Regimes of visibility and the affective affordances of Twitter

Literature review

A distant elsewhere for much of the Twitter audience, ² the Syrian conflict has seemingly reduced felt distance when the consequences of the refugee crisis are brought to the fore, primarily through images depicting atrocities. Most notable were the images of Alan Kurdi (born Aylan Shenu), a 3-year-old Syrian boy, who drowned during his migration journey from Turkey to Greece. His body washed ashore near Bodrum, Turkey on 2 September 2015. That same day heartbreaking photographs taken by Nilufer Demir were published on Twitter and attained viral status within minutes (Vis and Goriunova, 2015). As Twitter users helped diffuse these highly emotive images, they were also attuning affectively and aligning emotionally, constituting global, digital flows of affect (Döveling et al., 2018). Still, interpreting such flows of affect as a direct consequence of the affective potential of emotive images and their circulation on social platforms should be regarded with caution.

As Zakaria Sajir and Miriyam Aouragh (2019) explain, the content of images can be similar, yet result in very different courses of action or inaction. They compared the cases of Alan Kurdi and Omran Daqneesh, a 5-year-old pictured sitting in an ambulance after sustaining injuries from an airstrike in Aleppo, Syria. The authors discuss how the first case was followed by a surge of solidarity with refugees, while the latter, though also triggering a sense of compassion, did not garner significant support and instigated a perversion of compassion (Arendt, 1973; Sontag, 2003), resulting in apathy.

Contextual factors are inherently involved in determining whether an image has a profound impact on the general discourse. The idea of states ‘under siege’ by an influx of refugees (Hage, 2016) and the use of Islamophobic tropes widen the empathetic distance between Western public opinion and refugees. This leads non-governmental organizations (NGOs) to adjust their communicative strategies and, correspondingly, their photographic depictions of refugees. Though the present study is not about determining what images made a difference in terms of societal impact, many of the highly retweeted depictions of refugees tweeted by NGOs should be understood as a counter-narrative to negative news media frames. Rather, we focus on determining which images are elevated into prominence within the socio-technical environment of Twitter and what they communicate. Before doing so, we review the existing literature on representations of refugees and constructions of otherness, and situate the latter within the context of social media specificities for witnessing distant suffering, thereby addressing key technological affordances of Twitter.

Felt distance and otherness

Judith Butler (2006) advances a hierarchy of grievability that delimits who is deserving of our sympathy, often communicated via the construct of the suffering ‘other’ based on commonalities with the spectators (of suffering). To fully understand how this plays out, we use Lilie Chouliaraki’s (2013) concept of post-humanitarian solidarity as a theoretical framework to assess Twitter photographs within our dataset. Specifically, this concept outlines how distant sufferers are rendered invisible due to the self-reflexive nature of contemporary solidarity, which inhibits solidarity for suffering ‘others’ that are unlike us.

Both Chouliaraki’s (2013) ‘distant other’ and Annette Markham’s (2013) ‘similar other’ are ways to theoretically (re)construct suffering; particularly, how the suffering of (distant) others gets constructed through media depends upon a deeper understanding of otherness itself. Roger Silverstone (2006) proffers an ethics of care – an ethical proposal which begins by recognizing unfamiliar others as others with humanity. This ethics of care is based on a particular politics of representing otherness, dubbed proper distance, which is grounded by two ethical approaches to otherness. First, the ethics of a common humanity states that moral imagination is a universally shared capacity: ‘we,’ the self and the other, share a space of proximity that is all-inclusive of the human species. By contrast, the second symbolic ethics of strangeness suggests that both self and other emerge as effects of representation, which is why the other can never be fully transparent or intelligible to ‘us’. It is only by acknowledging this distance that we might begin to form moral bonds with the other (Eagleton, 2009).

Despite its profound influence, the notion of a common humanity fails to recognize that the proximity it celebrates is grounded within a Western discourse of ‘the human’. For instance, Chouliaraki (2013) demonstrates how proximity and distance are articulated in contemporary solidarity practices. Inspired by a quiz from Action Aid, Chouliaraki (2013: 1) suggests that a new emotionality – how one feels – steers contemporary solidarity: ‘There is no doubt that emotion has always played a central role in the communication of solidarity, yet . . . there is something distinct about the ways in which the self figures in contemporary humanitarianism.’ In such appeals for solidarity, the suffering of victims begets the emotions of the spectator. This self-reflexivity obfuscates the underlying causes of the depicted suffering, obscuring reasons to act as we are consumed by our own emotions. Chouliaraki’s (2013) proposed alternative of agonistic solidarity may be seen as an attempt to reach a cosmopolis: a space wherein we may imagine ourselves caring for others, not because they are reflections of ourselves but because they are different from us.

Expanding on this, social media images that allow their spectators to engage in self-reflexivity – enhanced when people can socially identify with the depicted (Reicher, 2001) – might be more likely to get elevated into prominence (i.e. through sharing) and become part of a regime of visibility. Such regimes align with the typology advanced by Chouliaraki and Tijana Stolic (2017), who studied news images of the refugee crisis of 2015. Therein, the scholars built upon a notion of visuality understood as: the public horizon of what we see and relate to in the media, the semiotic domain wherein a specific ‘politics of representation’, the ‘struggle over who is to be represented’ and how this representation should be interpreted, is played out (Mirzoeff, 2009: 76). Chouliaraki and Stolic (2017: 1167) understand regimes of visibility as:

the main analytical unit in that they provide the organizing principle underlying their reconstruction of specific visualities of refugees: visibility as biological life is associated with monitorial action; visibility as empathy associated with charitable action; visibility as hospitality, associated with political activism; and visibility as self-reflexivity.

We suggest that self-reflexivity combined with the personalized logic of social media creates an amplification of personalized messages of humanitarianism which taps into audience characteristics instead of the subject of images. Thus, victims are made visible only when they may be re-appropriated within a scene that is recognizable and identifiable. This also connects to Reicher (2001) and Wetherell (2015), both of whom assume social identification as a crucial prerequisite for ‘copying affect of others’, having consequences for the amplification of certain visual representations of distant others.

Twitter’s affective architecture

Essential for understanding the socio-technical dynamics of Twitter is its built-in asymmetric relationship between users, who are not ‘friends’ based on a reciprocal relationship, but instead subscribers (i.e. followers; Schmidt, 2014). For example, the UN High Commission for Refugees (UNHCR) follows 1,481 accounts but is followed by more than 2.3 million users. Thus, in the case of Syrian refugees, we encounter a distinct power asymmetry, whereby NGOs and mainstream news outlets have large followerships while following significantly fewer accounts themselves. The images introduced by these accounts thus resonate due to audience size and subsequent retweet activity, thereby exponentially increasing the affective potential of accounts with significant follower counts.

One way users may extend their reach beyond their own follower networks is through hashtags. Because hashtags are indexable, they connect content from users with no pre-existing follower–followee relationship (Bruns and Moe, 2014). For image tweets, hashtags can function as ‘captions’ that textually anchor to the meaning or content of an image (#children) or ‘relay’ whereby the tag adds information that cannot be derived from the image (Barthes, 1977). As Susan Sontag (2003: 25) outlined: ‘Whether the photograph is understood as a naive object or the work of an experienced artificer, its meaning – and the viewer’s response – depends on how the picture is identified or misidentified; that is, on words.’ Thus, hashtags – beyond acting as repositories of feelings and emotions – not only bring issues into focus, but also function as indicators of contextual meaning (for images).

Retweeting as signs of affective investment

Twitter users may bring messages from the hashtag level to the attention of their followers via retweeting: a performative affirmation of the contents of a particular tweet and a way of spreading a conversation more widely. As danah boyd et al. (2010) suggest, retweets do more than spread messages widely, they provide a structure for conversation and comment. A case study into reply-and-retweet behavior of Dutch politicians (data gathered from 1 February to 31 August 2012) found that, while replying was unaffected by party affiliation, retweeting was very much structured by it (Paßmann et al., 2014). Moreover, Dutch politicians preferred retweeting their own messages over retweeting messages by members from opposition parties. Retweets are thus often seen as a form of endorsement, while replies appear to be a mode of communication among fellow party members.

By affording an immediate response to content, retweets make visible how individual bodies self-report an affective charge of investment, of being touched (Cvetkovich, 2003: 49). As retweets may be understood as endorsements (Paßmann et al., 2014; Papacharissi, 2015) and signifiers of emotional investment – however small or ephemeral – we might infer that they co-create and/or contribute to emotional alignment.

Within the current scholarly debate pertaining to the capacity of mediated representations to humanize distant suffering (Linfield, 2010; Sontag, 2003), to mobilize public opinion and imagination (Hariman and Lucaites, 2007; Sliwinski, 2011), and to propose symbolic spaces of empathetic connectivity and solidarity (Chouliaraki, 2006; Silverstone, 2006), we contribute in describing which humanitarian imagery is ‘allowed to be seen’ through the socio-technical interplay of powerful actors ‘narrowcasting’ images and networked publics that, through their relational affective states, shape the prominence of certain images within Twitter. We thus advance the following research questions:

Methods

Although illustrations, cartoons, and meme-like imagery were part of the dataset of image tweets, they did not resonate as much as photographs. We therefore chose to focus on resonating photographs (and one video ranking number one in the globalized context) that construct refugees and their identities. We study these by analysing their place in a greater network that sorts images by their content elements (image-label networks clustered by automated labels) and proceed by analysing the construction of otherness for refugees and their suffering in resonating photographs.

Studying images on social media entails working with ‘qualitative data on a quantitative scale’ (d’Orazio, 2013), which introduces challenges related to the size of datasets and the amount of work required for coding and tagging materials. When analysing large numbers of images, both content analysis and cultural analytics have proven valuable; however, they do not account for the circulation and networkedness (Rose, 2016) of images within socio-technical environments. Generating rich insights into visual platform vernaculars is the plotting of images according to characteristics, such as time of posting or colour (Pearce et al., 2018). However, this too ignores depicted objects (and subjects) that symbolize – or signify – meaning; these elements cannot be ignored as they inform how we should read the visual just as much as text does. As Martin Hand (2016: 1) explains: ‘In social media, the circulation of visual data destabilizes research objects in ways that challenge visual analyses of textual meaning.’ Hand is part of a larger scholarly call (e.g. Leszczynski, 2019; Rogers, 2013; Rose, 2016; Schreiber, 2017) that urges social media researchers to consider the platform’s architecture when researching images circulating within them. Overall, our approach blends qualitative visual analysis (Rose, 2016), digital methods (Rogers, 2013), and network analysis (Borgatti et al., 2018; Decuypere, 2019).

Samples

Data for the present investigation was gathered through the Digital Methods Initiative Twitter Capture and Analysis Toolset (DMI-TCAT; Borra and Rieder, 2014), using relevant Arabic and English hashtag queries ³ to cull subsamples pertaining to Syrian refugee and migrant communities from a larger dataset pertaining to the Syrian War and refugees (see Appendix A for full list of query terms). We disentangled the language spaces, querying similar hashtags separately, to reveal potential differences in the discursive framing of migrant issues. While we do not suggest that language differences signify separate cultural groups or demographics, such an approach may reveal more localized perspectives (Rogers, 2019). The disentangling of tweets based on language resulted in tweets whose metadata revealed distinctly different locations; whereas the English language dataset contained tweets with international location metadata ⁴ (largely UK- and US-based) the Arabic language dataset predominantly included locations in the Middle East (namely Egypt, Lebanon and Syria).

Justification for the selection of particular hashtags within these queries emerged through frequency outputs, such that highly recurring hashtags with explicit mention of refugees or migrancy were included. Once these subsamples were identified, additional queries were used to extract images and parse Twitter content by text, co-hashtag occurrence, and retweet frequency. By triangulating visual content resonance in this way, our mixed-methods approach provided a snapshot of the Syrian refugee digital ecology within Twitter, which, as we will discuss, included both migrants and spectators from Arabic and English spaces. In total, these hashtag spaces included 117 images – 61 from the English language space and 56 from the Arabic space.

Automated, network, and critical analyses: a qualitative interpretation

Demarcating data based on hashtags means that the images collected are already contextualized in specific ways: their hashtags add meaning and reveal intentionality of tweeted content. Within the two demarcated, contextualized hashtags spaces (English and Arabic) we de-contextualized images by reorganizing them based on their denotative content elements: who was in the image, what activities were pictured, and what equipment and surroundings were included (Barthes, 1977). Such elements inform how we should read the several ways in which images can be read just as much as tags and words surrounding the images do. In a later stage, the connotative meanings were derived by assessing how denotative elements were contextualized by both hashtags as well as tweeted texts.

Denoted messages (i.e. who and what are depicted) embedded within visuals were examined through automated object-label annotation via the Google Cloud Vision API. We selected this API as it currently outperforms other Vision APIs vis-à-vis object-label specificity (Mintz and Silva, 2019). We made use of the Memespector Tool (Rieder et al., 2018), which taps into the Google Cloud Vision API and generates different analytical outputs, such as object labels and network visualization files.

Next, we created bipartite networks containing both images and their annotations in order to visualize and cluster images according to their embedded content. Furthermore, we incorporated retweet metrics to size images, thereby making visible their attendant objects and subjects. Networking images by their objective content (annotated labels of objects) and adding a retweet frequency dimension provided an opportunity to – using a large number of images – trace the objects through patterns of resonance. Such object and retweet patterns of resonance can point us to visual constructions of refugees and otherness that potentially lead to digital flows of affect (Döveling et al., 2018).

Although ‘liking’ is often equated to message endorsement, it does not further disseminate images across the platform; though, it does affect the algorithmic visibility of images. As the algorithmic properties of the Twitter feed are unknown to us, we focused on retweet counts (and not ‘like’ counts) to get a sense of the reach of certain images over others.

Co-occurring images and labels were qualitatively examined for visual themes, narratives, and other connective patterns. This approach partly aligns with the four steps of VNA (Visual Network Analysis) as proffered by Mathias Decuypere (2019): collecting and coding relational data; visualizing network diagrams through software; analysing the form of these diagrams; and interpreting the resulting visualizations by offering narrative readings of these forms. Thus, we emphasized the visual rather than the mathematical properties of networks for our analysis.

Finally, after network analyses of both annotated object labels and images, we examined the values and ideas expressed through resonant tweets in the framework of existing literature on the constructions of refugees; this allowed us to make sense of the visuality constructed by influential actors within the Twitter sphere. Thus, through qualitatively assessing images that share certain objects, we enhance our understanding of the use of certain visual and compositional elements, their connotative meanings, and how these relate to affective potentialities in networked publics.

Findings

We first describe the prominent objective elements as annotated and networked by the Google Cloud Vision API, then proceed by briefly examining visual themes pertaining to each object. Here, ‘object’ is used to denote any labelling derived from the API, meaning object labels can also pertain to machine-detected emotions and activities, such as play and happiness. After the qualitative interpretation of the networks, we will proceed by analysing the images themselves, while taking into account texts with which these images were tweeted. We assess these visual contents using the frameworks of witnessing distant suffering and constructions of otherness.

Overview of visual elements and retweet resonance

Several object labels were identified by the API such as vehicle, soldier, explosion, etc.; however, an in-depth analysis of each image and its affiliated labelling is beyond the scope of this article. Therefore, the results and implications proffered by the present investigation focus solely on photographs and one video featuring Syrian refugees, the selection of which is based on retweet metrics, pointing to greater affective potentiality (Figures 1 and 2).

OPEN IN VIEWER

Figure 1.

The network of images and labels for the English Twitter space; labels connect to annotated images and cluster by similarity; images are sized by retweet count and coloured by modularity class.

OPEN IN VIEWER

Figure 2.

The network of images and labels for the Arabic Twitter space; labels connect to annotated images and cluster by similarity; images are sized by retweet count and coloured by modularity class. This network is slightly cropped for legibility.

Resonating objects in the globalized and localized context

Whereas annotated image object elements were largely similar across both spaces – with significant space for children, aid workers, and aid initiatives in action, as well as bombardments and the destructive results thereof – the visuality of highly retweeted images, pertaining to a larger affective potential and playing out relational affective patterns between interacting agents (users and platform affordances), is different in both contexts (see Figure 3 for an overview of the top five most retweeted (RTd) photographs in both contexts).

OPEN IN VIEWER

Figure 3.

Overview of the top five retweeted images in a globalized, Western-biased construction of Twitter image content and in a more localized construction.

Over half of all images analysed were photographs depicting scenes within refugee camps, whereas only 7 of the 61 images analysed within the English language space featured such depictions. Indeed, everyday livelihoods in camps are more present in a localized context. Overall, the Arabic space is less personalized as refugees are often depicted in groups, whereas the English context was dominated by UNHCR and NowThis News (news outlet) tweeted portraits of refugees who are either posing with a recognizable prop (e.g. a Sesame Street figurine) or artefacts of particular personal importance to those portrayed.

Within the English space, the most resonating image depicted a female Syrian refugee making halloumi cheese, captioned as ‘food-processing’. The image – a still photograph captured from a video clip – is a typical approach (Gilligan and Marley, 2010; Van Dijk 2008) for countering negative stories of refugees through positive narratives, demonstrating how migrants contribute to their host country. It was originally tweeted by UNHCR and entered our dataset through a retweet by @JamesMelville, an active political commentator with 151,000 followers as of 12 March 2020. The featured protagonist in the clip, Razan Alsous, founded the Yorkshire Dama Cheese Company, a successful cheese company in Yorkshire, UK, producing halloumi from local dairy farmers. The tweet reads, ‘After escaping war in Syria, a refugee has used her initiative – and local milk – to make halloumi in northern England, creating jobs and winning awards along the way.’ Here, quite literally, the employment benefits are brought to the fore.

The narrative of a ‘contributing migrant’ is then expanded upon by Razan, who states, ‘Whenever people ask me where I come from, I say: Yorkshire.’ With this explicit testimony of belonging to the host country (leaving out descendance), UNHCR actively and intentionally addresses a Western audience: one perceived by the institution as unwelcoming or intolerant of migrants for diverging from Western cultural values. In this image, Razan – through proclaiming that she is from Yorkshire – is ‘overwriting’ her Syrian descent. Looking at the retweet metric distribution, this first image was retweeted 2,909 times while the second image has a retweet count of 1,206. This is interesting as the latter image was tweeted by an account with a larger followership than the account tweeting the former image (7.8 million as opposed to the 137,000 of @JamesMelville), which is unexpected if we can assume that larger followerships have a greater potentiality of getting retweeted.

There are several similar portraits tweeted by UNICEF during this time period, in which refugees talk about personal artefacts. Through their (micro)narratives, an attempt to (re)humanize the people is part of an intentional humanitarian strategy. One particular case, in which two images were shared by the same account in the same time span, allowed for a more direct connection to be made between image characteristics and resonance. The first tweet, with an image of a girl and her house keys, generated 1,206 retweets. The second tweet carries a quite ‘similar’ UNICEF portrait, but of a young woman, generating 97 retweets. The two main features of the first mentioned image include: (1) the soccer club logo on the key chain; and (2) her age, marking her as the youngest subject of the two photos discussed. From a close reading of the Twitter comments, the logo of AC Milan has led supporters to encourage the club to support her by inviting her to games; one of the comments reads, ‘You want to begin this new Chapter of success on the right foot? Do the right thing @acmilan and help her out!’

A second visual characteristic that might account for a greater relational affective pattern in the retweet response is the gender and age of the young child. Depicting women rather than men, and children rather than adults, is a common humanitarian trope of depicting the pain of others in ‘older media settings’ (Sontag, 2003; Zarzycka, 2012), and seems similarly applicable for images circulating on Twitter. This trope also cuts across language barriers; in both spaces we see more children than adults, although we see a balanced depiction of gender representation in both contexts.

Discussion and conclusion

Seeking to expand upon scholarly understandings of human migrancy in the digital age, the present study utilized a multi-methodological approach to investigate the visual networking processes through which users’ perceptions of social and political struggle take shape. We encountered different regimes of visibility in the two language contexts; there were similarities in image content, but the resonance and relational affective patterns that images generated differed significantly. While visual content elements of specific otherness – including livelihoods and geopolitical traits – were more prominent in the Arabic language space, their textual contextualization was similar to Liisa Malkki’s (1996) ‘speechless emissaries’ and Sontag’s (2003) notion of reducing the powerless to their powerlessness by excluding the names of depicted people in captions. Moreover, these images – fitting a regime of visibility as biological life (Chouliaraki and Stolic, 2017) – seemingly reduce refugees to groups of unfortunates living under dire or indigent circumstances, fully dependent on emergency (Western) aid. Like this, such visuals construct the refugee crisis as a humanitarian emergency and exclude critical reflections on failures of politics. Simultaneously, the English language sphere visually taps into the affective power of portraits. Although (literally) zooming in on the sufferer, portraits obscure geopolitical otherness, thereby allowing space for spectators to be self-reflexive and closing the space for voices that are unlike those of the audience.

In sum, what spurs the audiences in both language spaces are images belonging to different regimes of visibility as constructed by Chouliaraki and Stolic (2017); one fitting the biological life regime and the other fitting the self-reflexive regime. Under the affective affordances of the platform, visibility is likely to be positively affected by sharing behavior of Twitter users. As networked publics cannot be affected by images that do not show up in their feeds, networked publics ‘feed themselves’ what is shareable, resulting in a ‘vicious circle’ of amplifying two predominant constructions of refugees. These include refugees: (1) with whom we can personally identify; or (2) depicted as masses of unfortunates that we can help, but for whom we hold no further responsibilities in addressing or reversing political failure.

Regarding the earlier outlined power asymmetry enshrined in Twitter’s architecture, we encounter an interesting dynamic when comparing the two language spaces. Such an asymmetry seemed less apparent in the Arabic space, where individuals with relatively small personal publics were able to reach the top ten in retweet ranking, finding themselves among users with much higher follower counts: the top-ranked image came from a user with 2.5 million followers, whereas the third came from a user with 1,070 followers. This is most likely the result of a smaller total number of users within this dataset; in smaller networked spaces, the likelihood that of being heard above the fray is higher, which increases chances of ‘your tweet being picked up’ or going viral.

Humanitarian hashtags and the bundling of causes for forced migration

The workings of digital affordances, such as hashtags, also have a profound effect on how people are confronted with forced migrancy. Although tweets pertained largely to the Syrian cause for forced migration (i.e. based on initial targeted queries; see Appendix A), we also identified images of refugees within the dataset depicting the consequences of other causes for forced migration. In an effort to gain as much traction as possible, users include hashtags that are not (directly) related to the contents of the tweet, intensifying occurrences of entanglement with different issues in one stream of tweets. The danger lurking in this ‘lumping together of different cases and causes for tragedy’ aligns with Sontag’s (2003: 70) critique of Salgado’s images:

through grouping together a host of different causes, Salgado makes suffering loom larger . . . globalizing it . . . may spur people to feel they ought to ‘care’ more. It also invites them to feel that the sufferings and misfortunes are too vast, too irrevocable, too epic to be much changed by any local political intervention.

In our study, this lumping together of causes is a result of the hashtag #worldrefugeeday which dominated the time span for the duration of this study, and allowed us to assess how humanitarian tags combine diverging causes of forced migration.

It must be noted that there is a distinct difference in the ways in which people ‘consume’ these streams of images, as opposed to reading these images within other media contexts; this difference is difficult to tease apart. Due to the fact that feed algorithms remain largely obscure to academics, scholars cannot easily assess how users consume this hashtagged stream of ‘grouped misfortunes’ under the tag #worldrefugeeday. If the tag is searched, one encounters a more complete overview of the tagged tweets. But apart from this deliberate act of searching via hashtag, it is unknown how many of such ‘grouping’ tweets individual users are confronted with.

Limitations and future directions

As with any attempt to automate the analysis of visual data, the present study was not without its limitations. Google Cloud Vision API can provide inaccurate or inconsistent labelling of visual objects. For instance, the Arabic space contained a similar amount of ‘people’ labels (18, compared to 14 in the English space); however, people are shown in collectives, resulting in the annotation of background features, at times ignoring ‘people’ altogether. While this seems to be a limitation of the API, when paired with the qualitative assessment of similar images within network label communities, it provides a glimpse into the differences of how people are visually contextualized by the composition of images.

Another limitation of the automated analysis of visual data via Google Cloud Vision API is the program’s inability to separate and analyse multiple images within a single string or line of data (i.e. some tweets contained multiple image links, but the API was only able to analyse the first image within the sequence). Thus, future research should employ automated methods for separating this content prior to analysis.

The final limitation worth mentioning is the inherent difficulty in distinguishing the underlying impetus behind retweeting on Twitter. Our review of the literature demonstrates the unique architectural and affective affordances of retweeting as a mechanism through which users actively disseminate or endorse content. However, it remains unknown exactly why – beyond theoretical explanations – Twitter users engage in these practices; that is, because we focus upon aggregate data and lack the depth of insight from interviews or observations, our contribution here is inherently theoretical. In the present study, we attribute differences in retweeting activity between disparate language spaces to underlying socio-cultural perspectives. That is, the socio-technical interplay between Twitter’s affective affordances and platform users produces regimes of visibility that rely on the affective ‘shareability’ of images. Different language spheres, then, clearly pertain to different regimes of visibility; both fail to draw attention to the actual underlying (political) causes of refugees’ circumstances.