On 22 March 2016 the hashtag #StopIslam began to trend on the social media platform Twitter, after 32 members of the public were killed and 300 injured in terrorist attacks in Brussels for which Islamic State claimed responsibility. Although the hashtag had existed before these events, its use was relatively low-key; it was tweeted 1,598 times in the 50 hours following the Paris terrorist attacks (Magdy, Darwish and Abokodair, 2015) and did not gain visibility in the wider public sphere. In the immediate aftermath of the Brussels bombing, however, #StopIslam grew to prominence, drawing mainstream media attention after it was used in 412,353 tweets (including both posts and retweets) in the 24 hours after the attacks, with almost 40,000 tweets per hour at its peak.
This use of social media appeared to crystallize a European political context wherein overtly anti-Muslim narratives had become entangled with broader concerns about immigration, in the wake of the refugee crisis (Holmes and Castañeda, 2016; Khiabany, 2016; Wilson and Mavelli, 2016). In trending, moreover, #StopIslam resonated with a broader political climate in the Global North, where Islamophobia and white supremacism had become increasingly visible (Ouellette and Banet-Weiser, 2018; Feshami, 2018, this issue); increasingly intertwined (Hafez, 2014; Horsti, 2016); and increasingly legitimated in political terms. The rise of xenophobic nationalism, for instance, has most notably been evidenced in discourses surrounding key events such as: Brexit (Green et al, 2016); the near electoral victory of Austria’s far right Freedom Party (Rheindorf and Wodak, 2017); and the election of Donald Trump as US president (Kellner, 2016).
Against this backdrop of resurgent right-wing populism, the extensive use of the #StopIslam hashtag seems to justify concern that xenophobic sentiment has become progressively normalized (e.g. Kelly, 2017; Siapera, 2018). The circulation of #StopIslam resonates with concern that anti-immigration rhetoric is not just evident on the extreme right, but presented as a legitimate reflection of ‘public mood’ that politicians – at all points of the political spectrum – are compelled to respond to (Forkert, 2017). The hashtag’s popularity underlines the ways that social media have been particularly implicated in the circulation (Groshek and Koc-Michalska, 2017) and even fermentation (Farkas et al, 2017) of far right political discourse. However, a closer examination of #StopIslam suggests that it does not fit as neatly into narratives about right-wing populism and social media as might be expected. The reason why the hashtag trended and was reported on within the mass media was not because it was being used to spread hate speech. Instead #StopIslam had been appropriated by those seeking to criticize and challenge the hashtag’s original meaning.
This paper presents quantitative findings from an inter-disciplinary project between media studies, cultural studies and computer science, which contributes to these pressing political and theoretical debates surrounding the spread and contestation of hate speech on social media. We focus on #StopIslam as an especially visible instance of a hashtag campaign in which a counter-narrative emerged. The project gathered and analyzed tweets (n = 302, 342) that included the #StopIslam hashtag over the 40-day period immediately after the Brussels attacks, in order to explore who was involved in circulating the hashtag and identify dominant trends in how it was engaged with.1
The original aim of the project was to focus on the dynamics of the counter-narrative itself, with initial research questions asking what were the dominant messages and voices in the discourse, what was the relationship between key actors and whether the counter-narrative offered a platform for Muslim self-representation or was predominantly engaged with by would-be allies (who tried to speak for Muslims in more problematic ways).2 Our findings, however, forced us to shift our emphasis and change the types of questions we were asking. Counter-narratives are, by definition, relational in always being explicitly articulated in opposition to dominant narratives, and in this instance the narratives were so entangled, that it was impossible to focus solely on the features of the oppositional narrative. Instead, the presence and on-going influence of right-wing voices necessitated more of a focus on the relationships between the original and counter-narrative, in order to ask which possibilities for critique and resistance were opened up by these relations and which were foreclosed.
In response to this shift in emphasis our findings are organized into three sections. The first provides the context for the original #StopIslam narrative by delineating key features of the hashtag campaign as it originally emerged and began to trend. The second section elaborates on the dynamics of the counter-narrative by contrasting it with the affordances of uses of the hashtag in line with its original meaning. We then, finally, reflect on the additional set of relations that existed between these Twitter narratives and the mainstream media, in order to more fully tease out the political potentials that emerged and those that were undermined during the course of these events.
Our findings initially seem to fit with evidence that shows that online political discussion exists largely within ‘echo-chambers’ or assumes ‘trench warfare’ dynamics where even when opposing perspectives are brought together they speak past one another or clash in ways that entrench pre-existing views (Karlsen et al, 2017). In light of this argument it would be easy to dismiss the significance of ongoing uses of the hashtag in democratic terms. If the aim of right-wing propaganda is to extend and legitimize this agenda in the mainstream, this discourse might seem irrelevant if these voices are merely talking to themselves. We argue, however, that uses of social media to spread the values of the far right should not be trivialised; beyond this particular discursive event it is clear that increased political legitimacy is being afforded to these opinions (as represented most visibly in the election of Donald Trump).
Despite the counter-narrative achieving more mainstream visibility, our findings suggest that such campaigns are difficult to sustain in the face of tight-knit extreme-right networks. After the initial counter-narrative there has been little evidence of contestation of the ongoing use of #StopIslam in relation to more recent attacks in Manchester and London. The dynamics of the #StopIslam hashtag, along with others (Evolvi, 2017; Siapera et al, 2018 and the other papers in this special edition), demonstrates the strategic and instrumental use of social media by right-wing activists, where Twitter is just one component of a more complex media ecology that is working to push mainstream political discussion to the right. In this context, it is particularly important to understand the dynamics of critical counter-narratives and the political possibilities they can (or indeed cannot) offer.
In addition to drawing together urgent concerns about populism – and its contestation – the #StopIslam hashtag resonates with a number of longstanding theoretical debates within media and communications studies regarding the political potentials and limitations of commercial social media platforms. The contestation of #StopIslam, for instance, offers insight into debates surrounding the capacity of digital media to offer space and visibility for the self-representation of minority groups who are often excluded or misrepresented in the mainstream media and events surrounding the hashtag support the formation of counter-publics and counter-narratives. The hashtag also elucidates how these issues have become still more urgent in light of the flourishing use of social media by right-wing actors with links to white supremacist groups. Before turning to our own findings, therefore, it is necessary to examine these existing debates.
Visibility and voice
Questions about whether social media can be used to create counter-narratives need to be set against the backdrop of long-standing debates regarding the misrepresentation and marginalization of particular minority groups within the media (Cottle, 2000). The increase in the cultural diversity of nations should be seen in the context of a neoliberal, global order, riven by conflicts over struggles for hegemony, in which the movement of people is one outcome. Immigration has been seized on by populist groups as a scapegoat for increasing feelings of political and economic insecurity. As Political Islam was constructed as the new enemy of ‘the West’, following a post-Cold war political vacuum, Muslims have become the maligned ‘Other’ on which to project contemporary anxieties (Halliday, 1996). There is substantial evidence for the resulting demonization of Muslims in ‘Western’ media (used with all the caveats around the homogenisation of the West and its media: Poole, 2002; Richardson, 2004; Baker, Gabrielatos and McEnery, 2013; Ahmed and Matthes, 2017).
Within a ‘clash of civilisations’ (Huntington, 1996) and securitization discourse, Muslims have been represented as a cultural and security threat. Largely excluded from mainstream media, it has been argued that digital platforms might offer a space for self-representation to Muslims in order to contest these hegemonic media frames (Brock, 2012; Dawes, 2017). One of the aims of this project was to consider whether social media offered Muslims a place for their ‘voices’ to be heard (Couldry, 2010). Research conducted in the UK, focussing on opposition to the 2003 invasion of Iraq, for example, has pointed to the potential for particular media platforms to gain visibility for Muslim activists (we use this label recognizing the multiple identities at play) (Gale and O’Toole, 2013; Gillan, Pickerill and Webster, 2008). These studies found that digital media allowed Muslims to compete with other social actors over definitions of contentious issues and also offered connectivity, although this was largely symbolic (Gillan et al, 2008). For many theorists, it is the networking potential of digital media that offers marginal voices increased power (Papacharissi, 2015). However, it has also been emphasized that a further important aspect of voice is recognition and being heard (Couldry, 2010).3
In line with these questions of voice, the distinctive role of social media is often articulated as its relationship with the mainstream media. For instance, there is a strong tradition of work that has offered informative applications of late-Habermasian accounts of the public sphere to explore the – often complex and contradictory ways – that digital media can facilitate counter-publicity and the formation of counter-publics who can contest hegemonic media framings (Downey and Fenton, 2003; de Jong, Shaw and Stammers, 2005). This process is not straightforward and frequently involves compromises. For the sake of constructing simple, accessible slogans that have broad appeal, for example, activists often create calls to action that reduce the complexity of issues or appeal to more general values, rather than pushing for structural change (Birks and Downey, 2015). This, along with concern that the underlying infrastructure of the internet does not lend itself to egalitarian politics in the way that is often assumed, has resulted in public sphere approaches to digital media themselves being criticized (Fenton, 2016).
In general, hope for digital media to act as a platform for marginalized voices (Kahn and Kellner, 2004, 2005; Hands, 2010) has waned significantly throughout the first decade of the 2000s, especially in light of a perceived displacement of activist-produced alternative media platforms with commercial social media (Giraud, 2014). Hashtag campaigns are often seen as the apex of these shifts and are routinely characterized as epitomising superficial, fleeting forms of political engagement (Dean, 2010), wherein: ‘Collective solidarity is replaced by a politics of visibility that relies on hashtags, “likes” and compulsive posting of updates that hinge on self-presentation as proof of individual activism’ (Freedman, Curran and Fenton, 2016: 188). Here, in other words, online visibility is framed not as something that can support more sustained movement building, but that actively undercuts it.
Although it is important not to celebrate visibility in and of itself, we nonetheless argue that social media are not wholly reducible to this mode of politics, but play a more complex and messy role within broader communication ecologies (for a related argument, see Mercea, Ianelli and Loader, 2016). For instance, despite all of their shortcomings, hashtags can offer an affectively ‘sticky’ form of public engagement – to use Ahmed’s (2013) turn of phrase – around which publics can coalesce (Papacharissi, 2015). In doing so, these media can play an important role in creating a collective voice or identity for counter-publics and protest movements (Kavada, 2015).
In recognition of the messy but nonetheless significant role of social media, therefore, we take a lead from contemporary research that has sought to overcome polarized debates about whether digital media straight-forwardly support or undermine political action, which has drawn instead on lineages from social movement studies (such as Cammaerts, Mattoni and McCurdy, 2012; Treré, 2019), a field that has always foregrounded the ambivalence of digital media technologies for activists whilst still maintaining a sense of their role in political practice (Pickerill, 2002). This body of work has recognized that social media are not something that can be rejected or avoided due to being entangled with the fabric of political life (Giraud, 2018, 2019), shifting focus instead to how the frictions associated with particular platforms are navigated in practice (Ruiz, 2014; Barassi, 2015; Shea et al, 2015). Social media, from this perspective, are not something to be valorized but are nonetheless understood as having an important political role, as part of broader media ecologies where they work alongside and interact with a range of other media: from pamphlets and email lists to mainstream media outlets (Treré, 2012; Treré and Mattoni, 2016). This understanding of social media’s role offers an informative background for grasping the tensions that surround the role of specific platforms, such as Twitter, in articulating a collective voice, identity or counter-narrative.
Counter-publics and Counter-narratives
Debates about the capacity of social media to facilitate counter-narrative formation have proven especially significant in the context of race and racism.4 Despite all of the criticisms that have been levelled at social media, for instance, Jackson and Foucault Welles argue that they nonetheless ‘offer citizens most invisible in mainstream politics radical new potentials for identity negotiation, visibility, and influence’ (2015: 399). Likewise, Rambukanna contends that although the often-complex political engagements offered by hashtags do not represent the ideal speech situation that characterizes Habermasian notions of the public sphere, they nonetheless offer space for ‘publics that do politics in a way that is rough and emergent, flawed and messy, and ones in which new forms of collective power are being forged on the fly’ (2015: 160). In particular, Rambukanna draws attention to the use of hashtags such as #racefail to draw attention to problematic media representations of race, arguing that their use has created a highly visible ‘welling up of critical race discourse’ (2015: 169). Hashtags, however, have most prominently been cited in the literature not as something that affords new opportunities to critique existing media content, but as a means of drawing attention to what is missing from the mainstream media.
#Ferguson, for instance, has been referred to by a number of thinkers, including Rambukanna, as an instance of hashtags being used to raise awareness of racially-motivated violence and inequality in the US (Brock, 2012; Jackson and Foucault Welles, 2015; Rambukanna, 2015). #Ferguson rose to prominence in the wake of the police shooting of teenager Michael Brown as a means of drawing visibility to his death and police violence against African-American populations in the US more broadly. Jackson and Foucault Welles (2015) specifically point to the way that #Ferguson not only offered a useful rallying point for commentary about police violence, but argue that the hashtag’s success was in part due to shaping mainstream media narratives. Not only did the issue gain visibility in the mainstream media, frames established by #Ferguson set the tone of the narrative. Indeed, ‘early initiators’ in the discourse, including ‘African-Americans, women, and young people, including several members of Michael Brown’s working-class, [and] African-American community, were particularly influential and succeeded in defining the terms of debate despite their historical exclusion from the American public sphere’ (Jackson and Foucault Welles, 2015: 412).
Broadly speaking, therefore, hashtags have been seen as an affectively significant component of socio-technical assemblages through which race is enacted online (Sharma, 2013) and as holding capacity to support intersectional alliance building (Loza, 2014). At the same time, the problematic qualities of social media campaigns have been consistently emphasized and argued to encourage paternalistic modes of politics that often speak for others. In doing so, they reinscribe racial inequalities, rather than provide a platform for diverse voices (Torchin, 2016). Maxfield argues, for instance, that hashtags such as #BringBackOurGirls were ultimately appropriated by ‘White feminists in the Global North’ in a manner that ‘suggests not simply a reiteration of earlier colonial patterns, but an act of colonialization as it continues in the contemporary era’ (2015: 11).5
Appropriation, however, is complex and multifaceted in online contexts; while hashtags originating within particular communities are sometimes appropriated in ways that reinscribe racial inequalities, the case of hashtags such as #StopIslam elucidates how racialized or racist hashtags can themselves be appropriated. Jackson and Foucault Wells (2015) describe this process as ‘hijacking’, a term they use with reference to the appropriation of the #myNYPD. This hashtag was organized by the New York Police Department as a means of soliciting positive photographs and stories about people’s experiences with the police. The hashtag was also, however, adopted by those critical of policing, who used it in conjunction with images of police violence against African-American communities alongside parodic captions.
This appropriation of hashtags in order to create critical counter-narratives offers a helpful lens to approach #StopIslam. As described above, the original hashtag speaks to a political climate in which racialized narratives of Islam have become routinely articulated with anti-immigration sentiment and in which white supremacist narratives have become increasingly visible (Hafez, 2014; Horsti, 2017. In the rest of the article we set out and analyze data associated with #StopIslam in order to develop further empirical and conceptual understanding of the dynamics of counter-narratives on social media platforms, and to better grasp the conditions under which narratives against hate speech can emerge and be sustained.
Twitter is a useful object of study through which to approach issues surrounding counter-narrative formation, contrasting with social media platforms such as Facebook, where people of different political persuasions tend to exist in echo-chambers (Bücher, 2012). While (as we go on to elucidate) echo-chamber dynamics exist on Twitter to a degree, people without pre-existing connections are brought into conversation more frequently. The platform has, correspondingly, been associated with the formation of ‘affective’ (Papacharissi, 2015) or ‘ad hoc’ (Bruns and Burgess, 2011; Dawes, 2017) publics, which are brought together around particular issues and often gain a degree of political purchase, although the loose-knit ties that bind such publics together mean that any influence is often fleeting. The longstanding use of hashtags also means conflicting voices are routinely brought together, which make the platform ideal for studying the emergence of counter-narratives. A further reason for engaging with Twitter is its relationship to its constituent media ecology, where it has a well-defined relationship with the mainstream media (often used as a news source in its own right), which – as touched on above – has given activist counter-narratives opportunity for wider visibility (Jackson and Foucault Welles, 2016).
A substantial proportion of existing research on Twitter can be categorized as the study of ‘big data’ or ‘datafication’ as it uses large scale data sets which are then analyzed through computational methods. These methods include text analytics (e.g. word frequency distributions, pattern recognition, tagging/annotation, and machine learning techniques such as sentiment analysis that can categorize tweets into coded categories). As well as analyzing the tweets themselves, there are also methods for identifying relationships between users such as social network analysis via tools such as Gephi.6 Good overviews of tools and techniques for analyzing social media data for social science researchers are provided by Batrinca and Treleaven (2015) and Ahmed (2017). These techniques are not without their problems, with the main criticisms relating to the assumed accuracy, transparency and objectivity in the way the results are gathered and presented; interpretation is still an integral part of the research design and analysis which results in selective knowledge production (boyd and Crawford, 2012; Vis and Voss, 2013). When engaging with large data sets, it is, therefore, vital to recognize the ethical guidelines and limitations of data visualization techniques as well as our own ideological positions in their design (Kennedy, 2011; Kennedy et al, 2016). We locate our research within the tradition of critical data studies which acknowledges the social processes and power relations involved in data production, whilst maintaining such approaches can still offer valuable insights, especially if combined with other quantitative and qualitative methods (Dalton, Taylor and Thatcher, 2016).
To these ends, we adopted a mixed methods methodology incorporating the following approaches: computational analysis, manual quantitative content analysis and qualitative content analysis (Cresswell and Clark, 2007). By triangulating our methods in this way, the study sought to overcome some of the issues with relying solely on computers to gather data. One of the problems with using computational methods alone was highlighted by our initial use of sentiment analysis (an approach where computers use existing dictionaries to search for positive and negative words to establish the tone of particular tweets). Because the original tone of our hashtag, #StopIslam, was negative, many of tweets were returned as ‘negative’ because they included words such as ‘stupid’ and ‘ignorant’. These tweets, in our interpretation, however, should be perceived as ‘positive’ because they criticized the original tweeter for their hostile stance. In any event, the majority of tweets were returned as ‘neutral’ as they included both positive and negative evaluations (83%).
We used computational methods to search for frequently used words in the tweets and Twitter profiles of users as well as most commonly used related hashtags, most shared tweets, most prolific users and network data. But to avoid misinterpretation we corroborated these methods with traditional quantitative content analysis, manually coding the content of tweets according to pre-determined categories, an approach that also enabled us to delve deeper into significant issues exposed by our analysis (Sloan and Quan-Haase, 2016).7 For example, not only were we able to identify the locations of the users by coding biographies (where computer generated data is unreliable) but we were also able to code the tweets into topics to establish frameworks of coverage. We drew on further qualitative content analysis to provide a more detailed examination of the combination of structure, language, imagery and interactions of individual tweets, an approach that (in line with Deacon et al, 2007) proved valuable in establishing the construction of activist narratives and identity. Using this triangulated approach ensured that we could capture a comprehensive and robust picture of the dynamics within this hashtag as a ‘discursive event’ (Rambukanna, 2015). This article is based on the quantitative analysis of the data and further qualitative work is forthcoming.
Twitter, created in 2006, had, in the first quarter of 2018, around 336 million active monthly users (Statista, 2018) generating more than 500 million tweets per day (Internet Live Stats, 2018). Hashtags operate as a user-driven organisational tool for categorizing topics or events so they are easily searchable by others; they also tie users into an existing conversation, extending their networks (Dawes, 2017). Hashtags therefore structure both content and contributors into more cohesive groups, forming temporary communities around specific issues or a set of values (Bruns and Burgess, 2011). Acting as both framing and discursive devices, then, they enable both the production and circulation of ideologies and identities (Giglietto and Lee, 2017; Papacharissi, 2015; Rambukanna, 2015). For this reason, they are valuable research topics.
We used Twitter’s enterprise API (Application Programme Interface) platform GNIP8 to ensure the full data set was collected, that is all tweets using the hashtag #StopIslam from just before the attacks (on 22 March 2016) and for the forty days following it (21 March – 29 April 2016). After removing the 551,400 spam tweets we received from Twitter, we were left with 66,764 unique tweets and 235,578 retweets (shared original tweets), 302,342 in total. These unique tweets were analyzed both separately and with the retweets in order to identify similar and contrasting trends. As well as applying content and descriptive analytics, we also undertook a network analysis of those users who had retweeted others and been retweeted. This was followed by a manual quantitative analysis of a sample of the most-shared 5,000 retweets. We used a coding schedule to measure the date and time of tweets, location, gender, religion and political and/or institutional affiliation of the tweeter, topic of the tweet, and to ascertain whether this was part of the dominant narrative (against Muslims) or counter-narrative (that contested the original negative narrative). After sorting for deleted accounts, this left us with 4,263 tweets. To adhere to ethical guidelines, in the presentation of these findings we predominantly include quantitative data that has been processed in ways that do not identify individual users and only reproduce tweets that were shared by multiple users, which are largely memes or widely-shared slogans engaged with by the mainstream media.
The qualitative analysis of 150 most retweeted tweets and their comments (sometimes up to 500) is not provided here as it is beyond the scope of this article, which focuses instead on establishing the broader trends and patterns that we identified.
Establishing a discourse: #StopIslam and the political right
Before turning to the counter-narrative it is necessary to establish the context of the original emergence of the hashtag. The first thing to note from the data is that the activity took place mainly in the 24 hours following the attacks, which seems to be in line with similar Twitter activity, following the Paris attacks for example (Siapera et al, 2018). Figure 1 shows how uses of the hashtag peaked between 4–7pm on 22 March 2016, the day of the attacks.
In order to establish the content of the discourse within this hashtag, we undertook word analysis, using frequency as a measure of significance. This revealed the negativity of the hashtag, as one would expect, given its positioning as #StopIslam. The wordclouds (Figures 2 and 3) illustrate the negative content of a sizeable number of tweets, with the most frequently occurring words appearing larger. The words have been coded in the following way: negative evaluations in red, positive in green and descriptive nouns relating to the topic (Islam, Muslims, religion) in black.9 It was important to analyze the original tweets and retweets separately to identify the content of those most likely to be circulated. For example, it was clear through the analysis of retweets that a large number of these were used to reflect on what the original tweets were saying (the words that suggest this are marked in yellow).
We also conducted an analysis of the most common related hashtags to further establish topics and their framing; this also revealed significant actors in the discourse (Table 1). What was immediately evident was the intervention or even propagation of this hashtag by conservative actors, mostly based in the USA (coded blue in Table 1). These findings suggested the hashtag was being leveraged in distinct political ways: through its connection to other hashtags (e.g. wakeupamerica, makedclisten) it was being used to promote conservative agendas, while the presence of hashtags such as #trump2016 or #trumptrain indicated Islamophobic discourse was being used more specifically to promote and strengthen the presidential campaign of Donald Trump.
In Table 1 the Islamophobic hashtags are coded red; hashtags circulated by right-wing groups are coded blue (of course many of those coded red are being circulated by conservative groups but not exclusively). Hashtags showing mixed political sentiment are coded in purple, with only one outrightly positive related hashtag coded in green (#stopignorance) and neutral hashtags coded black. The appearance of #billwarnerphd shows the centrality of particular figures in supporting anti-Muslim discourse, often those who presented themselves as experts about religion and geopolitics (whose opinions were used to legitimize negative characterizations of Islam, in this case via the website Political Islam, which operates as a resource for right-wing activists).
We sought to explore and further confirm these findings through a word analysis of the biographies of users (Figures 4 and 5). These biographies are self-defined so it is somewhat unsurprising that people use positive evaluations to describe themselves and what they enjoy (coded green). The relatively high number of conservatives is again evident (coded blue); in unique tweets solely n = 814, all tweets n = 2858, with Christians also outnumbering Muslims (coded purple) (unique tweets n = 431:384, all tweets n = 1530: 1069). Table 2 reveals the preferences (in likes/loves) of these groups. There are more self-identified male actors than female (in unique tweets n = 1587: 868, all tweets n = 5783: 4396). However, it should be noted that this data only reveals those who explicitly use these descriptive terms with others using descriptors such as occupational terms (coded yellow).
To further understand the actors participating in the discourse we explored the collocations around these words which confirmed that the most prolific actors were politically on the right, even the extreme right (Table 2). Although these numbers are small relative to the corpus overall, they represent significant clusters; other users appear to be more differentiated. Only two of the top ten actors fall outside the category of most frequently posted.
The data is revealing of a particular (performed) political identity of a significant ‘community’ of tweeters who often identify not only as ‘conservative’ but ‘Christian’ and ‘patriotic’.10 By manually coding the biographies and tweets of those shared the most (n = 4263), we were able to gather a more accurate picture of the geographical distribution (Figure 6) of tweeters. Again, the US was confirmed as the largest single country using the #StopIslam hashtag, followed by the UK and Pakistan. Indeed, there were a large number of non-European actors involved in engaging with the hashtag which is illustrative of the transnational characteristics of these hashtag campaigns.
As well as focusing on who was participating in these anti-Islamic narratives, a key question we were concerned with in our research questions was what these actors were tweeting. Our content analysis showed that the only country where Islamophobic discourse significantly outweighed counter-discourses was the US (1,023 anti-Muslim tweets or ‘dominant narrative’ [DN] compared to 297 supporting Muslims, the ‘counter-narrative’ [CN]). The only other countries that had marginally more anti-Muslim tweets were Canada (28:18), Australia (21:13) and Germany in this sample (16:13) but the figures are less significant in proportional terms. Similarly, in this sample, 588 of the self-identified ‘Christians’ tweeted the dominant narrative compared to only 24 sharing the counter-narrative (Figure 7); atheists were also more likely to tweet the dominant narrative (67:6). The fact that not everyone chooses to include their religion in their Twitter biography adds weight to the idea that where religion was used as identity marker it was often a component of a broader performance of a particular political identity. Whilst it could be argued that these findings may be a result of the focus on English language tweets, other English-language countries were more likely to circulate counter discourse (the UK, for example: 331 CN compared to 105 DN).
Overall, our results mirror the findings of similar studies of Islamophobic and racist hashtags, which have demonstrated the strategic and instrumental use of Twitter by the far right to mobilize activists and extend the reach of their propaganda by connecting to other conservative groups (Dawes, 2017; Evolvi, 2017; Siapera et al, 2018). However, the prominence of the counter-narrative adds another layer to this research, in foregrounding moments when this discourse was contested.
Contesting #StopIslam: The formation of a counter-narrative
The rest of this article focuses on the counter-narrative that emerged in response to the hashtag, but in order to fully grasp its affordances it is still helpful to contrast it with the original narrative itself, both because it is difficult to disentangle from the original and because a comparative approach is useful in making sense of its dynamics. While there is clearly a clustering of right-wing groups circulating the dominant narrative, an analysis of the most shared tweets demonstrates a different pattern. The most retweeted tweets appear to be defending Muslims, establishing a counter-narrative against #StopIslam. Table 3 shows that in the top shared tweets, only one (number 9) carried the original meaning of the hashtag (only identifiable through a meme, which showed the wording was intended sarcastically, see Figure 8). The top shared tweet, defending Islam, was retweeted 6,643 times whilst the top tweet attacking Islam was shared 1,500 times; the next most shared dominant narrative was only retweeted 761 times.
|1.||y’all are tweeting #StopIslam when … (see image below)||6,643|
|2.||Why is #StopIslam trending? It should be #StopISIS||3,791|
|3.||I said this earlier today, but seeing this ridiculous hash tag made me want to re-share. #StopIslam open your eyes. (includes meme negating link between Muslims and terrorism)||3,420|
|4.||#StopIslam? Eyh, the muslim boys next door bring me tom yam whenever I’m sick. Why would I stop kind souls like them?||2,500|
|5.||Educate Yourself. #StopIslam is pathetic! Terrorism has no place in Islam. (includes meme with a quote from the Quran forbidding murder)||1,989|
|6.||Are white lives more precious than Muslim/Arab lives. Terrorism has no religion so focus on issue not #StopIslam (includes a meme on the hypocrisy of showing sympathy with only white victims of terrorism)||1,769|
|7.||Why is #StopIslam trending? It should be #StopISIS||1,755|
|8.||I wish #stopisis was trending instead of #StopIslam. The act of ignorant and bad people doesn’t mean we should blame all religion.||1,603|
|9.||#StopIslam is pure Islamophobia. We know Islam is a religion of peace and terror has no religion. (includes visual data that suggests Islamism is responsible for most acts of terrorism)||1,500|
|10.||“If you didn’t study Islam, Please don’t say anything about Islam” Islam doesn’t teach terrorism. #StopIslam (includes a photograph of the source of the quote)||1,463|
A key characteristic of the counter-narrative was the use of infographics, URLs and memes to undermine the dominant narrative. Our manual content analysis allowed us to further analyze the characteristics of these tweets. Geographically, counter-narratives were shared mostly by tweeters in the UK (74% of the UK sample) and the MENA regions (86%). 95% of tweets from Turkey shared the counter-narrative and 97% from Pakistan.12 There was no significant difference in the gender of those sharing dominant or counter-narratives. The counter-narrative was more likely to be shared by Muslims (99%). We further coded the top tweets (including retweets) into topics and then themes (Table 4).13 This analysis shows the framing of both narratives was extremely narrow as 95% of tweets could be categorized within these themes. The data also demonstrates how the right-wing actors have politicized this hashtag, tying it into anti-left agendas, whilst the counter-narrative focuses on defending Islam and Muslims.
|Negates the relationship between Islam and Terrorism||1,419||33.3|
|Islamification/spread of Islam||967||22.7|
|Islam as a negative force||704||16.5|
|Islam as a positive force||397||9.3|
|Muslims as victims/discrimination/Islamophobia||141||3.3|
|Reflecting on the # trending||127||3.0|
|Anti-far right discourse||22||0.5|
|Other identified themes||222||5.2|
Whilst Table 4 shows that the most popular individual topic of tweets included arguments that negated the relationship between Islam and terrorism, when topics are combined there is a fairly even split between positive (46.4%) and negative (44.1%) discourses about Islam. This could be because there was less diversity in the counter-narratives being shared. There was also a difference in information flow (Figure 9). As an existing hashtag (prior to this event), the dominant narrative had more longevity, while the counter-narrative was predominantly shared in the 24 hours after the attacks. The qualitative analysis demonstrates the amount of flak vociferously generated by the right to close down the opposing discourse. In terms of longevity, given that the dynamics of #StopIslam mirrors similar events such as the counter response to #JesuisCharlie following the Charlie Hebdo terrorist attack in 2015 (Dawes, 2017), this suggests that this could be the natural life of a counter-movement on Twitter.
The decline of the counter-narrative is significant in light of the ongoing use of the hashtag following more recent events, such as the 2017 terror attacks in Manchester and London. In these contexts, it has reverted back to its original function (to spread Islamophobic discourse and extend right-wing agendas). However, even though the counter-narrative was short-lived, the data does demonstrate the potential of digital media platforms to galvanize and create collectivities which can have discursive power. This is sharply illustrated by the selection and reporting of the dynamics of this hashtag in the mainstream media.
Reporting #StopIslam: A successful ‘hijacking’?
This hashtag first became visible to us due to its reporting in mainstream media, including CNN, Daily Express, Daily Mirror, Daily Star, Russia Today and The Washington Post among others. To examine the mainstream media uptake further, we analyzed the affiliations of the accounts of the top 5,000 tweets and also qualitatively analyzed the 100 users with the most followers (see Methodology). Most of these top 5,000 tweets were disseminated by individual users (78.8%), with a slightly higher proportion of accounts tweeting the counter-narrative (57.3% compared to 41.3% the dominant narrative). 9.6% tweets were circulated by what could be defined as ‘alt-right’ groups; this was the only affiliation of top tweeters that was more likely to support the dominant narrative (99.7%). Media institutions and celebrities (3.8%), in contrast, were more likely to report on the counter-narrative (67.4%).
If we look at the top 100 accounts in our dataset who have the most followers, 22 of these were media (news) organisations including Al Jazeera, CNN, Nigeria Newsdesk, The Independent and The Washington Post. As 64% of these institutions reported on the counter-narrative, this adds further weight to the argument that the hashtag was successfully appropriated by a counter-movement in order to gain visibility for anti-racist, inclusionary discourse. Other users with large numbers of followers were also more likely to support the counter-narrative. Only two of the 18 unverified organisations (including far right news site Breitbart),14 one of the 43 highly differentiated individual users and three of the 17 verified users (including far right anti-Muslim Dutch politician Geert Wilders) perpetuated the dominant narrative.15 Both the most shared tweets and the tweets disseminated by accounts with the most users were more likely to support or reflect on the counter-narrative. The success of this counter-narrative echoes the claims of a number of academics that particular uses of Twitter can enable grassroots collectives to change a story that seeks to spread hatred into something much more positive (Jackson and Foucault Welles, 2015, 2016; Dawes, 2017).
Networks and echo-chambers: Evidence of ‘connected communities’
The success of the counter-narrative seemed to suggest that, because of the particular characteristics of Twitter, people do often stray out of their ‘echo-chambers’ to interact with others and challenge views they dispute, and that the use of hashtags and retweeting provides this functionality. However, we ultimately found that the evidence for segmentation is still strong. Figure 10, for example, demonstrates the polarized collectivities that coalesced around #StopIslam.
The network diagram shows the connections between users in a retweet network. It is composed of 1,944 users who have had their own tweets retweeted (to reduce the network to ‘key’ users), with the size of each circle (representing a user in the network) related to the number of times they have been retweeted. The lines between each user link the user being retweeted and the user who is retweeting them. Through a closer analysis of the users (made anonymous here) it was possible to establish that those on the left represent the voices attacking Islam whilst those on the right are engaging in the counter-narrative. The left cluster reveals the large nodes of prolific and authoritative users of the hashtag and their influential networks. It also demonstrates the tightly-bound networks of the (political) right compared to the more diffuse supportive users. It is this structure that allows right-wing ‘serial transnational activists’ (Mercea and Bastos, 2016) to be so effective in spreading their ideologies. These contrasting dynamics of the ‘echo-chambers’ we identified may, through density or dispersion, block out or shut down the further circulation of counter-narratives, as evident in the longevity of conservative voices (Figure 8). However, the affordances of Twitter, based on its feature of retweeting, can create room for the emergence of issue-based counter publics in specific circumstances and moments. Whilst distributed by individuals, they produce a coherent message around #StopIslam, so become simultaneously fragmented and collectivized (Siapera et al, 2018).
The challenge offered by counter-narratives can be quickly subsumed, therefore, due to the ad-hoc nature of the communities that cohere around them, in contrast with right-wing online communities that have more established information-sharing patterns. What appears to be different about our study from existing research is that it shows how the mainstream media, usually part of the dominant narrative regarding the representation of Muslims, briefly featured a grassroots counter-narrative, offering a glimpse of future possibilities for progressive politics. Twitter alone may not be able to challenge established power structures, however, by adopting an approach which recognizes its role as a part of a system of ‘hybrid media’ (Chadwick, 2013; Treré, 2019), activist groups may be able to propel their voices into the public sphere. At the same time, the activities of right-wing groups here show how user-led political participation needs to be strategically supported to have a greater impact.
Discussion and Conclusions
In this article we traced the dynamics of the Islamophobic Twitter hashtag #StopIslam and found that at the point when the hashtag was shared the most frequently, it was not primarily being used to circulate anti-Islamic sentiment but had been appropriated by users (including both Muslim self-advocates and would-be allies) seeking to contest hate speech. A ‘discursive event’ (Rambukkana, 2015) that was originally an attack on Muslims, therefore, had been transformed into a defence.
Previous studies, examining racist and exclusionary discourse on Twitter, have similarly demonstrated the centrality of the US right in contributing to the virality of this discourse (Magdy et al, 2015). Densely connected and persistently used, #StopIslam was one of a range of interrelated hashtags associated with right-wing populist sentiment that were used by various semi-organized political groups campaigning for Donald Trump at the start of the US election period. We do not wish to homogenize the range of actors involved in propagating these discourses (as they seem to range from individuals who support Trump to more organized extreme-right news outlets), yet at the same time, and as Evolvi (2017) argues, the affordances of Twitter do appear to enable disparate right-wing voices to come together in ways that give the sense of a collective identity, an identity often established in explicitly antagonistic relation to minority groups in particular. For Siapera et al (2018) the capturing of long-standing hashtags to this effect is an example of how information flow on Twitter has become instrumentalized and manipulated strategically for political purposes. These ‘strategic publics’ are driven by an identity politics provoked by an affective, and often shared, response to a specific issue (Dawes, 2017).
Due to the random and fluid ways in which they develop, collectivities that emerge on Twitter have also been conceptualized as ‘ad-hoc’ or ‘networked publics’, or ‘connected communities’ (Bruns and Burgess, 2011; Dawes, 2017). The focus on a particular cause and the speed at which these communities can form allows them to be inclusive and can, at times, change the direction of a dominant discourse through the political participation by or on behalf of excluded groups (Brock, 2012; Dawes, 2017; Papacharissi, 2015; Rambukanna, 2015; Sharma, 2013; Siapera et al, 2018). Dawes’s (2017) analysis of the counter response to #JesuisCharlie – itself framed as an issue of freedom of speech after the attack on the French satirical magazine Charlie Hebdo – argues that the alternative hashtag #JeNeSuisPasCharlie demonstrates the heterogeneity of voices that are ‘connected’ by a shared reaction to the dominant frame. The emergence of such communities can temporarily subvert such frames, briefly allowing marginalized voices centrality. The loose-knit dynamics of these communities, however, can impact on how purposeful they are, hence the use of ‘connected’ rather than ‘collective’ to describe them (Dawes, 2017). Siapera et al. (2018), moreover, argue that the power of these counter publics remains ‘liminal’ as opposed to the structural, more permanent power of established groups and institutions.
Our research builds upon these existing findings, but in this case the prominence of the counter-narrative also offers an alternative frame for a number of mainstream media outlets – including CNN, BBC and Al Jazeera – in their portrayal of public sentiment after the attacks. The contestation of #StopIslam, in other words, does appear to be a productive instance of what Jackson and Foucault-Welles (2015) term online ‘hijacking’, wherein the original meaning of an online narrative is transformed through counter-public intervention in order to re-frame how an event is represented in the mainstream media. Whilst it could be argued that the power of the counter-narrative was often depoliticized in mainstream media in the way #StopIslam was reported (with articles often reflecting on the counter-narrative trending rather than its content) this momentary inclusion did offer some recognition to usually marginalized voices.
These findings suggest that more research is needed into how the media works as an ecological system, using methods that combine big data with a more qualitative approach to digital activism that could contribute to establishing the digital media practices that give specific narratives visibility and traction. As social media becomes more central to political participation (Kriess and McGregor, 2017) but continues to favour dominant and populist groups through its political economy (Groshek and Koc-Michalska, 2017) research needs to adopt a longitudinal approach to reflect on longer terms patterns and shifts in the dynamics of communication online.
The circulation and contestation of #StopIslam thus speaks to the present political context, in drawing together concerns about the rise of right-wing populism, white supremacy and normalization of nationalistic and xenophobic sentiment targeted at particular communities. In particular, the hashtag ‘campaign’ seems to bear out concern about the imbrication of social media in these discourses (Evolvi, 2017). At the same time, responses to #StopIslam also promise a glimmer of hope regarding the capacity of social media platforms to also be used to challenge hate speech, even if such uses are fraught with compromises and difficulties. The political significance of digital media has often been overstated (Freedman et al, 2016) and in analyzing the dynamics of #StopIslam we are not seeking to make broad claims about social media, but to develop a clearer sense of how counter-narratives against hate speech can emerge, circulate, and gain wider visibility. While the model of participation Twitter offers is clearly limited, this study demonstrates its potential for a more distributed production of discourses and the capacity for connected communities to participate in protest when an issue cuts across a broad range of socio-political identities. Although this type of counter-narrative formation should not be valorized in and of itself, therefore, we nonetheless argue that it is still an important component of contemporary media ecologies that needs to be better understood. In light of seemingly successful uses of social media by right-wing communities, including white supremacist groups, in extending their discourses outwards, it is particularly important not to dismiss these opportunities for contestation, however fleeting.