#NotDying4Wallstreet: A Discourse Analysis on Health vs. Economy during COVID-19

Abstract

This paper combines political/poststructuralist discourse theory with actor–network theory to explore dystopian visions in the context of a discourse around the hashtag #NotDying4Wallstreet. The call for protest against former US president Donald Trump’s demand to reopen the economy during lockdown dominates the discourse. The tweets were analyzed with quantitative discourse analysis and network analysis to identify key terms and meaning clusters leading to two main conclusions. The first (A) is an imaginary dystopic future with an accelerated neoliberal order. Human lives, especially elderly people, are sacrificed for a well-functioning economy in this threat scenario. The second (B) includes the motive of protest and the potential of the people’s demands to unite and rally against this threat. Due to the revelation of populist features, this (online) social movement seems to be populist without a leader figure. The empirical study is used to propose a research approach toward a mixed-methods design based on a methodological discussion and the enhancement of PDT with ANT. Thus, the article has a double aim: an update of contemporary approaches to social media analysis in discourse studies and its empirical demonstration with a study.

1. Introduction

During the SARS-CoV-2 pandemic’s containment measures, Twitter and other social media platforms became a critical “go-to news source” [1] (p. 6). The wide range, increasing usage, and shift of debates into the virtual space during the pandemic indicate that Twitter became a virtual room for social interaction and political opinion-forming [2]. This increased social media usage, paired with restrictions, and the fear of worldwide economic and financial turmoil led to heated discussions about the nexus of health and economy. Particularly, one debate in the United States in March 2020 during the first lockdown gained momentum after a statement by former US president Donald J. Trump. He claimed “that an economic crisis might result in more deaths, through suicide, than a global pandemic” [3]. Based on this assumption, he demanded a re-opening of the economy, although the numbers of infections and deaths related to COVID-19 were rising. A second claim by Texas Lieutenant Governor Dan Patrick that elderly citizens would be willing to self-sacrifice for the sake of saving the US economy on TV sparked more public outrage, which then led to the invention of the hashtag #NotDying4Wallstreet.

From a discourse theoretical point of view, this hashtag and public debate not only address the COVID-19 restrictions but also offer an interesting perspective on the nexus between health and economic wealth. The tweets include different positions and harsh critiques of established social practices in the context of work, social justice, healthcare, and state involvement in so-called crises. As illustrated in the two tweets in Figure 1, the debate also shows a new layer of trauma after the losses and restrictions in the context of the pandemic. The discussion about an imminent reopening of the US economy adds an economic layer to the destruction related to COVID-19 as well as an emotional layer in the case of a reopening, which is viewed as a prioritization of the economy over human lives.

How are the demands in the context of the online protests around #NotDying4Wallstreet articulated, and which social disruptions are represented in the discursive formation? Based on this research question, the paper explores the emotional level of trauma expressed by the threat of dying, the socioeconomic destruction tied to COVID-19, and the lockdown policies at the beginning of the pandemic. Tweets with the hashtag #NotDying4Wallstreet will be embedded in a mixed-methods design including qualitative and quantitative discourse analysis as well as network analysis. Due to theoretical and methodical inconsistencies in the current literature concerning tweet analyses, this paper also offers an approach to how to best analyze tweets, which will be elaborated on in part 2. Regarding the content-related aspects, the online discourse shows signs of populist characteristics, such as the unification of a group that views itself as the people who are in an antagonistic relation with the establishment constructed as evil. In part 3, I will introduce the notion of difference as developed in poststructuralism, particularly PDT¹, as a starting point of the theoretical framework. Drawing on, mostly, the works of late Argentinian theorist Ernesto Laclau as the base of this brand of poststructuralism, I will also give a brief introduction to his understanding of populism and contrast this with leader- and agency-focused populism(s). The overview of the theoretical vocabulary will also help to better understand the processes of hegemony and social change. Especially regarding the notion of crisis, this paper follows an approach that differs from the agency-focused, traditional definition of crisis in International Relations as “an unexpected event that has to be dealt with, managed” within International Relations [6] (p. 26). I draw on Dirk Nabers’ definition of crisis as constitutive and closely related to social/discursive change [7,8,9], which already has been used in other discourse theoretical works [10,11,12]. Since social media acts as a “place where power becomes decentralized, and the supremacy of the state and dominant institutions are challenged” [13] (p. 4), PDT as a theoretical frame is highly advisable to explore these struggles. This enhanced discourse analysis will help to trace identity formation, structures of self and other, frontiers of antagonism, and disruptions based on the critique and demands expressed by the analyzed tweets with this hashtag. The tweeting people unified by #NotDying4Wallstreet symbolize resistance against former US president Trump’s call to return to normalcy by going back to work during the lockdown. This particular aspect is an example of a leader’s failed attempt “to engage in crisis performance in order to advance their own political positions” [14] (p. 153).

Eventually, this paper is a proposal for an enhanced methodical approach to social media data in discourse analysis and the illustration of a crisis as a constitutive part of change in the context of the health/wealth nexus during the pandemic. Since “[p]opulism gains its momentum from the notion of crisis,” this case study with data from the beginning of 2020 not only offers new empirical perspectives on social (online) movements that are populist without a leader but also is aimed at developing theory and method one step further.

2. Approaches to Social Media Data

The main issue with social media data in discourse analysis is that “the theoretical and methodical implications of digital and social media have barely been considered thus far” [15] (p. 254). The data are approached as one-dimensional text, which is the most used source used for discourse studies, although Twitter exists since the early 2000s [16,17]². This lack of theoretical elaboration and adaptation to the specifications of interactive social networks, such as the impact of users or non-human users [18], presents an obstacle.

Often, method and theory are not combined, but a single fragment, such as grammar [19,20], influence measurement [21], sentiment and interests analysis [22,23], syntax practices [24], statistical analysis [25], or opinion research [26], is in focus. Seldomly, there are multilayered approaches that combine geolocation data, speaker positions in the network, and the content of tweets [27]. Most network analyses focus on mere structures and numbers, without further examining the content [28]. Studies with large corpora, even social network analyses (SNA) [29], are statistical and leave out the theoretical frame and the content of tweets [30]. Corpus and network approaches that include discourse analysis in their description (ab)use it merely as an umbrella term without depth, such as [31] and [31] including critical discourse analysis (CDA) based on critical discourse studies (CDS) [32,33,34,35,36,37,38,39,40,41,42]. Usually, CDA is combined with corpus linguistics (CL) when a larger corpus is analyzed [13,43,44,45,46]. At this point, not only reproducibility but also compatibility are in question since the existing method mixtures have theoretical inconsistencies due to incompatible premises [40,47]. Due to the enormous amount of data, “the empirical analysis of user-generated content typically requires the adoption of quantitative and automated data mining tools” since “semantic analysis methods also have limitations because most of these methodologies have been developed for more coherent and traditional texts rather than for analyzing conversations” [22] (p. 784). The vast number of shared content on Twitter—circa 500 million tweets sent per day³—as well as the complexity of user networks and the involvement of non-human users, such as social bots, are regarded as challenging for the common approaches of discourse analysis [48,49] or hurdles for qualitative methods [15]. Mostly tweets were not analyzed as embedded in a network; therefore, specifications regarding the positions and connectedness may be overlooked.

However, it is possible to combine a specific version of discourse theory with a method that is already used for larger amounts of data and add components of network analysis. I propose to adopt Wolfgang Teubert’s approach [50,51], which is adapted to the theoretical premises of PDT. Especially the concept of dislocation, at this point simplified as “the space of possibility for (progressive or repressive) change” [52] (p. 269, emphasis in original), plays a crucial role in understanding discourses as continuously struggling formations. Thus, PDT is deemed more suitable, since the tweets express a struggle with dominant beliefs and the discourse shows signs of dislocations, which are “processes of social change, as they produce structural gaps that have to be filled, situations of fragmentation and indeterminacy of articulations” [9] (p. 166). CL is useful to “identify linguistic patterns that occur across large sets of texts” [53] (p. 106), which is why I will use both.

3. Discourse Theory and Populism

Discourse studies bundle the approaches to explore the meanings of notions that are produced by the use of language (or other non-textual ways of communication), the processes and power-related contexts of shifts in meaning, and the practices produced by them [33,35,37,54,55,56,57]. As part of discourse studies and specifically the so-called Essex School, PDT is mostly based on Ernesto Laclau’s works, especially Hegemony and Socialist Strategy—Towards a Radical Democratic Politics [58]. The main pillars consist of “a combination of post-Marxist social thought and post-Saussurian linguistics” [59] (p. 3), namely the abandonment of Marxist ecological determinism, adoption, and adaption of Antonio Gramsci’s approach to hegemony and the appliance of semiotic theories to social processes in reference to Ferdinand de Saussure and Jacques Derrida.

In PDT, all meaning is viewed as changeable and solely constituted by all-encompassing discursive structures. This view is built upon Saussure’s separation of meaning and into the signifier and signified [60] as well as Jacques Derrida’s addition of the “arbitrariness of the sign” [61] (p. 47) and lack of essences [61,62]. Stemming from these assumptions, Laclau and Mouffe claim that meaning has to be generated negatively between objects, subjects, ideas, practices, etc., due to the absence of a perpetual and immanent core of meaning as well as randomness and volatility [58]. Meaning is constituted negatively—“to be something is always not to be something else” [58] (p. 115). For instance, an apple is an apple not because it has some inherent features that make it an apple but because it is not a cherry or a banana. An apple could also be an electronic device instead of a fruit in another discursive context [63] (p. 254). Thus, the production of meaning is also viewed as a relational “play, thought as absence of the transcendental signified” [61] (p. 50, emphasis in original). Each time a concept develops another meaning in this endless play, it brings along parts of its former meaning(s). This time factor is expressed by Derrida’s neologism différance [61].

In discourse studies, PDT is used to frame and inquire how exactly the play’s rules develop(ed) by assessing practices, happenings, and ideas. This is conducted practically “by analyzing the way in which political forces and social actors construct meanings” [57] (p. 129). However, the analysis of meaning formations is not a reconstruction of causal chains but a wholesome approach to viewing discourse as equal to the social. Thus, verbal expressions, actions, objects, subjects, and their practices have to be perceived in a larger and more complex frame, in which they all are linked to discourses with specific conditions built on hegemonic structures that have been established through sedimentation, the hardening, and naturalization of beliefs, practices, and relations [64]. In other words, “the social production of meaning, which is structured under the form of discursive totalities” [65] (p. 93) is understood as a system where discourse is “both process and product” [66] (p. 395). “The sedimented forms of ‘objectivity’ make up the field of what we will call the ‘social’” [67] (p. 35, emphasis in original), and the social is equal to discourse. Another aspect regarding the basic assumption is that in PDT, “a world external to thought” is neither denied nor positivistically validated. This is related to broader discussions on idealism, realism, and materialism, but Laclau and Mouffe claim that everything we conceive as subjects within discourses is made comprehensible through discourses⁴.

Having established that “our perception of reality and the character of real objects is mediated entirely by discourse” [59] (p. 3), I will focus on the components of this “differential and structured system of positions” [65] (p. 93). The process of structuring and partly fixing meanings within discourses follows a specific practice named articulation that transforms elements into moments. Hence, articulation modifies elements, components without a differential relationship between each other, to moments, components with a partially fixed meaning for a particular discourse that are identifiable by their differential relations to other moments. Differential positions, insofar as they are articulated in a discourse, are moments with a particular identity. Therefore “all identity is relational and all relations have a necessary character” inside a discourse [65] (p. 92). Considering that the moments are just identifiable because of their positions and relations to other moments, the establishment of structured meaning relies solely on the logic of difference and its complementary logic of equivalence. Equivalences make “differences cancel one other out so far as they are used to express something identical underlying them all” [65] (p. 113). Laclau illustrated this with the example of the populist People’s Party, where political activists in the 19th century distinguished themselves due to their heterogeneity (different classes, skin colors, and genders). Nonetheless, they have unified and could overcome their differences for the benefit of common aims [68]. Due to the interaction of the two logics, “all identity is constructed within this tension between the equivalential and the differential logics” [68] (p. 70) and never fully settled [58].

As a result of these practices, nodal points emerge. They reflect a discourse’s strive to “dominate the field of discursivity, to arrest the flow of differences, to construct a centre (sic!)” [58] (p. 98). They are temporally and partly fixed meaning formations in a discourse with a stabilizing function as representatives of chains. “[T]he demands of ‘peace, bread and land’ in the Russian Revolution, which condensed a plurality of other demands” [69] (p. 193, emphasis in original) are typical nodal points [65] or empty signifiers⁵. This empty signifier is a “signifier without a signified” [70] (p. 36). To develop into an empty signifier, one signifier in an equivalence chain empties itself from its previously associated particular meaning by “stepping in and becoming the signifier of the whole chain” [68] (p. 131). Thus, it is empty but at the same time full of the meanings of the represented chain. Consequently, structures start to manifest and become dominant enough to be called hegemonic [70]. Freedom is a good example of an empty signifier that acts as a placeholder for many meanings, such as the freedom to vote, travel, marry, or chose a job [9,63]. Regarding Twitter, hashtags such as #metoo [71] or #NotDying4Wallstreet could be categorized as empty signifiers, although some academics view hashtags as more active participants since they produce articulation points [15,72].

Another stabilizer next to the equivalential chain and empty signifier is the antagonistic relation, ensuring a constitutive order with a “distinction between discourse and general field of discursivity” through antagonistic forces or frontiers [58] (p. 121). Since a discourse can be perceived as a formation with reduced and sedimented meanings, every other meaning possibility and the surplus of meanings in other discourses are in this outside field from the perspective of one particular discourse. Laclau recommended “to think society as two irreducible camps” [68] (p. 83), which is comparable to a constitutive opponency of white and black chess figures.

The last part of PDT stimulates discursive change and disrupts sedimented and hegemonic structures. Once established, “a ‘forgetting of the origins’ tends to occur; the system of possible alternatives tends to vanish and the traces of the original contingency to fade” [67] (p. 34, emphasis in original), but a dislocation can weaken such a seemingly naturalized structure. As emphasized by Nabers, so-called crisis events should be theoretically reframed as a constitutive part of change within discourses, namely dislocations [7,8,9]. During the constant fight between stabilization and articulation, the temporary sedimentation of meanings within hegemonic structures can be dislocated, which then leads to “drastic recompositions” [64] (p. 82). During catastrophes, acts of terror, or other happenings that are perceived as violent, the existing structures face the difficulty of building links to additional moments. What happened must be made comprehensible within the established meaning structures. Since “all discourses are finite” [73] (p. 16), which means the variety of meanings is limited to the available particular meanings within a discourse, it is unfeasible to link every possible new meaning related to dislocations to the current structures. Thus, dislocations force discourses to change. They also affect the subject, its identity, and its position in the discursive formation. Since a subject’s identity is never fully fixed, during a dislocation, its “efforts to rearticulate and reconstruct the structure also entail the constitution of the agents’ identity and subjectivity” [67] (p. 51). In an attempt to overcome their “failed structural identity” [67] (p. 44, emphasis in original), subjects try to overcome this lack through rearticulation with new links and new constitutive antagonists, since “a constitutive outside facilitates the displacement of responsibility for the split subject’s lack onto an enemy, which is held responsible for all evil” [73] (p. 17). Each struggle and rearticulation results in a new formation, even if the order strives for preservation and reestablishment so that dislocations lead “not only to negative consequences but also to new possibilities of historical action” [67] (p. 39). To sum it up, dislocations are a drive for change and contingency [64,65,67,73].

The emergence of a social movement can be an example of new possibilities and change, which leads us to Laclau’s approach to populism. Since the data include several populist features, I will give a short introduction to populism definitions, especially the discursive version. Even though the notion of populism exists for several decades in the scientific literature [74,75,76], populism is debated in the scientific literature as well as the media. “The lack of consensus around a definition of populism” [77] (p. 2) is becoming increasingly obvious in the heated-up scientific debates of recent years [78]. A problematic issue is that through “conceptual stretching” [79], populism degenerated to “a catch-all term in the general public discourse” [80] (p. 440), [80,81], which is ambiguous [82] but mostly negatively connotated as “anti-democratic” and framed as threatening to “derig the liberal order” [83] (p. 44), [84,85]. The various definitions can be broadly categorized into branches that view populism as a political strategy [86], a certain (performative) style or rhetoric [87,88], ideology [89,90,91], or discursive mechanism [66]. Next to this conceptual level of the definitory complexity and confusion about populism, the “connection between charismatic leaders and populist mobilization is a central feature of most contemporary theories of populism“ [92] (p. 55), [93,94,95]. Especially case studies about Latin American populism show a strong focus on charismatic leadership, which is viewed as a definitory feature of populism [95,96,97]. According to Paul Taggart, populism even “requires the most extraordinary individuals to lead the most ordinary of people” [98] (p. 1). Considering that most definitions of populism include the concept of the people and their counterpart as the elites, the establishment, government, or a variation of a group/institution in power [96,99,100,101], I will focus on the formation rather than content-related definitions. Since “most of the time the term is used to describe any form of non-compliant political actor or movement” [102] (p. 26), a simplified approach without any essential features is deemed more fitting⁶.

True to his abandonment of essences, Laclau defines populism based on form rather than content. In the preface of On Populist Reason, he states that “populism has no referential unity because it is ascribed not to a delimitable phenomenon but to a social logic whose effects cut across many phenomena. Populism is, quite simply, a way of constructing the political” [68] (p. xi).

The form of populism is expressed by the division of society into two antagonistic camps: the people vs. “the dominant ideology” or the “existing structure of the power bloc” [103] (p. 173). The people, as a construct by the discourse, are defined as a group with just assumed homogeneity. They strive for the realization of their demands, which are heterogeneous but act as a unified group that gathered for a common reason. “Populist discourse does not simply express some kind of popular identity; it actually constitutes the latter” [104] (p. 48), and as a result of that unification for a shared aim, an empty signifier arises [68]. The criteria for Laclau’s definition are summarized in

Table 1. Conceptualization of populism’s logic following Laclau’s definition.

Formal Criteria	Laclauian Populism
Equivalence	The people being underdogs and under threat
Subject position	Member of “the people” as a solidarized group
Constitutive outside	Establishment/elites, dominant structures
Relation to outside	Vertical up/down, hierarchy, power, economy,…

below.

It has to be emphasized that the discursive approach by Laclau offers a way to circumvent the question of whether leadership is a necessary feature of populism. According to Laclau, any potentially existent leader evolves into an empty signifier as the symbol for all the meaning they represent in the equivalential chain of the united demands and identities of the unified people. Since in the online movement linked to #NotDying4Wallstreet, a leader seems to be nonexistent at first glance, in Laclau’s approach to populism in which every subject is eventually framed as a political symbol, the spotlight is transferred from agency and personal power toward social structures in a more wholesome way [92]. Lastly, I want to offer a simplified visualization, a snapshot of a formation where moments are in differential relations to each other and linked by chains of equivalence and one illustrative empty signifier in Figure 2 below.

Despite its complexity and flexibility, there are several issues with PDT regarding social network data. PDT acknowledges subjects but does not cover other types of participants, such as algorithms that partake in online discourses.

4. Actor-Network Theory

Due to the lack of understanding of Twitter as a network and defective theoretical framing for subjects and non-human participants in the literature, I will introduce parts of actor-network theory (ANT) to enrich PDT. In ANT, everything, nature, and society are linked through relations that form changing networks, which is similar to the linked and changing relational and differential meaning concepts and subject positions in discourses [105,106,107,108,109,110,111]. However, “ANT is not a theory” [112] (p. 194) but a wholesome and inclusive way of thinking and perceiving [113,114]. With this perspective, ANT helps to theoretically embed different types of actors: human users, social bot accounts led by algorithms, and hashtags that distribute content or connect people [15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,64,65,66,68,70,71,72,115,116,117,118]. Due to the features of computational agency, the “boundary between humanlike and bot-like behavior is now fuzzier” [49] (p. 99), [114,119,120,121]. Here, non-human actors are re-labeled as quasi-objects and actants [122,123]. Any analysis of these actor categories is conducted without a priori assumptions so that ANT works with a rather “flat ontology” [124] (p. 275), which I will briefly introduce and compare with PDT.

First, a “flat ontology rejects any ontology of transcendence or presence that privileges one sort of entity as the origin of all others and as fully present to itself” [125] (p. 245), which corresponds with PDT’s premises of no essences or eternally fixed meanings. Second, “there is no super-object” [125] (p. 246), which is also according to PDT. Third, although a flat ontology “recognizes that humans have unique powers and capacities” [125] (p. 246), we as humans should neither “put epistemology before ontology” nor “begin by negotiating conditions of cognitive access to the world” [126] (p. 65). Even imagined objects are regarded the same as any other object or subject, so world–mind dualisms become redundant. PDT emphasizes that everything is constituted in a way that “depends upon the structuring of a discursive field” [65] (p. 94). Here, subjectification comes into play. According to the flat ontology and ANT, both subjects and objects are subjectified. Since in PDT, everything is regarded as a part of discourse, objects as well as subjects, the term “subjectification” could be stretched to include bots, which are objects behaving like subjects. This fits the claim that humanity has already reached a level of globalization and technological progress, that there are no pure natural objects anymore. so “[n]ature behaves as a subject” [127] (p. 36). And lastly, it says that “all entities are on equal ontological footing and that no entity [...] possesses greater ontological dignity than other objects” [125] (p. 246), emphasizing the anthropodecentric stance⁷. Since these basic premises of PDT and ANT align, I will focus on actors and acting.

ANT defines acting as “any thing that does modify a state of affairs by making a difference is an actor” [111] (p. 71). This difference could be a heavy key tag forcing hotel guests to leave the key at the reception instead of taking it with them [128]. In this example, the hotel guests, keys, and receptionists are all active actors. In socio-technological networks, such as Twitter, algorithms equipped with learnable software can have unpredictable and complex effects on the network [49,107,129]. Regarding network participants, both ANT and PDT use “difference” to describe them. Thus, difference is constitutive for discourse or network participants as a relational sign for activity. By labeling social bots and hashtags as quasi-objects, I claim they make a difference in the social network’s actions. Bots can also be “subjectified” since the algorithms learn how to interact with users and other bots. Like human subjects, self-learning algorithms can be altered by the network they are embedded in so that the term “subjectification” fits these quasi-objects. Hashtags, however, are similar to empty signifiers that can have varying meanings and even overflow with meaning, such as in the case of the hashtag #metoo [71]. Still, hashtags are only content distributors, which is a complex but subtler way of acting and making a difference. Considering the enhancement of “subjectification,” I propose a gradual approach where quasi-objects are differentiated based on their abilities. such as learnability. Since social bots are constituted and constitutive, they can be subjectified and contribute to the subjectification of other discourse and network participants, which is why I would view them nearer to society and human subjects. This enables researchers to differentiate between actors and identify bots and allows conclusions about the message spread with the help of bots during important happenings. such as elections, catastrophes [130], or a pandemic [18,131].

5. #NotDying4Wallstreet in Analysis

Twitter poses two methodological hurdles, which are data collection and analysis. I will give a brief overview and introduce the features in

Table 2. Overview of Twitter’s communication features.

Author/User	A person, group, or social bot that is “connected through an underlying articulated network” [47] (p. 2), [136].
Mentioning	Reference to a person. A user practice that was incorporated into Twitter’s code; thus, Twitter is a performative network [24,137].
Social Bot	Algorithms that learn, interact, and pose as humans [49,138].
Retweet	Reiteration of an initial tweet, a new conversational practice [23,139,140]. In addition, Twitter “adds a new twist to the death of the author” [24] (p. 1) since the retweeted text is of higher relevance than the original author [47] (p. 4). “[T]he retweeter wants to not only rebroadcast another’s tweet, but also add commentary” [24] (p. 5).
Hashtag/#	Represents topics as a searchable tag and was performatively established by users [141].

. Concerning the theoretical implications, the download limits set by Twitter mean that any analysis is a fragment of the network online. In addition, tweets differ from traditional textual sources due to clustering with positive correlation so that users are divided into communities where “individuals belong to groups and are acquainted with others with whom they share those groups” [132] (p. 7). Followers ensure reciprocity and a wide reach in these communities, whereas traditional textual sources are monologues directed to the reader [24]. Since they are “distributed across a non-cohesive network in which the recipients of each message change depending on the sender, […] conversational structures are missing” and without “an ordered exchange of interactions, people instead loosely inhabit a multiplicity of conversational contexts at once” [24] (p. 10), leading to “conversational relaxation” [133] (p. 26) and a “continuous feedback loop” [134] (p. 323). Thus, we “formulate our thoughts more freely” [47] (p. 4), [135] in the digital realm.

Thus, the practice of retweeting is linked to issues concerning authorship ascription and the traceability of conversation patterns, since “conversations are distributed across the network, referents are often lost as messages spread and the messages themselves often shift” [24] (p. 1). This causes split or incoherent conversation strings, abbreviations, wrong assignments, premature conclusion making, and lacking syntax rules [24]. Hashtags also pose several issues. The analyzed tweets, for instance, are just connected by the hashtag as the only conversational link between them. The use of “searchable talk” [19,140,141] helps to detect, visualize topics, and enable “different dimensions of the discourse to be retrieved by search” [141] (p. 284). So, the ambivalence of connotations by the indirect interaction of users, who do not know each other, and the specific contexts of the words create new meanings. These initial meanings are also changed by hashjacking, the reinterpretation or occupation of a hashtag by an opposing group [142,143]. In the case of #NotDying4Wallstreet, the term ‘Wallstreet’ is not about the stock market or the financial district as a signifier for a place or institution but has been used as an embodiment of greed, forced profit making, and the neoliberal capitalist economy in general. Consequently, hashtags can degenerate into spam or evolve into a catch-all term/empty signifier [144,145] (p. 193). To detect such shifts in meaning, it is useful to focus on collocates, concordances, and clusters to find additional hashtags and notions in the context of the focused keyword [145] (p. 196), [146]. However, the unique features of hashtags enable users to “enact relationships rather than simply share information” [19] (p. 2), which sets social network data apart from traditional texts. Especially, topical hashtags are the digital equivalent of a “speech at a public gathering—a protest rally, an ad hoc assembly—of participants who do not necessarily know each other but have been brought together by a shared theme, interest, or concern” [147,148] (p. 18).

In summary, the features and practices show that Twitter is an effective dissemination platform that is said to have played an emancipatory role in hegemonic struggles, such as protests in Iran, Bahrain [149], and Egypt [150], since it can “facilitate things like grassroots political action in places where censorship and surveillance make such mobilization difficult” [47] (p. 5). The online communication is valued as an enhancement of public spheres to enact identities [151,152,153,154].

5.1. Methodic Approach

This paper’s mixed-methods study combines quantitative discourse analysis (lexicometry), qualitative discourse analysis, and SNA. MAXQDA [155] was used for data collection and the software Gephi [156] for network visualization. The data consisted of 4000 tweets based on the hashtag #NotDying4Wallstreet from the first day after the hashtag emerged on 24 March 2020. The original corpus includes circa 40,000 tweets from several weeks until the hashtag #NotDying4Wallstreet was not used more than a few times per day. To focus on the issues right at the emergence of the hashtag, the developed design used just tweets from the day of origin. The research design consisted of three steps: (1) data cleaning and sorting, (2) network and content exploration, and (3) in-depth analysis of formations, meanings, and relations. The first step identified original tweets, retweets, and replies and helped to build initial categories regarding type and content. Since the protesters view themselves as a marginalized group under threat by the government due to COVID-19 policies, the empirical data are a representation of marginal(ized) positions [115].

In step 2, I split the analysis into a network and a content level, as suggested by Lindgren and Lundstöm [157]. The network was mapped with a focus on different levels of information, such as user, retweets, or follower maps. After the most interconnected users were identified, in a sense the loudest voices, these actors and their content output could be analyzed more purposefully based on the assumption that quantity and interconnectedness suggest relevance [51]. Starting from these path markings within the network exploration, the tweets were searched for initial keywords and topics so that lists of the most used words and hashtags could be used as entry points for the content analysis. As visible in the tweets later, the authors often have attributes suggesting their roles in society as working people in contrast to politicians within the establishment. Such oppositional positions lead to hegemonic meanings and inside/outside relations based on antagonistic frontiers.

For the quantitative part, Teubert’s lexicometric corpus analysis [50,51] was used as the basis. He describes corpus linguistics as an analysis of “language from a social perspective” [51] (p. 2) and “an imperfect methodology to make sense of the discourse” [51] (p. 13). Since our access to the world is filtered through perception as participatory subjects within discourses, “the discourse, and not the world out there, is the only reality to which we have direct, unmediated access” [50] (p. 8). It is impossible to directly look into a subject’s mind, so the only way to inquire about discourse is an indirect path through text. For that purpose, Teubert’s primary tool for analysis is frequency based on the assumption that “recurrent patterns defined by the co-occurrence of words” and “complex units of meaning” [51] (p. 5) can be searched in large corpora since frequency indicates relevance. Although frequency will be used “for making general claims about the discourse,” Teubert’s quantitative approach is not solely focused on “statistical ‘significance’,” since “[l]exical items also have to be semantically relevant” [51] (p. 5, emphasis in original). Semantic relevance is understood as follows: “When we negotiate the meaning of a text segment, we do this within the discourse, not outside or on top of it” [51] (p. 7). This argument is of particular importance in the context of my earlier critique on the extraction of statistically relevant topics and keywords from corpora in other CL and CADS approaches based on reference corpora. Teubert suggests here that the criteria for relevance lie within the corpus so that meaning or frequency comparisons based on deviation from reference corpora used as indicators of a normal way of language usage would be illogical in the context of discourse analysis. This aspect is linked to the theoretical implications of PDT and distinguishes Teubert’s lexicometric approach from other CL approaches.

As noticeable, the tools for the qualitative part of the analysis were inspired by the understanding of coding and circular research based on Rainer Diaz-Bone and Werner Schneider. They propose practical examples of qualitative discourse analyses with data analysis software and define coding as the process of marking and labeling text in the corpus analysis so that categories and meaning structures may be crystallized from the masses of text [158]. Since the theoretical premises behind these analytical practices, which are based on grounded theory [159], were already adapted to discourse analyses, they can be integrated into this methodic approach.⁸

For the network part of the analysis, I adopted the perspective of Evelien Otte and Ronald Rousseau. SNA is a process of “investigating social structures” in a strategic manner suitable to search for links and “the social context of the actor” [29] (p. 441).

5.2. Findings

Step 1 revealed that the content of every tweet is linked to the general discussion on health issues in the context of COVID-19, the dangers of the pandemic, and their relation to possible economic damages in the United States. The data include over 3.5k retweets that constitute the majority, whereas nearly 400 tweets are original posts, and nearly 100 of the tweets are replies. The high frequency of adjectives and attributes indicates that the discourse is emotionally charged. Their wording establishes strong fault lines in combination with warlike/militaristic rhetoric, and the names of politicians who were either blamed or praised, for example, rich and poor or the Manichean dichotomy of good and evil, in combination with directly addressed persons, such as Trump, Fauci⁹, and Texas Governor (Dan Patrick). This Manichean divide has been categorized as a feature of populism in one of its many definitions made by Kirk Hawkins [160,161]. In addition to adjectives, family members (mother, mom, grandparents) and expressions related to national and political identity (people, folk, American, country, USA, vote) are also frequently used. Interestingly, pronouns are also instrumentalized to generate a distinction between the own group and the othered outsiders. Pronouns, such as you or we, indicate a sharp antagonistic inside–outside relation, and demands for action, such as want, need, and more dramatic requests to resist, kill, or sacrifice, are often combined with these pronouns.

In the next step, the network was explored with Gephi. The subfigures A–C (Figure 3) illustrate the modeling of a retweet–network of accounts that posted the tweets as source nodes and the accounts that retweeted their content as target nodes with a direct link. Following [148] this type of analysis is categorized as a macro-level network exploration [162]. The network exploration consisted of filtering and rearranging based on the following questions: (1) Whose content was retweeted most often? (2) Who retweeted the most content?

As visible in the screenshot in the middle, the network’s layout is dominated by a blue cluster from the account mmpadellan, who expressed his solidarity with the hashtag #NotDying4Wallstreet and gave advice on trustworthy persons during the pandemic by claiming that the people should listen to medical experts, their governors, and Fauci [163]. The cloud-like accumulation of nodes next to this account indicates that this tweet was retweeted by accounts that were otherwise not active in this context. The findings support the initial impressions from the word frequency cloud regarding dichotomies. Especially, Trump is constructed as the dangerous and unreliable antagonistic other who could cause the death of the people who are openly displaying their mistrust in him, although he was elected by the people. The frontier gets more complicated regarding the perception of the governors, since most governors, except for the Texan governor, acted differently to Trump’s orders during the pandemic and established state policies. Based on this, governors, Fauci, and medical experts are constructed as the ‘good people’ [164]. Although the shared tweets do not always include words such as ‘true’, ‘false’, or anything else to confirm or deny facts and hint at accusations, the structure of the clusters, for instance, that each of them is focused on different persons, enables the categorization of these politicians based on how the network looks like.

Next to this, accounts were examined regarding their behavior to detect unusual patterns, such as inhumanly high rates of tweet frequencies or unnatural speech patterns, so that social bots could be found [165,166]. Bots are often used in political campaigns [167]. The most connected accounts in this network, @ahdrag, and @HumbertoDeLaHo8, show signs of typical bot behavior¹⁰.

Finally, the corpus was analyzed qualitatively to find co-occurring words, clusters of specific word groups, and their relations. The most frequent word pairs involved persons such as ‘“the governor’, ‘Trump’ and the alleged accusation ‘you kill’. A detailed exploration of frequently retweeted sentences including ‘Trump’ led to several accusations. Figure 4 shows a word tree of sentences that follow the word ‘Trump’. The first sentence in the figure is about the discussion on re-opening the US economy, which is expressed by the word capitalism [168]. Although the pandemic is seen as a situation where the US economy should be secondary, the workers were asked to die for capitalism instead of staying safely in lockdown. It is noteworthy that the tweet addresses the readers directly with the pronoun “you,” which can be interpreted as an attempt to create a connection and form a group that includes all the working people against capitalism and Trump as its enforcer. The second sentence goes on as follows: “Trump is warping us into a 70s dystopian sci-fi movie by calling for human sacrifices on the altar of Wall Street, framing this as the ‘cure being worse than the disease’” [169] (emphasis in original). It indicates that next to the traumatic experience of a pandemic and the threat of sickness and death by COVID-19, Trump’s call for re-opening adds another layer of destruction. The situation is perceived as unjust since the economy is prioritized over human lives and at the same time traumatic and disruptive to a degree that it is imagined as the plot of a dystopian sci-fi movie by the tweeters und hundreds of retweeters.

Another connection with Trump can be found in additional hashtags that were often used next to #NotDying4Wallstreet, such as #TrumpLiesAmericansDie. Tweets with these two hashtags also include several more hashtags, such as #DiefortheDow, #reopenAmerica, and #TrumpCrash. These tweets also clearly construct Trump as the perpetrator, the evil and threatening other that is in an antagonistic relationship to the self/own group. The tweet on the left side of Figure 5 describes him as a member of the so-called 1%—a small group of people with great power and resources [170]. Trump is accused of prioritizing the maximization of profit over the well-being and literally the lives of people. Before I explain all involved parties and their relations in this antagonistic struggle, two things must be pointed out.

First, these other hashtags widen the discussion about COVID-19, reopening the US economy, and Trump’s policies since they are also included in tweets that do not have the #NotDying4Wallstreet hashtag. This hints at a potential limit of the search query based on one specific hashtag in this study. Tweets without this hashtag also have corresponding content that could add valuable information, which shows the complexity of network structures that may have overlapping contents but just indirect links through different hashtags. And second, the tweet involving these other hashtags includes a photograph of a poster that was stuck on in New York City on the 24th of March 2020 by a comedy duo named The Good Liars, who tweeted about this [171]. This poster shows Trump in the same pose as the iconic ‘Uncle Sam’ recruitment poster that was used during World War I. I will just focus on the text since this is a text-based study. There are two interesting aspects. First, the poster displays a demand with a similar wording as the tweets criticizing Trump’s policies. The ‘you’ addressed by the poster is a part of the 99% of working people who are demanded to die for ‘our’ economy. The pronoun ‘our’ clearly distinguishes the 1% owning the economy from the 99% threatened by COVID-19, which strengthens the antagonistic relationship between these two groups. And second, the tweeted poster shows an interlap between putting on posters in the real world and the digital realm of Twitter. This indicates that discussions and social practices may begin in one realm but can continue in the other. In such cases, it becomes debatable whether social media studies of protests should be categorized as mere online protests.

The last part needed for the analysis of the demands articulated around #NotDying4Wallstreet and the lines of social disruptions was a detailed examination of the we group that stands against the threatening other. Who are they, and what are their demands? A prominent combination is ‘we need’, with a specific demand to express the people’s requests for political changes with the help of leadership and strikes but also the need for an alternative vision for life. The people need protection from the threat of COVID-19, especially the health workers, as said in the tweet on the bottom right in Figure 5. These findings, in addition to the construction of the people and the establishment as their antagonists, show a clear populist character of the discourse connected to neoliberalism, healthcare, and social justice. Especially, the code ‘we need’ as a statement by a united group of persons constructing themselves as the people has populist characteristics since it includes a clear set of demands for political change and shows a bottom-up structure [68]. This impression is strengthened by the findings of the in-depth analysis with coding that leads to statements such as “#NotDying4WallStreet We need more out of life than being profit-producers for the sociopathic elite” [174]. Statements like this construct the “sociopathic elite” as the antagonistic other to the people as “profit-producers” who are currently threatened by the pandemic. In addition, shared solidarity as united working-class people can be found in tweets such as “We can do this, people. #GENERALSTRIKE NOW! US workers have done it before!” [175]¹¹.

These findings also relate to other works emphasizing how the pandemic has been constructed as populist¹². Based on the populist features found in the corpus, the content analysis had a stronger focus on typical words, such as ‘people’ and ‘elites’, ‘establishment’, or ‘government’. Eventually, as it was suggested at the beginning, this online movement is leaderless, which stands against many definitions of populism with a charismatic leader figure. The sole unifying factor is the hashtag #NotDying4Wallstreet, which presents the people’s diverse demands to act against the reopening of the economy as it was initially proposed by Trump. An interesting detail about the way these demands are articulated is that they are not demands for future changes but demands to get the people, the workers, what they would have deserved in the first place, namely “security in the workplace and social protection for families” [176]. The demanded “adequate health and safety measures [...] especially relevant for health workers” (Ibid.) are codified in Goal 8 Decent Work and Economic Growth within the Sustainable Development Goals (SDGs) launched in 2015 by the United Nations [177]. The word choice in the tweets, especially the emphasis on the unjustness of the prioritization of economic well-being over health protection, shows anger about being denied such basic human rights. Their anger is mixed with frustration since the threat of COVID-19 and the trauma linked to the pandemic are deepened by the trauma of getting sent back to work without health protection for the sake of the economy by politicians who were expected to prioritize human lives.

The following Figure 6 illustrates the sum of my findings: The left equivalential chain represents suffering due to the pandemic and issues related to it, such as job loss, health (insurance) problems, uncertainty, and helplessness, leading to accusations against politicians, such as Trump, and ‘Wallstreet’ as capitalism’s representatives. On the other side of the frontier (dashed line), greedy elites are constructed as enemies upholding an economic order and threatening policies, such as lifting the pandemic containment measures. Some dislocated parts of the chains try to cross the frontier (arrows).

6. Conclusions

This work’s initial aim was a profound analysis of the discourse around #NotDying4Wallstreet in a PDT context. The demands articulated around #NotDying4Wallstreet and the social disruptions represented by the discursive formations were in focus since the passionate online debate on the nexus of health and neoliberalism seemed promising at first glance. At second glance, however, the lack of theoretically and methodically well-developed research designs with tweets as data for discourse and network analysis manifested itself as a scientific gap. Based on this issue, this work followed a dual approach as a step toward a better analysis of tweets with the purpose of not only identifying problematic theoretical and methodical issues but also presenting a coherent theoretical enhancement of PDT with ANT and a mixed-methods research design, with tweets around the hashtag #NotDying4Wallstreet as an empirical execution of the proposed approach.

One of the new theoretical proposals is the expansion of the understanding of subjectification in PDT to include quasi-objects with the ability to not only be subjectified by others but also have a subjectifying effect on their environment based on algorithmic self-learning capabilities. This inclusion of a network perspective next to the discursive perspective goes beyond the subject positions of humans and focuses on the significance, position, and weight of the network’s participants (human accounts and bots) so that their role as content multiplicators in cluster dynamics may be acknowledged and theoretically framed.

Regarding the findings that were content related, the analysis revealed two main topics: a dystopic future of an accelerated neoliberal order in which human lives are sacrificed for a well-functioning economy and the motive of protest, including leftist solidarization inspired by the spirit of historic working-class movements, and populist features within the demands and construction of the antagonistic other(s). The frontier was drawn between the hardworking people who voiced their protests by using the hashtag #NotDying4Wallstreet and parts of the ruling elites, namely Trump and his governors, who demanded a fast re-opening of the economy instead of prolonged COVID-19 containment measures. The hashtag #NotDying4Wallstreet, initially a statement against the willingness to die for the sake of the neoliberal order represented by ‘the Wallstreet’, also included descriptions of unfair living conditions in comparison to the so-called rich and greedy elites. The protests in the context of the analyzed hashtag demonstrate that the call for a reopening of the US economy is perceived as a major threat that adds another layer of trauma next to the trauma of COVID-19. The analysis of the tweets and the resulting antagonistic relation showed that instead of being protected by their some of their elected politicians, the people are directly faced with the pandemic due to the prioritization of the economic well-being. Another finding is about the feature of leaderlessness of an online grassroots movement that managed to stand against the attempted ownership of Trump as a leader who is usually depicted as a (right-wing) populist. As initially cited, populist leaders “tried to politicize the pandemic to increase the antagonism between the people and the elites” [14] (p. 149). #NotDying4Wallstreet is the story of such a failed ownership. Although Trump may have attempted to claim the pandemic through television and social media performances, an attempt to perform himself as the leader of the people and the medical experts and the press as the antagonistic elites, another group of people standing against Trump emerged. Contrary to the antagonistic frontier in Trump’s version, the people from #NotDying4Wallstreet constructed him as their enemy and the medical experts as the good ones. The struggle between the attempted ownership of the pandemic and the frontiers not only emphasizes that the people are a discursive construct but also demonstrates the relevance of social structures and their analysis. “[I]f one defines populism as a personal strategy for power accumulation, social structures seem to recede into the background. If one defines populism as a social discourse, the strategic autonomy, and agency of the leader disappear” [92] (p. 66). By analyzing the tweets as part of an online movement in a larger discourse on the nexus of health and economy, their populist character could be explored, which led to two conclusions. First, populism is so complex that more than one attempted construction of the people and antagonistic frontiers can emerge. And second, populism should not be exclusively restricted to narrow definitions that just focus on leadership, but understood as social constructions with or without a leading figure.

Since my data corpus was limited to the first day of the online discussion around #NotDying4Wallstreet, the development—or rather disappearance—of this discourse over a few weeks could be another point for further investigation, especially in the context of the emergence and the process of disengagement of social movements. Due to the complexity of content relation and a variety of overlapping hashtags, I also recommend a widening of search words to find networks with overlapping contexts to circumvent the limits of sparse linkage between tweets. Due to the limits of my data corpus and the search parameters, this study is not representative but should be viewed as an analysis of a snapshot of changing discursive formations.

Finally, I would like to stress the importance of a coordinated theoretical frame and methodical approach as key factors. The enhancement of PDT with ANT and the adapted mixed-methods approach pose an update of discourse analysis for a new age of social network data. If we rightfully acknowledge that social networks are valuable sources to connect to the zeitgeist of online discourses, the theories to grasp and frame these discourses as well as the methods to access and explore the masses of data must be up to date. I propose the further development of both by taking other approaches into consideration. Since the tweets in this study were often accompanied by pictures, image analysis techniques could be a useful extension of the analytical tools. Another complementing approach can be found in the information and communication technology studies. One methodological example is computer-mediated discourse analysis (CMDA) [178], and another is critical technocultural discourse analysis (CTDA) [179,180], which is partly based on critical discourse theory. The proposed consideration of theoretical and methodical perspectives from “neighboring” fields of studies enables better access not only to Twitter but also to discourses on other social networks. Such as Facebook, Instagram, and TikTok as well.