Deep state phobia: Narrative convergence in coronavirus conspiracism on Instagram

Data collection

In order to validate our hypotheses, we analyze a dataset of 478,154 Instagram posts collected by means of the Instaloader Python library (Instaloader, 2021). These data were collected on the basis of a seed of 82 primarily English language hashtags related to conspiracies prevalent at the time of the coronavirus pandemic determined by a team of domain experts writing a book on the topic of COVID-19 conspiracy narratives (Figure 1), and from the accounts of 66 accounts that frequently posted using similar hashtags—many with followers counts in the tens and hundreds of thousands, even one in the millions. In compiling this seed list, our aim was to capture a representative sample of online conspiracist responses to COVID-19 on Instagram—primarily, though not exclusively, focused on Anglo-American content. Since most of the seedlist hashtags are recent vernacular innovations, one might assume that they would return a quite narrow sample, when used to query Instagram’s API. Significantly, we find that the posts in our dataset commonly used a dozen or more hashtags each. As an example of this practice of ‘hashtag stuffing,’ the following list of hashtags comes from a single post, picked more or less at random from amongst the top most engaged with content in the dataset, in this case a cartoon image of Bill Gates: #agenda2030, #control, #populationcontrol, #depopulation, #freemarket, #agenda30, #truth, #corruption, #agenda21, #nwo, #rothchild [sic], #rockafella, #newworldorder, #corona, #coronavirus, #covid, #covid_19, #covid19, #event201, #wedonotconsent, #vaccines, #antivaxx, #alternativemedicine, #antivaccine, #virus, #viruses. Given the number of posts that we collected, our starting points were thus overshadowed by the plentitude of connections contained in each post. That said, assembling an exhaustive dataset of every conspiratorial post for the period under investigation was not the objective, not least because of the inexactness of the very concept of conspiracy theory. Given the platform’s vast size and global reach, our dataset is only an entrypoint into the multiple worlds of conspiracism on Instagram.OPEN IN VIEWER

For each post, we retained the hashtags and the full text of the post description. We ran the script at four points across 2020: in mid-May, at the end of June, at the end of August, and at the end of October. By concatenating these data we created a single corpus of 707,203 posts. This resulted in a historically unique archive, since Instagram is known to have removed a large numbers of accounts deemed as conspiracist in late 2020. As our analysis focuses on Q1 through Q3, this ‘deplatforming’ is not visible in our data collection. However, upon verification, we found a substantial number of the posts discussed below to have been permanently deleted—as well as a third of the accounts—in what amounts to a radical erasure of the archival record for this controversial period on the platform. Against this context, we offer the analysis below as evidence of how, during the height of the pandemic, Instagram effectively functioned as a medium for the growth and spread of conspiracist narratives.

While social media platforms had already come under fire for amplifying extremist conspiracy theories prior to the pandemic (Phillips, 2018), our study appears to corroborate the claim that the pandemic acted as a catalyst for an even greater mainstreaming of conspiracist narratives (Stolper, 2020). Importantly, it should be noted that on October 4th of 2020, amidst attempts to manage the spread of misinformation on the platform, Instagram adjusted their API making it much more challenging for social media researchers to collect data from the platform. As relatively fewer posts were consequently scraped from Instagram for Q4 of 2020, data from this quarter were therefore mostly omitted from the analysis. For the purposes of the present investigation, we thus focus on a set of 478,154 posts from the first three quarters of 2020, with Q1 (January–March) marking the period leading up to the global coronavirus pandemic, Q2 (April–June) marking the actual outbreak of the pandemic, and Q3 (July–September) the first months into the pandemic. In addition to posts, we also used Instaloader to collect a dataset of 31,752 associated images.

As illustrated in Figure 2, the content retrieved from our seed query of hashtags and accounts dates back as far as 2011. Though a segment of the hashtag narratives retrieved thus predate the pandemic, we focus our efforts mainly on the year 2020. During this time, the pandemic evolved from a local crisis in Wuhan to into a global public health emergency. Although Instagram is more typically associated with images and identity rather than narratives and politics, the platform nonetheless came to be described as a conspiracy theory “hotbed” during the pandemic (Bell, 2020: np). Journalists and scholars noted for instance that, between January and August 2020, interactions and video views generated by QAnon-related Instagram accounts significantly outperformed other trending hashtags such as “MeToo” or “BlackLivesMatter” (Bloom and Moskalenko, 2021: 44). One possible explanation for QAnon’s success on Instagram is that the platform’s image sharing affordances spoke more directly to the emotions and identities than galvanized the QAnon narrative at this stage of its development (Bloom and Moskalenko, 2021: 71). Another factor to consider is the dynamics of social media ‘influence’ in the circulation and amplification of disinformation and conspiracy theories online (Phillips, 2018), where Instagram has been and arguably remains the dominant platform for studying an ‘attention economy’ of influence, in which status is based on page views and clicks (Marwick, 2015).OPEN IN VIEWER

Specifically, over the course of 2020, researchers observed the emergence of a female demographic of Instagram accounts who were drawn into QAnon through a narrative that imagined wealthy establishment ‘elites’ as secretly engaged in ‘ritual child abuse,’ which connected to an early pro-Trump conspiracy theory known as ‘Pizzagate’ (Robb, 2017; Argentino, 2021). Although these narratives were false, they frequently contained the kernel of a genuinely disturbing reality, as, for example, in the outrage over the case of Jeffrey Epstein, the American financier with ties to establishment figures—including the Clintons—who for years evaded justice for his serial sexual abuse, only to die in prison under seemingly mysterious circumstances (Cassidy, 2019). As this study shows, these and many other (often disturbing) narratives forged connections, which can be detected across multiple scales, from the image content of the posts and their textual descriptions, to the hashtags that are used to literally connect posts into larger conversations.

Hashtag analysis

On Instagram, one of the more explicit ways in which users can express their support for a certain narrative is through hashtags. Such new media affordances as the hashtag have been theorized as the basis of a profound historical transformation toward more “flexible” and “personalized” forms of political engagement (Bennett and Segerberg, 2012: 744). Indeed, digital methods research on Instagram treats such hashtags as markers of distinct political discussions or of an “issue space” (Rogers, 2021: 6). In the context of presumedly conspiratorial discourse, it is be fruitful to consider hashtags constitutive role in broader ‘political narratives’ (Patterson and Monroe, 1998), in this case compelling stories about the way that the world actually works, set against a broader context of extreme polarization and even existential uncertainty. In looking for empirical traces of convergence between such conspiracist narratives, we thus start off at digital hermeneutics’ ‘mass opinion’ scale by seeking to detect patterns connecting across all hashtags in the corpus. Our relevant hypothesis here is that when two otherwise distinct hashtags appear together in a single post, this may indicate a confluence of narratives.

General clusters and dynamics

Below, we graph the hashtag co-occurrences in the data as representation of the growth and convergence of conspiracist narratives over the course of the first three quarters of 2020 (Figure 3). In these graphs, hashtags are represented as nodes, and an edge is drawn between two nodes when the corresponding hashtags co-occur in the same post. Following the digital hermeneutics approach of combining data analysis with theorization, we then proceeded to use domain expertise to manually assign nodes to overarching conspiracist narratives (or metanarratives). Because Instagram users often combine multiple hashtags in their post descriptions (a phenomenon that can be referred to as ‘hashtag stuffing’), our initial seed of hashtags yielded a set of additional, co-occurring hashtags. While many of these were not ‘conspiratorial’ per se, they were manually clustered based on topical similarity in an effort to identify traces of convergence. In the resulting graphs in Figure 3, we thus plot the co-occurrence networks of the hashtags for the first three quarters of 2020, color-coded into seven ‘metanarratives’: Trump/QAnon (blue), Pizzagate (purple), Conspirituality (orange), NWO/New World Order (yellow), Covid (green), Bill Gates (red), and 5G (brown). By this method, we can thus see metanarratives appear to grow and overlap as new hashtags appear and shared hashtags draw them together. Finally, we visually compared the graphs for the first three quarters of our dataset in order to examine the potential convergence between narratives over time.OPEN IN VIEWER

The co-occurrence networks of the hashtags across the first three quarters of 2020 provide an overview of the diversity and scope of conspiracist narratives on Instagram. For one thing, we can observe a large number of very popular Trump-related hashtags in the networks. The top ten most used hashtags in our dataset over the first three quarters of 2020 are: “wwg1wga,” “qanon,” “thegreatawakening,” “obamagate,” “pizzagate,” “deepstate,” “newworldorder,” “q,” “billgates,” “maga.” What we find is that all of these hashtags except for “bill gates” and possibly “newworldorder” are phrases directly connected with Trump or his followers. The blue cluster at the bottom of the graph is made up of hashtags related to the partisan slogans of the antagonistic populist political style associated with Trumpism (e.g., “Trump2020,” “MAGA”). Interestingly however, the most commonly used hashtags at the center of this cluster are related to QAnon (e.g., “thegreatawakening,” “wwg1wga”). We can surmise from this observation that, during the 2020 U.S. presidential election, QAnon became central to Trumpism. While QAnon did to some extent internationalize in the course of 2020 (Gallagher et al. 2020; Callison W and Slobodian), the hashtags that make up this large blue cluster nevertheless give a sense of overall the prominence of right-wing American politics in this dataset, and conversely the presence of conspiracism in right-wing American politics at this point in time. Furthermore, this cluster is marked by its parochialism, focusing narrowly on Trump and QAnon, while other clusters can be seen to overlap with other narratives. That this cluster appears to remain comparatively stable across the three quarters, suggests that QAnon was also already well established on Instagram prior to the pandemic. We may interpret this as empirical proof of Instagram’s importance in developing and amplifying the QAnon movement—evidence which the platform has since set out to erase from the historical record.

We also coded a purple cluster made up of hashtags related to the child trafficking and pedophilia narrative (e.g., “savethechildren,” “pedogate,” “adrenochrome”), which appears above as a subset of the aforementioned blue cluster. As mentioned earlier, prior research into conspiracism on Instagram proposed that this narrative exploded on the platform in the fall of 2020. We call this the Pizzagate cluster, in reference to the conspiracy theory that predated QAnon and which initially developed much of its central narrative of ritual child abuse, via the coordinated efforts of ‘researchers’ on 4chan at the time of the previous US presidential election (Tuters et al., 2018). These hashtags do indeed appear as a more dynamic network within the larger QAnon movement, although the growth in Q2 occurs earlier than has previously been reported, which suggest that the narrative had quite some time to consolidate relatively unobserved. Finally, with regards to QAnon, we see that COVID-19–related hashtags, both neutral and conspiratorial (“lockdown,” “plandemic,” “filmyourhospital”) occur primarily in the top half of the graph, relatively isolated from both Pizzagate and QAnon-related hashtags. This is a remarkable finding as it contrasts with claims of earlier research that the use of hashtags related to Covid mitigation measures had energized the QAnon movement (Bloom and Moskalenko, 2021: 149–174). Taken together, these two clusters suggest that the QAnon movement was energized by Trump as well as by a resuscitated version of the earlier narrative of Pizzagate.

If the hashtags in the bottom half of the hashtag graph may all be understood as essentially relating to the broader QAnon narrative, then those hashtags at the top of the graph might point toward a relatively distinct conspiracist narrative. Earlier reporting on QAnon’s presence on Instagram made much of its overlaps with supposedly more ‘feminine’ New Age concerns—a convergence that had already been termed ‘conspirituality’ some years earlier (Ward and Voas, 2011). Having used this term to label hashtags in the orange cluster (“mindcontrol,” “questioneverything,” “awakening”), in contrast to the prior research we find little overlap with QAnon. Instead, we find overlaps with the green Covid cluster and the yellow NWO/New World Order cluster, the latter related to conspiratorial theories concerning global governance (“newworldorder,” “illuminati,” “event201”). Additionally, at the top of the graph is a brown 5G cluster (“stop5g,” “5gkills,” “radiation”), which gradually becomes engulfed by the Covid cluster. This supports the interpretation that pre-existing concerns around 5G became connected with COVID-19 in 2020 (Bruns et al., 2020). Last, we have the red Bill Gates cluster (“fuckbillgates,” “billgatesisevil,” “saynotobillgates”), which unlike any of the others remarkably appears out of nowhere in Q2. The dramatic emergence of Bill Gates in our dataset seems to provide initial evidence for our hypothesis concerning the role of shared antagonists in the process of narrative convergence.

By Q3, all of the clusters in the top of the graph—in distinction to those related to QAnon at the bottom—appear to have become more integrated with one another, with hashtags for ‘the New World Order’ forming their core. Although the term ‘New World Order’ has long been used by anti-semitic conspiracy theorists (see, for instance, Goodrick-Clarke, 2003: 279–302), arguably, we may refer to this apparent convergence of narratives as ‘The Great Reset’ metanarrative. Originally taken from the name of the World Economic Forum’s 50th annual meeting (Schwab and Malleret, 2020), as a conspiracist metanarrative, the Great Reset imagines the coronavirus pandemic as part of a secret plan to impose global governance on the part of “the globalists, the world economic elite” (Mercola and Cummins, 2021: 103). Involving speculations about IT (including “rfidchip” and “5G”) as technologies of “populationcontrol,” the metanarrative also frames “vaccination” as a corporate strategy to replace individual rights with a mass surveillance regime—orchestrated by the “illuminati.” The Great Reset thus appears to make a conspiracy theory by combining bits and pieces of long established right-wing conspiracy narratives with leftist critiques of ‘globalist’ corporations conspiring to advance a neoliberal economic agenda (Slobodian, 2018)—a point to which we will return in the discussion below.

In general, our analysis of hashtag co-occurrences suggests an increased connectivity between pre-existing conspiracist narratives around the time of the outbreak of the pandemic in Q2. Significantly, our graph for Q2 reveals a dramatic multiplication of covid-related hashtags, which seems mirrored by the sudden appearance of predominantly antagonistic Bill Gates–related hashtags. While antagonistic discourse about Bill Gates likewise predates the dataset under investigation, this finding supports the claim of previous researchers concerning “the emergence of ‘Bill Gates’ as a key actor” with the coronavirus outbreak (Shahsavari et al., 2020: 298). Indeed, our subsequent findings situate Bill Gates at the intersection of a broad range of conspiracist narratives, for which we consider some possible explanations below.

QAnon and the Great Reset

When comparing QAnon and the Great Reset as relatively discrete conspiracist narratives we see a number of differences. There is a tonal difference here between the fantasy element that underlies many of the QAnon hashtags. More stable over time, the QAnon narrative has featured more of its own hashtags which function as rabbit holes into a broader interconnected narrative. Often based on references to fantasy literature (“followthewhiterabbit,” “adrenochrome”) or film (“redpill,” “wwg1wga”) these hashtags have the quality of seeming to be part of a scripted narrative, as many have observed of QAnon more generally (Thompson, 2020). By contrast, on Instagram the Great Reset appears to be anchored by the long pre-existing conspiracist narratives about global governance (“NWO,” “illuminati,” “markofthebeast”), which in turn seem to speak to a more diverse set of concerns centering on COVID-19 (“plandemic,” “covidhoax”), technology (“5G,” “rfid”) and what we might call technocracy (“event201,” “id2020,” “agenda21”).

Figure 4 also illustrates how the narrative of the Great Reset gains consistency as formerly relatively distinct narratives overlap. Anchored by the yellow NWO (New World Order) cluster, by Q3 the Great Reset includes anti-vax hashtags, anti-5G hashtags, and anti–Bill Gates hashtags. At its core, the Great Reset is a conspiracist interpretation of actually existing plans for global governance—for example, “id2020” refers to an NGO advocating digital identity for undocumented people, “agenda21” refers to a UN resolution for sustainable growth and “event201” refers to a pre–COVID-19 pandemic exercise funded by the Gates Foundation. As with the term ‘the Great Reset’ itself, what we find then is a conspiratorial reading of existing governance initiatives. In some cases, we can literally see this convergence figured in the Instagram image posts themselves, for example, in Figure 5, which presents a triangle of Big Tech, Big Science and Big Pharma, respectively, connected with “total surveillance,” “economic devastation,” and “neurological damage,” within which is a Venn diagram of 5G, the coronavirus and vaccines at the overlapping center of all three is “depopulation.” In the comments on the post we can see the prominence of both Christian religious and New Age interpretations with reference to both “lucifer” and “the love inside you.”OPEN IN VIEWER

OPEN IN VIEWER

Figure 5.

Narrative convergence in Instagram imagery. Image by Anonymous.

The patterns of hashtag co-occurrences described in this section seem to present the Great Reset as a more dynamic and multi-issue metanarrative than QAnon/Pizzagate, which furthermore appears surprisingly insulated from the pandemic-related concerns. In contrast to the largely right-wing American QAnon, the Great Reset seems both more international as well as more ideologically diverse. There is however one significant overlap, the “deepstate” hashtag in Figure 6, which actually moves from the core of the QAnon cluster to become a bridging node between the two metanarratives. The concept of the Deep State, to which it refers, first emerged in Turkey, a country with a long history of coups, as a term for a parallel government of bureaucrats and military officials that exert influence on the elected government later to be used in the American context by the noted conspiracy theorist and Trump supporter Alex Jones (Den Bulk and Hyzen, 2019; Gingeras, 2010). In fact, noted scholars have even gone so far as to claim that, while “continuities persist [with] similar theories of the past […] the Deep State narrative is unique to the Trump era” (Phillips and Milner, 2021: 12). Whether or not this claim is historically correct, it certainly speaks to the centrality of the “deepstate” hashtag within our dataset—and perhaps to the concept of the Deep State more generally, in the post-Trump era. In order to triangulate these findings, however, we need to move beyond the level of hashtags. Following the multi-layered ‘digital hermeneutics’ approach proposed in Romele et al. (2020), we therefore also consider the actual contents of the Instagram posts, that is, post descriptions and images.OPEN IN VIEWER

Text analysis

Typically, ‘digital methods’ work on the level of structured objects, such as hashtags or hyperlinks (Rogers, 2013). As was already explained earlier, mapping the co-occurrences of conspiratorial hashtags on Instagram leads to high-level insights about the convergence between metanarratives. However, we argue that in addition to hashtags or images, the study of online conspiracy theories can also gain a lot from examining the actual full text of an Instagram post, that is, the written description that accompanies an image or video. As mentioned above, this is the scale of analysis referred to as ‘latent opinion’ (Romele et al., 2020: 80). For the purposes of our ‘narratological’ analysis of texts on Instagram, we operationalize the NLP technique of named entity recognition (NER) to computationally identify co-occurring names of persons and organizations mentioned in post descriptions. Conceptually, this method recalls structuralist approaches to story analysis, in which the relations among named entities filling in specific roles or ‘actants,’ make up the core structure of a narrative. As was recently demonstrated by Tangherlini et al. (2020), networks plotting the interaction between such entities can yield insight into the deeper structures and dynamics of conspiratorial narratives and conspiracies. Building on our earlier observations, our hypothesis here is that in conspiracist Instagram posts, certain actors will take up more central positions in the networks, either as protagonists (e.g., Donald Trump) or as antagonists (e.g., Bill Gates) of distinct narratives.

In order to plot the co-occurrence networks of the key actors in our dataset, we first extracted named entities from the Instagram data using the spaCy natural language processing library in Python (spaCy, 2021). We then proceeded to clean up the retrieved entities, removing any noise resulting from the model, and manually classified the most frequently occurring persons and organizations into categories such as government officials, politicians, or Hollywood actors using a bottom-up coding scheme (DMI, 2021). Finally, we represented the co-occurrences of entities in the 2020 dataset as a graph using Gephi. Here we assume that if two or more entities are mentioned in a single post, a meaningful relation between these entities exists, and a corresponding edge can be drawn between both entity nodes. Correspondingly, an actor’s shifting position in the overall networks over time might provide meaningful insights into that actor’s role in bringing diverging conspiracist narratives together.

The graphs in Figure 7 show aggregated overviews of the persons and organizations mentioned in the textual descriptions of our full dataset of conspiracy-related Instagram posts (2011–2020). While seeming to represent a broad-ranging conversation that goes well beyond the starting point of our hashtags, we can find a general pattern of fear of the ‘Deep State,’ which manifests as antagonism toward multinational institutions and ‘globalist’ individuals. When we examine the most frequently mentioned organizations, we find that most of them are state and governmental entities, (e.g., “senate,” “state,” “pentagon”), big tech companies (e.g., “Amazon,” “Google,” “Microsoft,” and “Facebook”), big pharma companies (“Big Pharma,” “Johnson and Johnson,” “[Bill and] Melinda Gates Foundation”), or media organizations. Among the most frequently mentioned persons, “Trump” and “Bill Gates” stand out. Additionally we find names of perceived opponents of Trump (“Hilary,” “Obama,” “Joe Biden”) as well as figures associated with conspiracy theory either as authors and proponents (“Judy Moskowitz,” “David Icke,” “David K Miller”) or as key antagonists (“Tom Hanks,” “Oprah,” “George Soros,” “Anthony Fauci”).OPEN IN VIEWER

The co-occurrence networks of named entities over the year 2020 (Figure 8) display dynamics of convergence that reflect some of the patterns that we identified in our hashtag analysis. One salient observation here is the centrality of the figure of Donald Trump, who appears to be discussed alongside religious entities (“god,” “jesus”). As the year progresses, we also see Trump’s perceived Deep State opponents move from the periphery increasingly toward the center (e.g., “george soros,” “epstein,” “clinton,” “rothschilds,” “rockefeller,” and “bill gates”). Here the ‘us vs them’ antagonism of right-wing populism of conspiracism and of anti-semitism share a similar “narrative as felt,” or “deep story” (Hochschild, 2016: ix). We may interpret this as additional evidence that while Deep State narratives existed before the pandemic, there was a confluence in 2020 as entities became more interconnected. This is most clearly visible around the time of the initial outbreak of the pandemic (between Q1 and Q2), but what is also notable is that this interconnection remains in place as it seems to become part of the U.S. election narrative.OPEN IN VIEWER

Entities and celebrities

Some images in our dataset also contain striking examples of narrative convergence. The picture in Figure 12 for instance brings together otherwise relatively distinct concerns about 5G and vaccines with New World Order and chemtrails conspiracy theories. To help us identify such traces of convergence in our dataset of 31,752 images, we used methods from the field of computer vision offered through the Google Vision API (Vision AI, n.d.). Specifically, we used the entity extraction endpoint to computationally extract a list of all ‘entities’ (such as objects or persons) that figure in images, as well as the ‘celebrity detection’ endpoint to detect faces of well-known persons. Similar to the aforementioned methods, we were particularly interested in detecting co-occurring actors or entities in the dataset’s imagery. Additionally, subsets with images of interest (such as those representing some lead antagonists) were identified with the help of the PixPlot image clustering tool (PixPlot, 2021), and then further examined in a more qualitative manner. This technique for ‘reading’ Instagram posts at scale helped us to visually explore how conspiracy appeals to emotions (e.g., through colors, recurring faces, etc.). While this pattern of emotional appeal has been remarked upon in the literature (Van Prooijen and Douglas, 2018), there is a relative paucity of empirical evidence for how this works through the grammar of images.

Figure 9 shows a bipartite network representation of our image dataset for Q2 of 2020, where nodes represent images and edges represent co-occurring entities. This visualization supports many of the observations that were previously made on the basis of hashtags and post descriptions. On the left-hand side of the image we see a Bill Gates cluster with edges connecting to images containing the word “microsoft” and “vaccines,” as well as to a somewhat lesser degree with edges connecting to images containing the words “george soros,” “fauci” and “epstein.” Once again, but now at a different layer of analysis, we see that certain named entities function as a kind of “boundary object” that is shared by different communities, and that is “plastic enough to adapt to local needs and constraints of the several parties employing them, yet robust enough to maintain a common identity across sites” (Star and Griesemer, 1989: 393). Although Bill Gates forms the largest cluster, the most central boundary object in this network is “epstein,” in reference to the aforementioned Jeffrey Epstein. Connecting Bill Gates with Clinton and QAnon, the centrality of Epstein is consistent with the observed emergence of the Pizzagate narrative around Q2 in the above hashtag analysis. Indeed, Epstein featured as one of the central antagonists in this Pizzagate narrative. Also consistent with the latter, Epstein connects into a network of Democratic politicians and organizations seen as connected to the Deep State—indeed, the named entity “state” is amongst the largest clusters in the network. ¹ OPEN IN VIEWER

Moving on from quantitative analysis of the text contained within the images to the images themselves, celebrity face detection reveals that Bill Gates is the single most commonly occurring celebrity in the image sample. This supports the previous findings concerning Gates’s role as leading antagonist in conspiracist narratives (Figure 10). Unsurprisingly, given the prominence of Trumpism in the above hashtag analysis, the other main figure that comes to the fore is Donald Trump, the protagonist of the QAnon narrative. Among the other frequently detected faces, we primarily find figures from American politics who fit into a Manichean narrative, either as members of the globalist elite, like Bill Gates, or else as Trump allies. While it is clear that American politics, particularly of a right-wing valence, are highly represented in this dataset, the face detection technique also picks out major celebrities.OPEN IN VIEWER

Bill Gates

Out of all three of our analyses (hashtags, texts and images), Bill Gates emerges as the most significant single actor in the dataset in general and in the process of convergence in specific. Cast in the role of a shared antagonist, Gates connects a diverse range of conspiracist narratives, and his infamy only grows over time. Among the most reposted images in the dataset with the hashtag “billgates,” Figure 11 is illustrative of this common pattern of convergence: it features Gates at the intersection of various conspiracist narratives. The image is presented in the format of an ‘image macro’ meme template, which includes a setup and punchline at the top and bottom taken from American popular culture (Nissenbaum and Shifman, 2018). In this case the top and bottom of the image feature an avuncular character known as ‘the Stranger’ from the cult film The Big Lebowski, whose role is to comment on the absurdity of the narrative and to offer sage advice. In this case, the Stranger meme template is used to seemingly comment on the real-life movie of Bill Gates’s career as interpreted through the lens of conspiracism. The middle of the image presents five seemingly unconnected claims based on Gates’s post-Microsoft investment career—big tech, big pharma, big agriculture, geo-engineering, and contemporary art—that when taken together make Gates appear as a villain.OPEN IN VIEWER

Each of the claims in this image can be traced back to a hashtag or set of hashtags from the former layer of analysis. This again supports the idea of Gates as a focal point of narrative convergence. Gates’s investment in the biotech company Monsanto implies a plot to reduce the world’s food supply and thus diminish its population (as exemplified by hashtags such as “depopulation,” “populationcontrol,” “depopulationagenda”). An image of a smiling Gates with a syringe and a W.H.O. shirt connects him to anti-vaccination narratives that are highly prominent in the dataset. Gates is furthermore connected with the narrative of Satanic pedophiles (“spiritcooking,” “thesepeoplearesick,” “podestaemails”), through his having “sponsored Marina Abromovic,” whose artwork was at the center of the Pizzagate narrative (Tuters et al., 2018) as well as for his hubristic “plans to block out the sun” (“geoengineeredsky”) and for having “advocated microchipping” (“microchip,” “rfidchip,” “billgatesrfid”). While this is one example from a large collection of similar images, this meme shows how Instagram works as a visual networked storytelling medium, and how that affordance feeds into often bizarre conspiracist narratives—for example, about vaccine microchipping, which while erroneous are often based on kernels of truth (cf EUvsDISINFO, 2021).