Online as the New Frontline: Affect, Gender, and ISIS-Take-Down on Social Media

ABSTRACT Using a dataset of more than 80 accounts during 2015, this article explores the gendered ways in which self-proclaiming Twitter Islamic State of Iraq and Syria (ISIS) supporters construct community around “suspension.” The article argues that suspension is an integral event in the online lives of ISIS supporters, which is reproduced in online identities. The highly gendered roles of ISIS males and females frame responses to suspension, enforcing norms that benefit the group: the shaming of men into battle and policing of women into modesty. Both male and female members of “Wilayat Twitter” regard online as a frontline, with suspension an act of war against the “baqiya family.” The findings have implications for broader repressive measures against ISIS online.

90,000 active ISIS support accounts. 3 These consisted of both "official" propaganda accounts, as well as thousands of "unofficial" accounts owned by sympathizers and supporters across the world. Content was expressed in a variety of languages, originating from, and aimed at, different global audiences. 4 Twitter's popularity was in part due to its diverse functionality. Twitter enables users to adopt an anonymous online identity, which can easily be customized to facilitate identification by "like minds." Account holders can micro-blog messages up to 140 characters in length, in which media can be embedded. Users may "follow" others, and be "followed" back. Twitter allows users to share content, either by retweeting others' messages or endorsing them through "likes" (previously "favorites"). Accounts may be rendered visible to a select private audience or the public at large, and using past online activity, Twitter suggests similar accounts of interest to follow. With this complex functionality, Twitter therefore offered the potential for the transnational growth of ISIS networks, for active recruitment, for the dissemination of content including videos, images, and documents, and the ability to tailor messaging to different audiences, private as well as public. 5 As a result, open source social media and Twitter material have become a useful and muchrelied-on research resource. 6 Twitter provides data-rich insights into the lives, preoccupations, and beliefs of those posting. This is not to assume that all those professing allegiance to ISIS online are what they appear; the recent case of Joshua Goldberg, who simultaneously represented himself online as jihadi, Nazi, and feminist activists shows that identity online can be deceptive. 7 Nor are there established links between online activity and offline violence, or indeed offline activism. Few studies have assessed empirical data to understand the relationship between virtual and offline extremism. 8 However, those that do suggest the Internet does not necessarily accelerate the process of "radicalization." Von Behr et al. found online activity was no substitute for in-person meetings; 9 meanwhile, in a study of lone actor terrorism, Gill et al. found the rise of the Internet did not correlate with a rise in lone actor terrorism. 10 In a further study of the online behaviors of 227 convicted U.K. terrorists, 96 percent of whom were male, Gill et al. conclude that the vast numbers of online extremists do not progress to terrorism. 11 However, those who do commit terrorist acts "regularly engage in activities in both [online and offline] domains." 12 Therefore there is no easy translation of Twitter-or any other online-extremism to the offline environment. What research of Twitter has, however, enabled, is the demonstration of the ways in which ISIS's use of the forum are consistent with the historical use of both media, and social media, by terrorist groups more broadly (propaganda, recruitment, notifications, claims of attack, calls to attack). 13 This study again makes no claim to conclusions that will assist offline counterextremism methods; it instead offers insights into the gendered effects online of a particular initiative within Countering Violent Extremism (CVE) tactics: suspension.
Given the success of the Internet in drawing both male jihadi fighters and female jihadi state-makers from across Europe, Asia, the Middle East and the Americas, Internet companies have an interest in understanding their own role as a facilitator in this process. 14 Social media providers also experience pressure from policymakers and police to act. In 2016 the UK Home Affairs Select Committee suggested far more needed to be done by social media companies to counter the "Wild West" that is the Internet. 15 In May 2017, a subsequent Home Affairs Select Committee report suggested social media providers were repeatedly failing to remove illegal content, even when alerted to its existence, and needed to take far greater responsibility. 16 Twitter's policy is to prohibit abusive behavior and misuse by extremist groups or individuals promoting violence or hate. 17 Aware of its growing colonization by ISIS supporters, Twitter has also recently instigated a targeted policy, intensifying the account-takedown or "suspension" of active ISIS supporters, employing more staff with this remit, and instituting a series of technical programs to identify offending accounts. 18 Since 2015, when this research was conducted, Twitter has aggressively pursued a suspension strategy, reducing numbers by as many as 360,000 to August 2016. 19 Since the instigation of this policy, Twitter ISIS networks, recruitment, and funding have all been hit hard. 20 As a "take-down" process, suspension constitutes a repressive rather than preventive counterradicalization measure. It can remove accounts but cannot stop "returners," who frequently reestablish new accounts post-removal, known as the "whack-a-mole" effect. 21 Despite frequent "return," other social media platform providers have also favored "suspension" or "removal" policies, viewed by many as an effective means to limit the online expansion of ISIS networks. 22 Berger and Morgan support this approach in their ISIS Twitter Census, the widest study to date of ISIS behavior on the forum. 23 While noting the prevalence of return accounts, and an estimated 8 percent of account activity reportedly dedicated to "account rebuilding," Berger and Morgan deny this invalidates suspension as a policy. 24 They liken the tactic to weeding: "The consequences of neglecting to weed a garden are obvious, even though weeds will always return." 25 Through interrupting online activity, and requiring effort for ISIS supporters to find networks and followers once more, account suspension "wastes their time." 26 Berger, however, notes the possibility of unwanted side-outcomes to suspension, such as further radicalization.
What has not been considered, however, is the way in which gender is present in CVE tactics such as Twitter suspension of ISIS supporters. Gender is a key gap in research on online extremism, as will be explored in the next section. Analysing gender involves both an engagement with differences in the ways in which self-identifying men and women behave online, the social expectations of that behavior, and any differences in how tactics such as suspension impact on male and female supporters. As Sjoberg asserts, "While sex categorisation is a part of gender analysis, gender is often described as a social construct … constructed by a particular culture … because people agree to behave as if it exists or to follow certain conventional rules." 27 Such rules govern the Salafi jihad, dictating male/female roles, behaviors, spaces of access and aspirations. 28 The majority of violent actors may be men, but women are present as "extremists" in large numbers online. In existing discussion of Twitter suspension, quantitative assessments of impact have additionally been prioritised, with less research into the qualitative effects on ISIS supporters as a group. This piece aims to address this gap, by focusing on the gendered effects of counterextremism methods on Twitter, first, because they are under-explored, and yet reveal interesting dynamics in identity and networks within online ISIS communities; second, as they have implications for similar restrictive actions elsewhere on the Internet. ISIS has demonstrated the ability to repeatedly evolve, and the lessons learned from ISIS support-and take-down-on Twitter, are important for counterextremism across platforms. The analysis also focuses on the affective impacts of suspension, and from the perspective of ISIS supporters, in order to reveal this support group not simply in numerical terms, but as a community, and one that is constructed around gender. The next sections will set out why gender and affect are significant for study of Twitter and suspension, before moving to the methodology and findings themselves.

Gender, Women, and ISIS Online
Studies on how gender factors in online radicalization are still in their infancy, and there are calls for further research. 29 Most of the work to date has focused on the role online of "ISIS women." This is due to the concerted involvement of women by ISIS in its Caliphate project, which marks a key difference between ISIS activity and previous Islamist groups, including Al Qaeda. The state-centric goals of ISIS, and the aim of creating a functioning polity within a-besieged-territory, have necessitated the recruitment of women, who are desired within this space as state-makers, home-builders, supporters, mothers, propagandists, and wives for fighters. 30 While women from across the world have traveled to Syria and Iraq to live within the Caliphate, 700 from Tunisia for example alone, many more women ISIS supporters are highly active online, a long-seen trend in female Islamist activity. 31 Al Qaeda, for example, used propaganda magazines such as Inspire to promote women as core online supporters. Their lack of public visibility was regarded as a key advantage to their ability to covertly raise support, as well as funds. 32 This is to some degree replicated in ISIS relations with women. In her large-N sample Twitter study, Klausen considered the actions of female ISIS supporters online, and pointed to the "Umm factor"-a reference to the frequently adopted honorific-which suggests women are key secondary disseminators of extreme material. 33 The relative prevalence of female Islamist engagement online is linked by a number of authors to key differences between the operation of Salafi-jihadist gender norms in the offline versus online space. Opportunities for public engagement in offline space are limited for both Salafi-jihadist women, and within some traditional Muslim cultures more broadly. Women may face restricted movement and expression outside the home, and "free-mixing" between sexes is not encouraged. Authors including Sageman suggest the online space is therefore popular among female extremists as it operates as a comparative site of liberation, free of the gender boundaries and restrictions frequently found offline, particularly regarding the expression of female aggression. Consistent with this, Bermingham et al. point to the intensity of extremist messaging by women online, the Internet allowing them greater freedom to demonstrate aggression when compared with men, in the absence of constraining offline gender norms. 34 Women's engagement in jihadist interaction online may therefore fulfill a "safety valve" function. 35 Contrasts between online freedoms for women and offline restrictions are not trivial. There is already some evidence that gender differences online are impacting on offline behavior, with women for example more likely to be recruited to ISIS through online approaches. Claudia Dantschke founded the German counselling program Hayat to help those involved in radical Salafist groups. She works with women and families in deradicalization and associates an increased female "vulnerability" to ISIS recruitment online with women's better-developed online social networks, and an ability to snowball recruit via online friends. 36 Other research notes that women are perhaps less likely to be recruited offline in part due to the aforementioned cultural restrictions on physical freedoms and access to public space. 37 Gyms, mosques, and streets, examples of locations of male recruitment, 38 are not necessarily "spaces of vulnerability" for women. Indeed women exploit their presence online to serve as far more than just supportive "fan-girls," 39 and have proved themselves active recruiters of other women. Saltman and Smith's analysis of 12 social media female jihadis concluded they were committed to a recruitment role. These women were invested in the principles of violence, even if Salafi-jihadist gender restrictions disbarred them from violence itself. 40 The importance of less restrictive gender norms in offline space, and the impacts on female and male expression and behavior are so far, however, broadly absent from assessment of CVE approaches online. It is long acknowledged that Muslim men have most often been suspected of terrorism, and been subject to CVE programming or tactics such as the United Kingdom's Stop and Search policy in the offline space, with disproportionate impacts on them. 41 The prevalence of ISIS-supporting women online similarly suggests that CVE measures in this virtual space may disproportionately affect them, or at least affect them in particular ways. Indeed, to understand the online space in gendered terms and to recognize the differences in male and female norms and behaviors raises important questions for any repressive actions there. This article provides tentative answers to those questions, by considering gender differences in the impact of suspension.
Affect, Community, and the "Baqiya Effect" Online studies of ISIS females online to date have focused on the roles taken, risks posed, and activities carried out, in terms defined by the frameworks of terrorism research (e.g., propagandist, media-spokesperson, recruiter). The meaning given to the online space by ISIS activists, as constructed by them, has been neglected. Twitter ISIS supporters can, however, be considered part of a broader online "radical milieu," supportive of the aims of ISIS, and engaged in constant attempts to express this, despite user regulations on "terrorist" groups and online abuse. 42 Users themselves are aware of their creation of an online milieu, and the particular group norms of the Salafi-jihadist ideology. In particular, gender provides a useful framework through which to consider the interactions of ISIS Twitter supporters, as they themselves broadly segregate into communities of male and female users. These communities do mix, forming shared networks; however, both men and women tweet caution over too much "free-mixing" online, and many accounts suggest members of the opposite sex should not "DM" ('direct message') them with private messages. Indeed, the demarcation of particular roles for men and women, and gender segregation, is a feature of Salafi-jihadism, and gender has an evident function in group ideology as well as the practicalities of recruitment, propaganda, and legitimate access to organizational violence. 43 The sense of online community has also led many ISIS supporters to describe the Twitter space as a "wilayat" or "province." Members of this province are constructed as the "baqiya family," baqiya being a term colonized by ISIS and meaning "persistence", or "endurance". 44 This term emerges in this research as a centrally unifying concept around which Twitter supporters cohere in the face of adversity. Study of the affective nature of the relationships within this "family" enables another reading of gender, beyond that of difference between "women" and "men," instead considering power and status rooted in particular conceptions of masculinity. Affect can be understood as emotion created in interactions between people, encounters producing self, identity, and also bodies. 45 It can also be read as a "state" of being, the effects of which are amplified when shared, for example, a contagious yawn.
Affective studies reconceptualize feeling as communal and shared, rather than experienced by individuals alone. 46 Communal feeling can be invoked by terrorist groups, who create collective "moral shock," through the dissemination of distressing images or narratives. 47 Gender is evident in such messaging through association with status, status-lowering emotions-such as humiliation-associated with feminization/emasculation, and status-raising with masculinity. 48 Additionally, non-conscious affective dimensions of emotion nourish other conscious feelings, such as shame or humiliation, which can be explicitly instrumentalized to create radical identities, as in the context of the violent jihad. 49 Affect is crucial to understanding both offline and online space, where social networks are constructed around emotion, intimacy, and affectual relationships. 50 The naming of emotion can be instrumental in creating online communities of feeling. 51 As Ahmed emphasizes, shared anger, pain, grief, or love are inherently politicizing emotions offline and online, both "shared spaces of dwelling." 52 For Ahmed there are two important consequences: the need to reconsider the public manifestation of sentiment assumed to belong to the private sphere; and the contestation of the dichotomy, emotion versus rationality. 53 In the online space the result can be the creation of "tribes of shared emotions," in which everyday encounters are key. 54

Article Aims
This article incorporates the above insights to offer an analysis of the gendered and human impact of suspension on Twitter communities, bridging a research gap. It considers the sideeffects of suspension and asks, what else is happening in online milieus when supporters are removed en masse? What are the supporter perspectives on this action? Does it deter them, or encourage? Additionally, given the ways in which Twitter ISIS communities self-segregate according to gender, how do groups of men and of women respond to suspension? This article aims to go some small way to address the research gap in gender and extremism, with an exploration of the impact of suspension on a discrete sample of ISIS supporters on Twitter through the month of July 2015 and a focus on gender and affect. It challenges the conceptualization of suspension as an event outside the "core business" of Twitter ISIS supporters online, presenting it instead as an integral feature of the online lives of ISIS sympathizers. This is evidenced in the incorporation of suspension into online representations of identity, avatars, and handles.
The article also challenges the terms in which impact and influence online should be constructed, suggesting the importance of "immeasurables" based on affect and emotion, such as "community" and, in this sample, the "baqiya family." This emphasis emerges from consideration of the meaning of Twitter, as constructed by ISIS supporters themselves, and independent of analyses seeking to categorize and delineate particular behaviors, according to existing counterterrorism frameworks. First, it considers the activities observed in the gendered communities evident online. Second, it focuses on the impacts of suspension, arguing that: suspension bolsters a dominant recruitment narrative, that the mainstream persecutes Muslims and the Ummah, enabling Twitter to be constructed as an extension of the offline battlefield; ritual practices and norms are in turn constructed around suspension, and suspension is a Twitter life-event around which gendered community forms. It adds to growing studies suggesting suspension is disruptive, but also constructive, when regarded from the perspective of the group. 55 It additionally understands the online space as one in which, as Kuntsman suggests, "digital 'structures of feeling' work together. with broader political forces." 56 The analysis considers gender through differences in male and female behaviors, and practices, representative of different values associated with men and women through the Salafi-jihadist ideology adhered to by ISIS. These gender values are malleable, subject to definition according to culture, group purpose, and location. 57 By using gender as a tool of analysis, this article builds on the work of many scholars who believe the consideration of gender is integral to understanding militant groups. 58 It also heeds a recent call by Conway for research on online extremism to pay closer attention to gender, a somewhat underexplored aspect of investigation. 59 It considers the concept of baqiya as one uniting supporters of ISIS across social media platforms, suggesting that lessons learned from Twitter therefore have implications for future sites of ISIS activity. It recommends a gendered awareness of the possible outcomes of suspension, given the different behaviors of self-identifying men and women online. While Twitter suspension has clearly been very effective in removing content and users, the article reflects on the potential impact of this removal on morale, and community. As ISIS supporters are removed from Twitter, some may find other social media platforms to occupy, which must also confront the challenge of extremism. Additionally, the findings may have implications for other ideological movements on Twitter and other social media. 60

Methodology
Twitter has served as perhaps the most researched site of ISIS activity, to the neglect of other platforms. 61 This article, however, explicitly sets out once again to engage with Twitter ISIS support, and for three reasons. First, despite a relative focus on Twitter, there are still many unanswered questions about extreme movements' behavior there, particularly relating to gender. Second, this work aimed to address a gap in analysis of scale. There have been few small, random sample, qualitative analyses with line by line manual coding of entire lifecycles of accounts, from first to last tweet. Large-N studies, considering a vast quantity of data, have analyzed discrete sets of tweets, and not entire life-cycles of a given account. For example, Klausen considers only the last ten tweets of influential individuals in her 2015 study, Berger and Morgan only the last 200 in their census. 62 Third, it focused on random users, rather than those well known in the media, although "celebrity" characters were found within the networks. Fourth, as the site on which ISIS supporters were most active, there are lessons to learn from research on Twitter for other social media platforms perhaps absorbing its users as a result of the suspension clamp-down.
It should be noted that one limitation of any social media research is that the chosen sample may not represent the wider population of interest as a whole. Indeed, Mislove et al. suggest that Twitter users, when contrasted with the general population, tend to be disproportionately male, white, and from densely populated geographic regions. 63 While this bias is obviously unlikely to straightforwardly apply to ISIS Twitter samples, the general question of how to avoid sampling bias remains unresolved. 64 Meanwhile, there are specific issues for terrorism research, in which participants are criminalized, removed from social media sites and their communities are hard to access, online and off. With that caveat, Twitter nonetheless provides a rich resource for the understanding of social relations and identities, at the level of the populations constructing them. 65 The results presented here therefore suggest interesting and unexplored further paths for larger-scale research, as well as theoretical avenues for future study.
This research was undertaken throughout July 2015, with three data harvests carried out over the one-month period: one at the outset, one midway through July, and one at the end. To locate a cross-section of the ISIS-supporting milieu online an initial hashtag search was undertaken to identify five seed accounts for each sex, from which other accounts could be harvested through snowball sampling. The hashtags used were those popular among ISIS supporters on Twitter at the time. Some were gendered, such as #Teamhijabi, others were not, such as #baqiya. Others were temporary, and hijacked other causes, such as #sevenseven, the hashtag commemorating the Al Qaeda London Transport attacks of 7 July 2005, also used by ISIS to call for further violence.
The final data-set size aimed for was between 30-50 accounts for each of two categories: self-identifying men and self-identifying women, the smallest usable size for a qualitative study in which data was coded line by line by the researcher. Eventually 47 women were included and 41 men. There is, however, no independent way in which to verify the sex of a Twitter account-holder. Twitter users personalize profiles in a variety of ways, all of which present a means to express gender. These include: the profile names, handle (the "@" name address), avatars (profile picture), carousel (profile backdrop picture) or user biography, a short description. 66 Some handles are "kunyas" or "noms de guerre" and those beginning "Umm" or "Abu" are used to denote gender, 'Umm' being the Arabic for 'mother', and 'Abu' for 'father'. Many are not. The self-identified sex of the user can also be determined through examination of interactions between users, who greet one another in gendered ways (e.g., "Akhi", the Arabic for "my brother" and "Ukhti", the Arabic for "my sister"). Additionally, there is no independent way to verify account identity. As far as was possible this difficulty was mitigated through the selection of accounts with a large network of followers, some of whom could be verified through media reports, and frequent tweeting. This suggests acceptance by other members of networks, who themselves are suspicious of accounts that may be "spies," police, journalists, or researchers. Where "men" and "women" are used in the analysis, the phrase "self-identifying" should therefore be understood. The research also excluded "bots" or automated tweet generators. These are relatively easy to identify through the pay-for Twitter analysis site used, Twitonomy, in which their tweeting patterns emerge as square, non-organic, patterns.
The main inclusion criteria were that accounts were active, public, and that support of ISIS was the primary function of the account. "Support" was established in a number of ways: tweet content, Avatar and Carousel images (ISIS flags, the Black Standard, ISIS symbols), locations including Mosul, Raqqa, Dowla Islamiya, Islamic State, or in Arabic, . The research sought accounts with more than 100 followers. However, after the first harvest the latter criterion was relaxed, in order to include account-holders reactivating after being suspended. Mass suspension frequently affected Twitter accounts supporting ISIS at this time and this affected data, removing tweets, followers, and altering the numbers of favorites and retweets of remaining accounts. Another important inclusion criterion was that the account user should be predominantly tweeting in English. Many English-dominant accounts also use small amounts of Arabic, or other European languages, and retweet tweets in other languages.
The sample consisted of a diverse age range, apparently spread across a number of countries. Some users state their country of origin via names, or handles (e.g., "al Amriki", the American, or "al Britaniyah"); however, many are ambiguous, and given information may not be accurate. Some claimed to be in Syria and Iraq (nine men and seven women); others in Europe, Africa, Turkey, and the Americas. Tweets may contain the location of the user, but only if Global Positioning System (GPS) facilities are turned on, which is rare among ISIS supporters. Additionally, Twitonomy was used to provide the time-zone of accounts, and main language tweeted in. Four or five seed accounts were located for each set, "men" and "women," and their followers and following lists explored to identify a further set of approximately 40 accounts per gender for analysis. The seed accounts, since expired, were @Umm___Maryam, @saraEUK, and @Urwatulwuthqa_ AllahuMawlakum for women. The men's seed accounts were @The_Ansar_4_ISIS, @omarabdullah_00, @Salafi_ Jihadist, @AbuHudayr03, and @BurrningOne. From here snowball sampling was engaged to grow and include a variety of other accounts fulfilling the key inclusion criteria.
Exclusion criteria were also applied. The first excluded category was intellectuals, radical imams, and known preachers, such as Anjem Choudary or Mizanur Rahman, both United Kingdom-based, both relatively well known to the media, public, and security services, and both of whom have subsequently been convicted for terrorism offenses. They did not count as "ordinary" community members, and their accounts, while containing tweets in support of Sharia and ISIS, were not primarily aimed at support of this movement. The hashtags also brought up a second category that needed to be excluded, Twitter-users following an orthodox Salafi Islam, without any evidence that they advocated violence or supported ISIS.

Analysis
This is a qualitative study, with some quantitative analysis relating to basic statistical properties of the sets of tweets, male and female, in order to measure influence alongside affect. Twitonomy was used to produce Excel spreadsheets of content and basic quantitative information on tweet patterns, and screenshots were taken of each avatar and carousel image. These images were also coded. Quantitative information was gathered regarding the action categories on Twitter: Tweeting, Retweeting, Mentioning, Replying, Favoriting (which has since become "Liking," now indicated by a small heart image, but previously a star), Tweeting Hashtags, and Tweeting Links. Comprehensive analytics can be found in Table 1. The content and tone of these tweets was analyzed and coded manually.
This article is the result of recognition of a significant aspect of the study: the frequent suspension of accounts in the dataset, and the affective ways in which that was differently engaged by male and female ISIS supporters. It should be noted that the initial aim of this research was to fill a gap in the knowledge of female ISIS supporter behavior on Twitter, compared with male. The focus on gender in suspension emerged through gradual awareness of the ways in which suspension was impacting accounts. July 2015 saw several mass suspensions by Twitter, partly due to the actions of anti-ISIS activists "Anonymous," who mass-reported accounts. 67 This presented challenges for the maintenance of the dataset, although it did not permanently deter the sample. In all, 16 of the 47 women in the dataset were suspended during the study, and 12 of the 41 men, with women's suspension rates slightly higher per person on average (0.33 to the men's 0.3). Some of the accounts initially identified for the sample had also previously returned from suspension. Given the analysis of full life-cycles of accounts, using the data provided by Twitonomy, 68 suspension emerged as an important part of content and this impacted the direction of the study.
An analysis was carried out on users' tweet content, the aim of which was to consider the impact of suspension on accounts. This was measured by assessing the number of tweets dedicating to re-establishing accounts, the proportion of tweets dedicated to this task, and how this was achieved, including the use of emoticons, for example: "@BxxxRxxx24 waiyyaki ukhti..˘˘˘" The first tweet back after suspension was also analyzed, to assess the tone, and to record how frequently this was retweeted.

Findings: Gendered Communities, Gendered Roles
Before addressing the specific gendered impact of suspension and the ways in which ISIS communities of self-identifying males and females respond to this tactic, it is necessary to explore the highly gendered nature of the research sample. This consisted of a broad interconnected community, with subdivisions evident along gendered lines. These gendered communities encountered in this research visibly served the function of propagating offline ISIS norms-relating for example to women's dress, or men's duties to engage in battle-in the online space, with a high level of coherence between norms in the two domains. These norms serve to keep the Islamic State functioning through the prescription of female domestication alongside ideological commitment, and a male preoccupation with battle. Conformity to such norms is needed to ensure the future of the state, and the online communities therefore act in ways to ensure individuality is repressed. Male-female divisions in obligations are repeatedly emphasized. Women police one another into maintenance of ISIS norms, and men encourage one another to battle. Each gendered community also acts to shame and police the other sex, as well as their own, into their role. These insights are further developed later in the paper, with consideration of the specific ways in which suspension impacted on these communities.

Gendered Communities, Gendered Influence
Men and women emerged as enjoying different influence, and roles, using gendered communities in different ways, with different audiences, figures and reach, as shown in Figure 1. Males had greater scores in a number of measured categories suggestive of greater online "influence." These are presented in the format (male:female), and all are median values. They had more followers when contrasted with women (473:344); the ratio of followers/ following was higher (3.44:1.71); they have more user mentions (63:41) and more mentions per tweet (0.44:0.28). The percentage of their tweets being retweeted was higher (20.69:13.64), as was the total number of retweets (88:76.5), and the number of times any given retweeted tweet is retweeted (3.78:2.88). The percentage of male tweets being favorited was also higher (38.78:33.4), and each favorited tweet was favorited more than for women (2.18:2). All this suggests men are the key Twitter influencers with their tweets more likely to be disseminated, mentioned, and favorited. After suspension, it is common for users to announce their return to Twitter with "returning" tweets. Male "I'm back" announcements after suspension were retweeted a mean average of 24 times, versus women's 16, while male return announcements were favorited a mean average of eight times, and women's seven. Men seem to "count" more online, according to standard interpretations of influence.
To some degree a user cannot control which of their tweets others choose to retweet or favorite. However, men clearly acted to promote their influence, for example, employing greater use of hashtags, which give tweets greater reach and ensure they are retrievable to a wider audience. Considering median values, they used hashtags eight times as much as women (8:1), and three times as much when hashtags per tweet are examined (0.03:.01). Men's use of hashtags clearly weights their ability to be retweeted, to be mentioned, and to gain followers in the way in which they appear to be doing. Men also tweet more links per tweet, when the median score is considered (again, 0.03:0.01), which may also give their tweets greater appeal.
Women's sense of community was by contrast more "insular," demonstrated in their higher scores in certain categories than men. Again, considering median averages, expressed (Male:Female), they favorited more tweets than men (35:43.5), and a higher number of their tweets were favorited (77:110.5). Additionally, they have a higher number of favorites per 100 followers (22.71:37.68). One explanation could be that women are followed predominantly by other women, although this could not be verified; another is that women appear to prefer favoriting as a mode of tweet approval. Favoriting a tweet, while visible to the original tweeter, who receives a notification, is primarily a private act, shared between fewer people than either retweeting a tweet or mentioning it in another tweet. Female approval-giving therefore appears as a more private act, for their community only, than for men. 69 This may also indicate more personal relationships within female networks on Twitter.

Shaming Men, Policing Women
The gendered communities online are highly conformist, evident in the similar choices of avatar and carousel images. Both men and women favored particular images, including the ISIS fighter, the flag, ISIS slogans, religious slogans, and weaponry. Some avatar images had specific gender associations. Men chose the ISIS fighter as the most popular image. However, there was less conformity around the men's choice of avatar, with this image appearing 19 times. Meanwhile, the most popular women's image was the black burqa-clad woman, which appeared 29 times as an avatar image. This is consistent with ISIS dress codes for women in all their Wilayat, with obligatory black burqa, gloves, and niqab. Women dressed in this way were associated with "true beauty" and contrasted with-immoral-Western constructions of beauty, perceived as synonymous with vanity and display of flesh. One user, "RemoveYourFaceAvis," committed her online presence to pressuring others to ensure female online supporters did not post images of their faces. She tweeted, for example, "We advice brothers [sic] to not RT UNCOVERED women. 1) sin themselves by disobeying Allahs command to lower gaze2) let all their followers sin." This tweet was retweeted 27 times and favorited 12 times. Conformity around established ISIS norms better enables activities such as the policing of ISIS women supporters' behavior, and regulation of male commitment to the physical task of violent jihad. Female ISIS Twitter supporters spent time in engaging in enforcement of group conformity around dress, modesty, and the rules of interaction with men on Twitter.
Men also policed women's behavior, with enforcement of these norms frequently justified in theological terms. For instance, one user tweeted "Two of the worst things to come across: 1. A sister without Haya 2. A brother without Ghırah." Haya represents Islamic modesty and Ghırah, a protective jealousy exercised by men, around "their" women. Haya requires male Ghırah, and Ghırah requires the high valuation of women's modesty and honor, in order to have any meaning itself.
Men meanwhile also exerted pressure on each other. Their focus was, however, on the honor of eventual martyrdom, and the necessity of violent jihad. This was the central focus of male tweets, wherever in the world, as this user suggests, "Wallaahi I cannot wait for the day that I meet all of my twitter bros, standing together in the front lines of battles." Women were expected to support this aim, even if it meant losing a husband. This male user tweeted, "Our Women Don't expect Us to Return Home with Roses in Hand, Rather Klash Hanging from our shoulders, faces full of dust,body full of injury." Women served to regulate this through shaming. Umm Jxxxx retweeted this message for example, "15 years old teenagers [sic] making hijra while 40 years old grown adult men shaking when the word 'jihad is said near them…'"

Relationship to State and Offline Gendered Roles
Such roles are crucial to the future of the Islamic State as formulated by its leader al Baghdadi. Binary roles are enforced for both men and women, without which, the state, still battling for its survival, could not endure. The central aim of all official activities and norm propagation is to ensure the future of the movement in Iraq and Syria. In this, men have a clearly central aim as fighters, and women as migrants, and state-builders, wives and mothers. Women support violence, but only by men; only one woman in this sample was in favor of active female participation in violence. Their attitude toward their own responsibilities was consistent with ISIS norms set out in propaganda such as Dabiq, and centered on having children as future fighters, in lieu of participation in violence. Women frequently tweeted about this, as in this tweet from Umm Jxxxx who writes, "I may not be able to join in jihad and fight but I can raise the next line of Mujahideen InshaAllah and I encourage sisters to do the same." Such sentiments were often accompanied by photographs depicting this next line of fighters, as shown in Figure 2. Relationships with men were also central to their online conversations, particularly through the use of images. Loving relationships were frequently anthropomorphised, with images of lions or other animals used to depict affection between male and female.
Communities were oriented toward three things: fantasies of a past Utopia, which they were endeavoring to recreate; the much-cited hardships of the present struggle; and a future life where Islam could be followed in the way they preferred, with strict adherence to Sharia. The binary offline norms of Islamic State, in which men have predominantly warrior roles, and women are monitored and policed, are clearly reproduced online, constructed as a virtual "frontline." Findings: Suspension, Emotion, and Gender: Fear, Anger, and Resistance in Wilayat Twitter The previous section explored the ways in which gender is crucial to an understanding of the sample of ISIS supporters: gender shapes their communities, behaviors, and the relationship of Twitter ISIS followers to the norms of the offline Islamic State. Suspension thus impacts not merely on ISIS supporters, but on sub-communities of males and females. This highly gendered online landscape contextualizes the findings presented in this section, which focuses on the key findings that the impact of suspension is affective, generating an emotional response that is incorporated into the online identities of the dataset, becoming a significant event in their "Twitter lives." This was an event around which emotions of fear, anger, and outrage were constructed; this in turn fostered community and identity, and in gendered ways, with clear differences between responses in self-identifying male and female account-holders. Most importantly, the communities that consistently promote the core ISIS philosophy appeared to be strengthened by suspension, an act around which they cohered, in gendered terms.

Gender and Suspension in the Creation of Online Identities
Identities are created and displayed online in a variety of ways. First, specific identities are created through the biographies and handles of many of the accounts. As mentioned, Twitter users may have an "@" address or handle, and a name, which can be changed during a Twitter life. Users may also include a brief biography. Post-suspension, many users created both biographies and handles that represented their resilience in the face of having "survived" suspension. This was particularly true of the self-identifying men in the sample, of whom 27 percent directly referenced suspension in their Twitter profile information. By contrast, only 17 percent of the sample's women used profile information to notify others of past suspension.
The majority of the sample referred to Twitter's use of suspension as a specific tool to persecute Muslims. For example, one user tweeted "RT @_xxx_4_ISIS: @xxx_2 Dear akhi, could you please give me a shout out? There has been a major Twitter purge targeting Muslims." July 2015 also witnessed use of the hashtag "radicalized." Comments implicating Twitter in that process included, "I got #Radicalised when me and my people started to get suspended on every social media platform." Suspension permits Twitter ISIS supporters a specific avenue for resistance, as seen in the biography of Umm Q (Figure 3). This spirit of resistance is also evident in the first tweet of another female user, "They made their best to defeat us but kuffar doesnt know they can't let us our spirit down we came back," which was accompanied by violent emoticons, accentuating the emotional tone of the language used ( Figure 4). Perceived as an oppressive action, suspension is reproduced by ISIS followers according to a subversive agenda, as seen for example in the handle of QxQx alBxxxxxxx, who humorously names himself "@omgnotagain313". Oppression, through suspension, is no longer an action simply exerted by external forces on ISIS followers; it is reconstructed as part of a long history of "kuffar" aggression against Muslims, and incorporated into the identity of those online, not weakening, but strengthening them. For example, Abu B and Abu D reference past suspension in their image headers ( Figure 5). The male approach to suspension is highly masculinized according to hypermasculine criminalized norms, reconstructing Twitter disruption as the online equivalent to offline jail-time. Many men in the sample used biographical information to record the number of times that they had been suspended and returned. The ability to "survive" and return from suspension demonstrated "baqiya", the enduring nature of the Islamic State. For those who had returned despite challenge from the "kuffar", the display of past suspension became a badge of honor, both celebrating those who made it back and defying the intention of the practice.
While many men followed this pattern, others emphasized the difference between the Twitter battle-online and therefore insignificant-and the "real" battle taking place in Syria and Iraq-of high significance. Indeed, an emphasis on place was apparent in the images used to represent identities, with a number including images of desert, nature, or landmarks of Shams to remind what and where the fight was about. Some users in fact mocked male accounts boasting of suspension return. Abu TLT for example scorned fellow Twitter users, tweeting, "Whats up with this new trend of 'Twitter Veteran' in Bio, Get A Life Fellas."  Identity is an important aspect of ISIS supporters' online presence and one that the sample strove to maintain, despite suspension. The repeated use of favorite avatars, carousel images, or specifically worded biographical information persisted throughout suspension attempts. This had two functions: first, it enabled other users to recognize returning friends; second, it also served to emphasize the significance of the continuity of visual aspects of identity online. For example, € Ummãmîñ a identified herself in her initial biography as a "Scots lassie" living in Algeria. Her tweet rate was extremely high, in excess of 68,000 tweets. She seemed to have escaped suspension due to her location outside Europe. However, in July 2015 her account disappeared. When it reappeared a few days later, it was possible to identify her, as a number of characteristics remained the same or were tonally similar. These included: the same carousel image, a similar avatar image (light slogan in English on dark background of images), similar stylistics for name and handle, including the use of redundant accents, and a new name, which rhymed with her old (Figure 6).

Gender and the Ritual of Twitter "Return"
Suspension is not simply evident in its incorporation into the visual identities of ISIS Twitter supporters. Analysis of tweet content of whole life-cycles of accounts reveals its prominence as a significant Twitter "life event" and therefore as a conversation topic. Those returning from suspension adhere to ritualistic formulae when re-establishing a Twitter presence with a new account. This was repeated over and over in the sample, with 73 percent of the women returners following this pattern, and 75 percent of men. The ritual is formalized in stages. The first stage is the announcement of the returnee that they are back. The second stage is reliance on both "shout-out" accounts and other ISIS supporters to help them rebuild followers. "Shout-out" accounts are those explicitly set up to enable returners to find old networks, to let others know they are back, and to widely disperse news of suspensions and returns. The third stage is one of back-and-forth in which shout-outs are retweeted, and those assisting in this process are thanked.
Gendered differences in approach to this ritual were evident in the sample, primarily through the different tone engaged by men and women to shout outs for help, and to thanks for it. The central quality to this ritual among women is an affective display of warmth and love for any other women helping them. While "brotherhood" characterizes male desire to engage in the ISIS community, and indeed the radicalization mechanism as conceptualized in Sageman's "Bunch of Guys" model, the importance of the notion of sisterhood to ISIS women has also been noted. 70 This "Ukthi-hood" is distinguished through the prolific use of feminized emoticons, such as hearts and flowers (Figure 7). This warmth and sense of built community through affect and shared emotion is one ISIS Twitter supporters are conscious of and strive to create, as one female user expresses through a retweet, "It's when you're suspended that you feel the love and support from other Muslims on Twitter the most. I love being suspended:) ." The unity of community through ritual and conformity is expressed in instruction of new members in the particular suspension rituals of the group. Female users, for example, tutor new recruits to the use of Arabic language they may not be familiar with, such as "jazak Allah" instead of the English "thank you." Arabic phrases pepper even the accounts of users who are clearly only fluent in English. The quality of the returning tweets of male ISIS supporters is less deliberately warm and emotional, with fewer emoticons. Many men instead engaged in the "game" of suspension with humor, for example, "Wallah im loving it enraging the kuffar come join in loool I love shahdah so ready to get suspended."

Suspension and Community: Male and Female Roles
Such differences-male sample members adopting an outwardly more blas e stance toward return after suspension, engaging humor, against the women's more feminized discourse within all-female groups-contributed to the construction of particular forms of online ISIS community. The sample network consisted of both men and women. However, this could be further subdivided into particular communities of males and communities of females, engaging with one another in different ways, on different themes, one of which was suspension itself. Women in the sample were more likely to be suspended in the month of study than men. Tweets on the subject of suspension, or explicitly following suspension, constituted 24 percent of the output of those women who experienced suspension, and appeared to fulfill a number of functions, including re-establishing one's own account and others' accounts; discussing the consequences, likelihood, and causes of suspensions; speculating as to the whereabouts of other users post-suspension. The figure for those men who were suspended was 20 percent.
However, "suspension conversation" (retweets of others shout-outs, support tweets for accounts, shout-outs for one's own account and tweets in which content speculated on suspension) was not limited to those who themselves experienced suspension. The topic accounted for 16 percent of the total female tweets across the sample and 8.5 percent of the male. Attempts to divine the logic of account suspension preoccupied ISIS supporters, and what caused suspension was the source of much speculation. Some attributed it to time spent online, as one female user tweeted, "Ahaan … I ain't got suspend. … Weired [sic] My this account lasts 1.5 month… lol." Another theory expressed was that suspension related to numbers of followers, as this female user in Libya speculated, "I worry when my followers gets to 400 means my suspension is on the way." Another male user tweeted, "Yes twitter i know i have reached 900 followers no need to remind me munafiq i know u'r going to suspend me today." If suspension is an event around which community builds, it is also one that tests the boundaries of those communities, which are rigorously protected. Trust among followers is hard to gain, and particularly after a mass suspension, credentials are frequently required to ensure people have the identities they claim. This can be frustrating, as this tweet shows, "RT @skxx45: This was prepared from me and now for those who again ask me photos and doubt go & see that its me & not fake." Suspicion among ISIS supporters on Twitter is high, as a similar tweet from Abu H expresses: "The politics of twitter and everyone second guessing if next person is legit or a fed is highly irritating. I miss the pre ban Twitter." Many sample users created back-up accounts as part of their armory in this battle, left dormant until needed, but their addresses shared, so that followers would easily find them. For example, this male user "@Backup_109_109" tweets "… i have some special codes n words with brother which make them identify real me." Suspension enforces the sense of in-and outgroup, the subterfuge surrounding return emphasizing the subversive nature of the ISIS community online, versus everyone else. The re-establishing of accounts is a way of enforcing boundaries between "us" and "them." While tweets call for support for some accounts, they also fulfill an important role in warning against others, in order to protect the "baqiya family." This was most prevalent in the female sample, which most visibly and publicly policed itself, as is evident in the conversation below. A user replies to two fellow female Twitter users who apparently warn her against accepting new followers too quickly: A: Yes that's true. I always block them. Also those who locked their acc and follow many tweeps. Jazak Allah khair akhi for… … Reminding us. Especially in the last days did kuffar a lot to cause problems B: exactly ukhti. It's a lot of work to block all and to find out who's a member or not but better to be careful.
C: than having a troll among us. And especially the new sisters have to checked … it's better to block too many than too less and have problems after that. I've learned that I can only trust … our tweeps who belong to the fam for a while. All those new ones have to prove that they're not trolls This sense of "tweeps who belong to the fam[ily]" was often reiterated. The perception of "community as family" was prevalent, as another female user expressed, "People keep telling me that twitter ISIS too important to me but that's not true, I don't care about twitter, I care bout my baqiyah family." The need to protect this family is not just desirable, but a duty. LadyG tweets, "if they block u ten times, keep coming back a hundred times. We r here to support each other bi'idhnillah." This is further enforced by another female Twitter account holder who tweets, "Follow and support our sister and fulfill your obligation to expand baqiya family…" This "family" assumes even more importance to those whose blood families have disowned them for their loyalty to ISIS, as one user tweets, "I might have lost my family due to me following Islam but wallahi Allah blessed me with two families one ISIS the baaqiyaah family and (1)… I have another family so Alhamdulillah all praise ISIS due to Allah swt (2)." Wilayat Twitter: Shared Perceptions of "Online as Frontline" Resistance to suspension through its inclusion and appropriation into online identity fulfills a specific purpose for both male and female supporters, confirming a sense of moral righteousness in the brand of Salafi Takfiri Islam propagated by ISIS, and its broader battle. Twitter and its suspension policy on the other hand represents the dominant authorities, and the broader community of non-believers or "kuffar", as this user highlighted on his return: "New Account You know the drill. Make Allah happy and anger the kufar! Follow and spread" followed by, "… the kuffs are so desperate. Each time they suspend us they take great sin upon themselves." The act of suspension also serves as a metaphor for the greater trials endured by the Ummah, as perceived by ISIS supporters. For example, male user Abu H used his biography to compare perceived persecution via Twitter suspension to the trials of the Prophet, writing, "Muhammad [PBUH] was stopped from giving Dawah by the Quraysh today those upon the haqq [truth] are also stopped banned or suspended. Anyone not stopped is preaching batil [lies]." Additionally, while much analysis points to the propagation by ISIS of the Islamic State as a Utopia, in which the perfect Islamic lives can be lived according to Sharia law, and in a land where Muslims of all colors and races are equal, both males and females in the sample presented much evidence of another perspective: a committed but realistic portrayal of the hardship of the path. The same endurance required to combat repeated suspension was pointed to by many Twitter supporters as necessary in the offline environment, where only the hardiest of ISIS supporters would prevail. This "baqiya" was repeatedly cited in relation to the harsh reality of the Islamic State and the need for the "true Muslim" to keep faith and endure, as this fighter expresses, "Driven by d command 2 fight, angered by d oppression of Muslims n desirin 2 establish Allah's shariah; we left the comfort of our homes. #ISIS." As with Twitter suspension, so with the real struggle. Only those with utter commitment and devotion to Allah would remain with ISIS, "This path was never meant to be rosy. Don't go to jihad intending to live a life of comfort. There'll be blood and fear but never give up:)."

Discussion and Conclusion
The key issue raised in this article is that an important-and neglected-aspect of suspension is not around the numbers making it back, but about the function of returning, as understood by returners themselves. This article suggested that suspension has a number of affective impacts on Twitter supporters, generating both emotional meaning and community, expressed in particular gendered ways. It does not suggest that suspension is ineffective in reducing numbers of supporters. This is clearly not the case. Rather, it seeks to emphasize other less quantifiable ways in which suspension may reinforce the feeling of shared purpose felt by those supporting ISIS online, whether on Twitter or on other social media platforms. It argues for a reconsideration of how we assess the impact of suspension or take-down methods on other sites, recommending a shift toward the recognition of the power of affect, emotion, and online community, as well as quantifiable influence, such as numbers of followers and networks. This represents a shift in thinking about the meaning and function of social media community, with the emphasis away from the more rigid and behavioral categories of terrorism studies ("recruitment," "propagandizing," etc.) and toward an emphasis centered on the meaning that ISIS supporters create for themselves. Central to this is the recognition of social media ISIS supporters as occupants of a highly gendered domain; this is how Twitter ISIS supporters themselves construct the Twitter space. It is relevant however not just for Twitter, but any domain colonized by Islamic State.
The analysis began by suggesting the ways in which gender is evident in ISIS online, by outlining conceptions of suspension as an effective method of combating online extremism, "wasting the time" of Twitter users, and by suggesting that the dynamics of Twitter ISIS fans indicate suspension may impact on their communities in gendered ways. It then challenged the notion that suspension is purely disruptive to ISIS online, first by suggesting gendered communities on Twitter fulfill important roles that have consequences in the offline space; second, by outlining differences in the behaviors of men and women online, and their relative influence; third, by bringing these ideas together to emphasize the affective and community-building effects of suspension. Suspension effects must be viewed not simply in terms of network-disruption; the emotional and gendered effects of suspension also require recognition. These may be formative in radicalization processes, relating as they do to a sense of belonging, us/ them othering, and grievance, all assumed to be pivotal in the path to extremism. 71 This piece contributes to what is known about the constantly changing world of online ISIS support and the function of gender within that. Further research is required on the costs to ISIS supporters, in terms of network, versus the benefits, in terms of community-building, of online take-down as an approach; and how this sense of both "baqiya" and community is ported to other social media platforms. While "baqiya" -or endurance-is the watchword of Islamic State, there is no room for complacency in counter-terrorism policy, when it comes to success either on-or off-line. What can be learned from Twitter must be transposed to the wider online arena.
The piece also suggests the importance of thinking about counterextremism measures such as "take-down" in gendered ways. The impacts on self-identifying male and female Twitter supporters of suspension are different. In the online context, "community" is clearly not synonymous with "network." It is a collective experience, with affective and highly gendered qualities, which cannot be measured in quantitative ways. This is particularly important when gender differences in recruitment are considered. Few cases of migration to Islamic State, whether of state-building women or fighting men, have challenged the theory that the Internet alone is "sufficient" for radicalization. However, there is emerging evidence of the power of the Internet to specifically radicalize young women online, given their sometimes lesser access to the offline public space, where men and boys more easily make the contacts enabling travel to Islamic State. An online sense of community can have an offline effect. If collective identity, gender and community are important factors in understanding offline radicalization-and they are-then these concepts must be regarded as equally important in the online environment too. It is a hope of this analysis that these considerations will be routine in further research on radicalization online.
Annex. Analytics for the data sample.   (32) (*) beside a figure denotes the higher value (male or female); (F) shows that female users had higher mean and media scores in a category; (M) indicates that male users had higher mean and median scores in a category; (A) indicates that there was ambiguity between mean and median scores. Within the sample 2 'super-users' were outliers in terms of numbers of tweets, and length of time on Twitter. Averages in brackets exclude these super-users.