right content to viewers of more moderate right-wing channels to maximize user engagement with the site (40). Similar techniques include optimizing search engine keywords so that interested parties will more readily find ideologically biased results (41) and the use of fake accounts and bots to imply widespread consensus on social media (42). Because journalists often rely on engagement metrics such as Twitter’s “Trending Topics” to determine which stories should be covered and how they should be framed, successful algorithmic manipulation may help to set legacy media agendas (22). Second, in response to deplatforming, shadow banning, and content moderation by Big Tech, some right-wing actors have migrated to “alt- tech” equivalents that offer more permissive moderation. These include social media sites dedicated to right-wing communities, such as 4chan and 8chan, the Twitter alternatives Parler and Gab, and the YouTube alternative BitChute, as well as more ideologically neutral platforms such as Discord and Telegram (35). Although alt-tech platforms are much smaller than their mainstream counterparts, they al- low partisan and fringe communities to exist without opposition from alternative viewpoints. Studies have demonstrated a high prevalence of hate speech on 4chan (43), Gab (44), and BitChute (45), which is typically moderated on more mainstream social platforms. These spaces allow more extreme viewpoints to thrive, whereas mainstream social media pri- marily host less extreme content designed to reach wider audiences (22). The most relevant implications of the dif- ferences between how left- and right-wing ac- tivist networks reach their respective audiences derive from their very different relationships with the platforms they use. The left largely engages directly with traditional and social media, using them as primary communication venues to develop and distribute activist mes- sages. These outlets and platforms present themselves as what Cass Sunstein called “gen- eral interest intermediaries” (46), information environments that admit a wide range of per- spectives. Consequently, left-wing ideas tend to connect with individuals and institutions along a much broader range of the ideological spectrum than the right, including much of the center (14). By contrast, the right has created and used its own ideologically exclusive media ecosystem and digital platforms even as it continues to engage with the best-known tech platforms and news outlets out of necessity. These developments in turn (along with other nondigital factors) fuel what scholars have called “asymmetric polarization,” the proposition Freelon et al., Science 369, 1197–1201 (2020) 4 September 2020 3 of 5 June 2014 July August September October November January 2015 February March April MayDecember Tw ee ts p os te d pe r da y 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 Million Initial protests after Michael Brown was shot and killed by police officer Darren Wilson in Ferguson, Missouri Death of Freddie Gray in police custody in Baltimore, Maryland Decision to not indict New York City police officer Daniel Pantaleo, whose chokehold killed Eric Garner Decision to not indict Darren Wilson Fig. 1. Daily tweets about police violence and Black Lives Matter, June 2014 to May 2015. Reproduced with permission from (8). See (58) for the data and code used in creating this figure. DEMOCRACY IN THE BALANCE o n A p ril 5 , 2 0 2 1 h ttp ://scie n ce .scie n ce m a g .o rg / D o w n lo a d e d fro m http://science.sciencemag.org/ that conservatives have grown more extreme over the past few decadesthanliberals(14, 37, 47). Asymmetric polarization’s broader consequences include less common ground between opposing political sides, increasingly extreme policies when conservatives are elected, and more opportunities for ideologically branded mis- and disinformation to spread on the right, which we discuss further in the next section. Emerging research on asymmetric disinformation Since the 2016 U.S. and U.K. Brexit elections, scholars, the news media, and international publics have become increasingly concerned with the problem of false and misleading political content (14, 22, 48, 49). This general pheno- menon has multiple variants with a variety of labels, including the ubiq- uitous and ambiguous “fake news,” which we avoid. Here, we will focus on disinformation, which we define as “all forms of false, inaccurate, or misleading information designed, pre- sented and promoted to intentionally cause public harm or for profit” (48). Unlike misinformation, which refers to misleading content spread inadver- tently, disseminators of disinformation knowtheirmessagesaredeceitful.Actors behind such deceptive content seek to spread conspiracy theories, false rumors, hoaxes, and inflammatory opinions to promote their ideological viewpoints, decrease trust in mainstream insti- tutions, and recruit others to their causes (22). The relevant literature offers three types of evidence in support of the proposition that disinformation is more prevalent on the right than on the left, although to our knowledge this has not been directly tested. First, evidence from psychological studies indicates that conserva- tive individuals are more likely than liberals to prefer the kinds of closed media environ- ments (sometimes called “echo chambers”) that facilitate the spread of mis- and disinfor- mation (50), believe conspiracy theories when cued by official denials of conspiratorial causes (51), and tolerate the spreading of disinforma- tion by politicians (52). Second, analyses of false news diffusion on social media have generally shown a tendency for conservatives to share such content more than liberals (53, 54). Third, the most visible mainstream news media out- lets, upon which the left relies much more heavily for political information than the right, have a long history of fact-checking norms that largely prevent disinformation from thriv- ing there (14), which is why understanding how the news industry operates helps individ- uals avoid disinformation (55). Existing research provides numerous exam- ples of conservative-targeted disinformation, in which right-wing media ecosystems around the world are often centrally implicated (49, 56, 57). In the United States, the alt-right, unapologetic white nationalists, and others on the rightmost fringe attract relatively small audiences and must rely on media outlets at higher levels of the ecosystem to help circulate their disinformation and other extreme ideas broadly (14). The fringes are not always suc- cessful; in particular, conspiracy theories im- plicating a Washington, DC, pizza parlor as the center of a Democrat-controlled pedophilia ring and accusing a left-wing activist of mur- dering a counterprotester at the 2017 Unite the Right rally were not endorsed by the ecosys- tem’s upper echelons (14, 38). The ranks of disinformation stories that achieved greater notoriety include the Seth Rich conspiracy, in which a Hillary Clinton staffer was allegedly murdered because of what he knew about her emails. (Rich was killed in Washington, DC, on 10 July 2016 by unknown assailants, but no credible evidence links his death to Clinton.) The story originated among fringe ecosystem users on Twitter and Reddit in the weeks after Rich’s death (14). Sean Hannity covered the conspiracy multiple times in 2017 on his eponymous Fox News program, although the network eventually retracted the story. More recently, our analysis shows thatthe top ranks of the Twitter network discussing the debunked 2020 documentary Plandemic (which makes unsubstantiated and scientifically unsound allegations about COVID-19) in April and May of 2020 prominently includes right-wing media ecosystem members such as GatewayPundit (@gatewaypundit) and commentators for Fox News (@greggutfeld) and Infowars (@liberty- tarian) (58). In this way, the right-wing media ecosystem circulates sensationalistic content to an ideologically friendly audience free of the sorts of editorial practices that would prevent the spread of false information. The goal, as with much disinformation, is to support the in- group and denigrate the outgroup, even at the expense of verifiable truth. Perhaps because of the implications of the research reviewed above, very few studies have directly investigated online left- wing disinformation or conspiracy theories at scale. The studies show- ing a conservative-leaning asymmetry in social media false news sharing largely draw their data from before the 2016 election (53, 54). If liberals have changed in their susceptibility to dis- information in the ensuing years, pos- sibly because of incentives introduced by strong anti-Trump animus, we do not yet know. This could be a case of failing to find that which is not sought. The implications of such research are highly relevant to democratic practice: For one, they will help us understand the extent of the problem, who is most acutely affected, and under what con- ditions. Understanding the ideological and psychological antecedents of dis- information susceptibility is an impor- tant first step in targeting interventions to counteract it. To the extent that we as citizens value a democracy free of fraudulent attempts at opinion manip- ulation, we should investigate all con- texts in which it might lurk. Two existing studies, along with our own analysis of recent Twitter data, offer some evidence that left-leaning disinformation may not be as rare as the literature suggests. First, research published by Buzzfeed in October 2016 found that although conservative Facebook pages posted nearly double the proportion of false or partly false content as liberal pages, such content garnered much higher median shares per post on left-wing pages than on right-wing ones (59). (We should note that this report only analyzed six Facebook pages in total, its data were not made public, and it is possible that false content on right-leaning pages accrued more shares in total given that there was more of it.) Second, a recent study found that tweets posted by Russian disinfor- mation agents masquerading as left-wing African American activists attracted more attention on a per-tweet basis than either those by conservative identities or non-Black left-leaning identities (60). This demonstrates a level of vulnerability to disinformation on Freelon et al., Science 369, 1197–1201 (2020) 4 September 2020 4 of 5 0 5 10 15 20 25 30 35 40 45 50 Far left Center left Center Center right Far right P er ce nt ag e of f ra gm en te d us er s 41,176/41,176/ 325,771325,771 8998/8998/ 187,861187,861 106,008 /106,008 / 224,381224,381 632/632/ 54,93254,932 632/ 54,932 12,343/12,343/ 260,248260,248 12,343/ 260,248 41,176/ 325,771 8998/ 187,861 106,008 / 224,381 Fig. 2. Percentages of fragmented users retweeting media outlets across five ideological partitions. The denominator for each percentage is the number of users who retweeted (shared content from) at least one media outlet in that partition, whereas the numerator is the number of users for whom at least 80% of their retweets were of outlets in that partition (i.e., “fragmented” users). This figure depicts the behavior of the 1.82 million unique Twitter users in the dataset who retweeted three or more media accounts. The dataset upon which this figure is based comes from (37) and contains >88 million tweets about six major news issues throughout 2017. Media outlets and corresponding ideological classifications come from (14). See (58) for the data and code used in creating this figure. o n A p ril 5 , 2 0 2 1 h ttp ://scie n ce .scie n ce m a g .o rg / D o w n lo a d e d fro m http://science.sciencemag.org/ the left that is not often acknowledged. Third, we find that tweets mentioning the key words “anonymous” and “trump” posted between 31 May 2020 (when the Anonymous hacktivist collective released a cache of documents pur- porting to prove, among other accusations, that Donald Trump was involved in child sex trafficking) and 2 June 2020 were retweeted >1.1 million times, more than double the total retweet count for Plandemic in our analysis above (58). In contrast to the Plandemic net- work, the most-retweeted users on this topic are overwhelmingly nonelites with few fol- lowers (except for @youranoncentral, which is ostensibly controlled by Anonymous), not well- known liberals or mainstream news outlets. We acknowledge that these findings are pre- liminary and raise pressing validity questions— many of the attention metrics boosting these stories could have been generated by bots, for example—but we include them here for lack of more rigorous research on the matter. Taken together, we believe that they suffice to justify further investigation into disinfor- mation aimed at the left. Conclusion and future research This review offers three main sets of conclu- sions. First, people participate in online activ- ism along a wide spectrum of commitment levels, from liking and sharing content, to the back-and-forth of political discussion, to involvement as core movement leaders. Low- cost online actions do not harm activist goals; on the contrary, they help to boost activist topics and concerns to the levels of public visibility necessary to enact or prevent change. Both the left and right benefit from this basic dynamic of online activism. However, there is still much to learn about how clicktivism op- erates; for example, we still do not know how frequently hashtag-based conversations or signal-boosting extreme perspectives change people’s minds or behaviors. Second, the left and right generally engage in two distinct styles of online outreach: hashtag activism and online advocacy spearheaded by the right- wing media ecosystem, respectively. The iso- lation of the far right from the rest of the ideological spectrum results in asymmetric polarization and complicates the process of governing ideologically diverse polities. Key areas for future research here include mea- suring the relative capacities of these two styles in reaching, persuading, mobilizing, and antagonizing elites and nonelites on both sides. Third, disinformation distribution ap- pears to be one of the key functions of right- wing media ecosystems. However, the marked lack of research on left-wing disinformation leaves many questions about how it operates, who is most at risk, and how serious a problem it is, making such research an urgent priority. The very limited number of studies on right- wing online protest and activist hashtag use is similarly glaring. Moving forward, research- ers should endeavor to discover whether our current empirical understanding of left- and right-wing activism online represents reality faithfully or is a product of systematic gaps in case selection. REFERENCES AND NOTES 1. C. Tilly, L. J. Wood, Social Movements 1768–2012 (Routledge, 2013). 2. S. J. Jackson, M. Bailey, B. F. Welles, #HashtagActivism: Networks of Race and Gender Justice (MIT, 2020). 3. E. Morozov, The Net Delusion: The Dark Side of Internet Freedom (PublicAffairs, 2011). 4. D. S. Lane, D. H. Kim, S. S. Lee, B. E. Weeks, N. Kwak, Soc. Media Soc. 3, 2056305117716274 (2017). 5. H. G. de Zúñiga, M. Barnidge, A. Scherman, Polit. Commun. 34, 44–68 (2016). 6. S. Boulianne, Y. Theocharis, Soc. Sci. Comput. Rev. 38, 111–127 (2018). 7. D. Karpf, in A Research Agenda for Digital Politics, Elgar Research Agendas, W. H. Dutton, Ed. (Edward Elgar Publishing, 2020); pp. 123–132; http://doi.org/10.4337/9781789903096. 8. D. Freelon, C. McIlwain, M. D. Clark, “Beyond the hashtags: #Blacklivesmatter, #Ferguson, and the online struggle for offline justice” (Center for Media and Social Impact, American University, 2016); https://cmsimpact.org/blmreport. 9. S. A. Myers, J. Leskovec, in WWW ’14: Proceedings of the 23rd International ConferenceonWorldWideWeb(AssociationforComputing Machinery, 2014), pp. 913–924; 10.1145/2566486.2568043. 10. H. Margetts, P. John, S. Hale, T. Yasseri, Political Turbulence: How Social Media Shape Collective Action (Princeton Univ. Press, 2015). 11. N. Cohn, K. Quealy, The New York Times, 10 June 2020; https://www.nytimes.com/interactive/2020/06/10/upshot/ black-lives-matter-attitudes.html. 12. P. Barberá et al., PLOS ONE 10, e0143611 (2015). 13. P. N. Howard, S. Savage, C. F. Saviaga, C. Toxtli, A. Monroy-Hernández, J. Int. Aff. 70, 55–73 (2016). 14. Y. Benkler, R. Faris, H. Roberts, Network Propaganda: Manipulation, Disinformation, and Radicalization in American Politics (Oxford Univ. Press, 2018). 15. S. Zannettou et al., in IMC ’17: Proceedings of the 2017 Internet Measurement Conference (Association for Computing Machinery, 2017), pp. 405–417; 10.1145/3131365.3131390. 16. J. Kaiser, A. Rauchfleisch, N. Bourassa, Digit. Journal. 8, 422–441 (2019). 17. A. Heft, E. Mayerhöffer, S. Reinhardt, C. Knüpfer, Policy Internet 12, 20–45 (2019). 18. S. Boulianne, Commun. Res. 0093650218808186 (2018). 19. L. Bode, Gateway political behaviors: The frequency and consequences of low-cost political engagement on social media. Soc. Media Soc. 2056305117743349 (2017); http://doi. org/10.1177/2056305117743349. 20. S. C. McGregor, L. Molyneux, Journalism 21, 597–613 (2018). 21. D. M. McLeod, J. K. Hertog, Discourse Soc. 3, 259–275 (1992). 22. A. Marwick, R. Lewis, Media Manipulation and Disinformation Online (Data and Society Research Institute, 2017). 23. J. Schradie, The Revolution That Wasn’t: How Digital Activism Favors Conservatives (Harvard Univ. Press, 2019). 24. Schradie (23) found evidence of an alternative pattern of digital activism on the left: Specifically, volunteer left-wing labor organizers in North Carolina created digital presences that were less interactive and drew less attention than their well-funded conservative counterparts. Although the prevalence of this pattern is currently unknown, her research suggests that we may need to look beyond ideology to organizational stability and funding sources to assess the efficacy of digital activist strategies. 25. F. Belotti, F. Comunello, C. Corradi, Violence Women 1077801220921947 (2020). 26. M. H. Zaber, B. Nardi, J. Chen, in LIMITS ’17: Proceedings of the 2017 Workshop on Computing Within Limits (Association for Computing Machinery, 2017), pp. 51–58; 10.1145/3080556.3080557. 27. I. Lopez, R. Quillivic, H. Evans, R. I. Arriaga, in Human-Computer Interaction – INTERACT 2019: 17th IFIP TC 13 International Conference, Paphos, Cyprus, September 2–6, 2019, Proceedings, Part II, D. Lamas, F. Loizides, L. Nacke, H. Petrie, M. Winckler, P. Zaphiris, Eds. (Springer, 2019), pp. 733–743. 28. R. R. Mourão, E. Thorson, W. Chen, S. M. Tham, Jpn. Stud. 19, 1945–1956 (2018). 29. M. D. Watts, D. Domke, D. V. Shah, D. P. Fan, Communic. Res. 26, 144–175 (1999). 30. T. U. Figenschou, K. A. Ihlebæk, Jpn. Stud. 20, 1221–1237 (2019). 31. S. Nygaard, Jpn. Stud. 21, 766–782 (2020). 32. P. Bhat, K. Chadha, J. Int. Intercult. Commun. 13, 166–182 (2020). 33. A. J. Bauer, A. Nadler, in News on the Right: Studying Conservative News Cultures, A. Nadler, A. J. Bauer, Eds. (Oxford Univ. Press, 2019), pp. 1–16. 34. P. J. Hasson, The Manipulators: Facebook, Google, Twitter, and Big Tech’s War on Conservatives (Simon and Schuster, 2020). 35. R. Rogers, Eur. J. Commun. 0267323120922066 (2020); http://doi.org/10.1177/0267323120922066. 36. J. Wihbey, K. Joseph, T. Coleman, D. Lazer, in KDD’17: Proceedings of Data Science + Journalism @ KDD’17 (Association for Computing Machinery, 2017); https://kennyjoseph.github.io/papers/dsj.pdf. 37. D. Freelon, “Tweeting left, right, & center: How users and attention are distributed across Twitter” (Knight Foundation, 2019); https://knightfoundation.org/wp-content/uploads/ 2019/12/KF-Twitter-Report-Part1-v6.pdf. 38. P. M. Krafft, J. Donovan, Polit. Commun. 37, 194–214 (2020). 39. S. Milan, Inf. Commun. Soc. 18, 887–900 (2015). 40. M. H. Ribeiro, R. Ottoni, R. West, V. A. F. Almeida, W. Meira, in FAT*’20: Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (Association for Computing Machinery, 2020), pp. 131–141; 10.1145/3351095.3372879. 41. M. Golebiewski, D. Boyd, “Data voids: Where missing data can easily be exploited” (Data & Society, 2018); https:// datasociety.net/library/data-voids/. 42. L. Luceri, A. Deb, A. Badawy, E. Ferrara, in WWW’19: Companion Proceedings of The 2019 World Wide Web Conference (Association for Computing Machinery, 2019), pp. 1007–1012; http://doi.org/10.1145/3308560.3316735. 43. G. E. Hine et al., in Eleventh International AAAI Conference on Web and Social Media (AAAI, 2017); https://www.aaai.org/ocs/ index.php/ICWSM/ICWSM17/paper/view/15670. 44. S. Zannettou et al., in WWW’18: Companion Proceedings of the Web Conference 2018 (Association for Computing Machinery, 2018), pp. 1007–1014; http://doi.org/10.1145/3184558.3191531. 45. M. Trujillo, C. Buntain, B. D. Horne, What is BitChute? Characterizing the ‘free speech’ alternative to YouTube. arXiv:2004.01984 [cs] (4 April 2020). 46. C. Sunstein, Republic.com 2.0 (Princeton Univ. Press, 2007). 47. C. A. Bail et al., Proc. Natl. Acad. Sci. U.S.A. 115, 9216–9221 (2018). 48. High Level Expert Group on Fake News and Disinformation, “A multi-dimensional approach to disinformation: Report of the Independent High Level Group on Fake News and Online Disinformation” (European Commission, 2018); https://ec. europa.eu/digital-single-market/en/news/final-report-high- level-expert-group-fake-news-and-online-disinformation. 49. M. Bastos, D. Mercea, Philos. Trans. A Math. Phys. Eng. Sci. 376, 20180003 (2018). 50. J. T. Jost, S. van der Linden, C. Panagopoulos, C. D. Hardin, Curr. Opin. Psychol. 23, 77–83 (2018). 51. A. M. Enders, S. M. Smallpage, Polit. Commun. 36, 83–102 (2018). 52. J. De keersmaecker, A. Roets, Pers. Individ. Dif. 143, 165–169 (2019). 53. A. Guess, J. Nagler, J. Tucker, Sci. Adv. 5, eaau4586 (2019). 54. N. Grinberg, K. Joseph, L. Friedland, B. Swire-Thompson, D. Lazer, Science 363, 374–378 (2019). 55. M. A. Amazeen, E. P. Bucy, J. Broadcast. Electron. Media 63, 415–432 (2019). 56. E. Ferrara, First Monday 22 (2017). 57. M. Hameleers, Politics Gov. 8, 146–157 (2020). 58. Data, code, and documentation used to conduct the original empirical analyses for: D. Freelon, A. Marwick, D. Kreiss, False equivalencies: Online activism from left to right, Harvard Dataverse (2020); 10.7910/DVN/ZH1EWA. 59. C. Silverman, L. Strapagiel, H. Shaban, E. Hall, J. Singer-Vine, BuzzFeed News, 20 October 2016; https://www.buzzfeednews. com/article/craigsilverman/partisan-fb-pages-analysis. 60. D. Freelon et al., Soc. Sci. Comput. Rev. (2020). ACKNOWLEDGMENTS We gratefully acknowledge the research assistance of K. Adams and M. Reddi. Funding: The empirical analysis shown in Fig. 1 was supported by grant no. 201600019 from the Spencer Foundation. The empirical analysis shown in Fig. 2 was supported by grant no. GR-2018-55703 from the John S. and James L. Knight Foundation. Author contributions: D.F. wrote the initial draft of this review and conducted all original empirical analyses. A.M. and D.K. contributed to writing and editing the review. Competing interests:The authors declare no competing interests. Data and materials availability: All data, code, and documentation used to conduct the original empirical analyses in this review (Fig. 1, Fig. 2, and the “plandemic” and “anonymous trump” analyses) are available on the Harvard Dataverse (58). 10.1126/science.abb2428 Freelon et al., Science 369, 1197–1201 (2020) 4 September 2020 5 of 5 DEMOCRACY IN THE BALANCE o n A p ril 5 , 2 0 2 1 h ttp ://scie n ce .scie n ce m a g .o rg / D o w n lo a d e d fro m http://doi.org/ https://cmsimpact.org/blmreport https://www.nytimes.com/interactive/2020/06/10/upshot/black-lives-matter-attitudes.html https://www.nytimes.com/interactive/2020/06/10/upshot/black-lives-matter-attitudes.html http://doi.org/ http://doi.org/ http://doi.org/ https://kennyjoseph.github.io/papers/dsj.pdf https://knightfoundation.org/wp-content/uploads/2019/12/KF-Twitter-Report-Part1-v6.pdf https://knightfoundation.org/wp-content/uploads/2019/12/KF-Twitter-Report-Part1-v6.pdf https://10.1145/3351095.3372879 https://datasociety.net/library/data-voids/ https://datasociety.net/library/data-voids/ http://doi.org/ https://www.aaai.org/ocs/index.php/ICWSM/ICWSM17/paper/view/15670 https://www.aaai.org/ocs/index.php/ICWSM/ICWSM17/paper/view/15670 http://doi.org/ https://arxiv.org/abs/2004.01984 https://ec.europa.eu/digital-single-market/en/news/final-report-high-level-expert-group-fake-news-and-online-disinformation https://ec.europa.eu/digital-single-market/en/news/final-report-high-level-expert-group-fake-news-and-online-disinformation https://ec.europa.eu/digital-single-market/en/news/final-report-high-level-expert-group-fake-news-and-online-disinformation https://10.7910/DVN/ZH1EWA https://www.buzzfeednews.com/article/craigsilverman/partisan-fb-pages-analysis https://www.buzzfeednews.com/article/craigsilverman/partisan-fb-pages-analysis http://science.sciencemag.org/ False equivalencies: Online activism from left to right Deen Freelon, Alice Marwick and Daniel Kreiss DOI: 10.1126/science.abb2428 (6508), 1197-1201.369Science ARTICLE TOOLS http://science.sciencemag.org/content/369/6508/1197 CONTENT RELATED http://science.sciencemag.org/content/sci/369/6508/1147.full http://science.sciencemag.org/content/sci/369/6508/1192.full http://science.sciencemag.org/content/sci/369/6508/1188.full http://science.sciencemag.org/content/sci/369/6508/1183.full http://science.sciencemag.org/content/sci/369/6508/1181.full http://science.sciencemag.org/content/sci/369/6508/1179.full http://science.sciencemag.org/content/sci/369/6508/1176.full http://science.sciencemag.org/content/sci/369/6508/1174.full REFERENCES http://science.sciencemag.org/content/369/6508/1197#BIBL This article cites 30 articles, 4 of which you can access for free PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions Terms of ServiceUse of this article is subject to the is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience Science. No claim to original U.S. Government Works Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of o n A p ril 5 , 2 0 2 1 h ttp ://scie n ce .scie n ce m a g .o rg / D o w n lo a d e d fro m http://science.sciencemag.org/content/369/6508/1197 http://science.sciencemag.org/content/sci/369/6508/1174.full http://science.sciencemag.org/content/sci/369/6508/1176.full http://science.sciencemag.org/content/sci/369/6508/1179.full http://science.sciencemag.org/content/sci/369/6508/1181.full http://science.sciencemag.org/content/sci/369/6508/1183.full http://science.sciencemag.org/content/sci/369/6508/1188.full http://science.sciencemag.org/content/sci/369/6508/1192.full http://science.sciencemag.org/content/sci/369/6508/1147.full http://science.sciencemag.org/content/369/6508/1197#BIBL http://www.sciencemag.org/help/reprints-and-permissions http://www.sciencemag.org/about/terms-service http://science.sciencemag.org/