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.
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.”
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.
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”).
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.