Simulating the world

Computer simulation in the age of the world picture

In his famous essay ‘The Age of the World Picture’, Martin Heidegger seeks to comprehend the essence of the modern age by determining the particular mode of ‘seeing and questioning natural events’ that is proper to modern science (1977: 117). Modern science, Heidegger argues, distinguishes itself from ancient or medieval forms of knowing in that it fabricates a world picture. By ‘representing’ or ‘setting-before [vorstellen] himself’ the world as a picture, ‘modern man’ gains an overview over a totality of which he is himself a part (Heidegger, 1977: 129). ¹ According to Heidegger, earlier periods in history did not organize their knowledge in this peculiar way. Only in early modernity did the world come to be grasped in the form of a picture and was thereby turned into something ‘at hand’ [zuhanden].

This diagnosis has led scholars to analyse how our experience of and access to the world are being shaped by specific pictorial means. Peter Sloterdijk (1999: 885–940), for instance, highlights the importance of early modern nautical charts in this respect. In his genealogy of the global – which is aptly read as a thousand-page footnote to Heidegger’s essay – nautical charts are presented as media for setting up a world that is planetary in scope and that is involved in a process of ‘terrestrial globalization’ (Sloterdijk, 1999: 801). In a similar vein, cultural theorists have examined how Google Earth has transformed our contemporary world picture. Different from the famous photographs of the Whole Earth taken from space in the second half of the twentieth century (Lazier, 2011), Google Earth re-fashions the planet as a manipulable digital object, turning viewers into users of software applications (Helmreich, 2011). Both nautical charts and Google Earth thus make the world observable and navigable by creating ‘operative’ world pictures (Krämer, 2013: 278). By providing a particular global vision, they form technologies of the globe.

This article adopts this analytical perspective on technologies of global vision to approach computer simulations of pandemic disease. The aim is to examine the kind of world picture that they delineate: how do pandemic simulations configure the horizon of our planetary situatedness, what form of global observatory do they produce by digital means, and in what ways do they render life on Earth amenable to interventions?

As recent work in the field of science studies has shown, computer simulations are a key technology through which our current world is depicted, circumscribed and rendered knowable. In his path-breaking analysis of climate modelling, Paul N. Edwards underlines their extraordinary status as infrastructures of the global: ‘Everything we know about the world’s climate – past, present, and future – we know through models’ (2010: xiv). ² Whereas the climate is clearly a preeminent site for using digital technologies to chart global dynamics, for social scientists, the simulation of infectious disease is of particular interest. In sociology, epidemiological contacts have long been considered to be markers of social contacts, up to the point at which social interactions have themselves been conceived of in terms of contagion (Wald, 2008: 68–113; Opitz, 2015). Today, ‘contagion’ has evolved into a dominant ‘trope of globalization itself’, referring both to the spread of pathogens and to the spread of social phenomena such as fashion trends, protest movements or financial panics (Magnusson and Zalloua, 2012: 4). This tendency to capture the sociological in terms of the biological is reflected in the approach of computational scientists. To simulate disease transmission, they have started to model the interactive behaviour of agents. ³ The infectious dynamic becomes rooted in a model of the social, the observation of in silico epidemics is irreducibly tied to the simulation of agential intercourse.

Intertwining the biological with the sociological by computational means in order to turn the world of contagion into something ‘at hand’ has intriguing implications. They can be illustrated by looking at the landmark study by Joshua Epstein and Robert Axtell. Working at the Brookings Institution in the early 1990s, they developed a simulation of an epidemic which is based on what they call an ‘artificial society’:

We shall refer to agent-based models of social processes as artificial societies. In this approach fundamental social structures and group behaviors emerge from the interaction of individual agents operating in artificial environments under rules that place only bounded demands on each agent’s informational and computational capacities…The aim [is] to discover…micro-mechanisms that are sufficient to generate the macroscopic…collective behaviors of interest. (1996: 3–4)

Given that computer simulation has also become a key method in sociology (Moss and Edmonds, 2005), it is important to note the shift in perspective at stake in this article. Analysing simulation technologies in terms of the world picture that they produce differs fundamentally from applying them as methodological tools. In the following argument, nothing will be said about the quality of data generated through simulations or the (dis)advantages they have compared to other methods. Instead, the aim is to explore how the simulation of pandemic disease places the world ‘in the realm of man’s knowing and of his having disposal’ (Heidegger, 1977: 130). Such an analytical perspective resonates with what has recently been called ‘the social life of methods’ (Ruppert et al., 2013). Under this programmatic heading, scholars examine social science methods as ‘modes of “making up society”’ (Savage, 2013: 5). They draw attention to the diverse array of techniques by means of which the social is both known and acted upon. For example, scholars have shown how population policies rely on data gathered through statistical census (Ruppert, 2011), how security agencies deploy social network analysis to trace terrorist groups (de Goede, 2012), or how marketing strategies use surveys to address consumers (Law, 2009). In a similar vein, pandemic simulations make global connectivities visible to provide information to public health authorities. They implement from within the social a certain relation to the social by producing a particular world picture of the social – and it almost goes without saying that the technologically induced observatory of global contagion, when taken seriously as animated social theory, will modify inherited notions of the social. Accordingly, pandemic simulations are to be studied as forms of societal self-observation. They adjust their software-coded frames to the realm of planetary life in order to provide guidance for certain interventions into the very reality they present.

To analyse the world picture generated by the simulation of pandemics, this article makes use of Niklas Luhmann’s social theory. In fact, Luhmann not only has identified processes of global self-observation as a key mechanism in articulating world society (Heintz and Werron, 2011; Guy, 2016), he has also decomposed the very notion of the world itself in a way that makes it amenable to sociological scrutiny. ‘World society’, Luhmann (2012: 87) once wrote laconically, ‘is the occurrence of world in communication.’ Influenced by phenomenological thought, he understands world society’s world as a correlate of social operations and, as such, as the recursive effect of meaningful events (Luhmann, 1982; 1997). ⁴ By making meaningful references, we situate ourselves in the world, and we do so in four dimensions: in the factual dimension, in the social dimension, in the temporal dimension and in the spatial dimension (Luhmann, 1995: 74–82; Stichweh, 2000: 184–206). Meaningful events can therefore be characterized by their particular referential structure in these four dimensions. ⁵

The subsequent four sections adopt this analytical frame to investigate the specific kind of world that ‘occurs’ in pandemic simulations. In doing so, they seek to determine the mode of societal self-observation that pandemic simulations allow for. The following questions will structure the overall argument: first, how do pandemic simulations change the way we refer to the fact of global disease transmission? (This is ‘the factual dimension’.) Second, how do they change the way we conceive of our own inter-relatedness? (This is ‘the social dimension’.) Third, how do they change the way we position ourselves in global space? (This is ‘the spatial dimension’.) And fourth, how do they change the way we refer to the past, the present and the future? (This is ‘the temporal dimension’.) Elucidating the very mode in which the simulation of pandemic disease re-addresses the world in these four dimensions provides insights into how humankind currently is learning to see itself positioned in a planetary environment riddled by potential crises. At the same time, examining computer simulation as a technology of global self-observation may also contribute to an understanding of the ‘technologically induced shift of meaning’ immanent to the ‘neo-cybernetic regime of truth’ (Hörl, 2012) that lies, at least according to some cultural diagnostics, at the core of our contemporary post-human condition (Hayles, 1999).

From representation to the doubling of reality: a god’s eye view on contingency

To determine the particular kind of facticity that pandemic simulations produce, it is instructive to start by inquiring into the problem they are supposed to solve. Today, pandemic simulations are expected to respond to the ontology of life enclosed in the notion of ‘emerging infectious disease’ (Davis et al., 2017). ⁶ This notion – that became prevalent in the 1990s (Lakoff, 2008) – designates a fundamental limit to the capacity of knowing disease since it characterizes microbial life as inherently unstable. Simply put, the term indicates that the matter of concern is no longer an identifiable actual disease but the event of emergence itself: pathogens may suddenly enter human populations through zoonotic spillover or change unexpectedly through genetic mutation. Due to the emergent properties of disease, one not only does not know when the next pandemic will occur or where it will originate. Moreover, one does not know what it will be, i.e. what kind of viral strain will emerge, how symptoms might develop, how transmission takes place, what sort of vaccine might offer protection or how the infection process recursively influences the population in which the infection takes place. The ontology of emergent life thus constitutes a ‘general crisis environment’ in which potential pandemics loom large as highly unpredictable threats (Massumi, 2009: 154).

Simulation technologies are supposed to explore this environment whose complexity increases with intensified global interconnectivity (Heesterbeek et al., 2015). They are a knowledge technology designed to tackle the unknown. Yet they do so not by simply repealing the problem of radical uncertainty which is irreducible to emerging infectious diseases. Simulations rather contain ‘conjectural’ features in addressing the emergence of unpredictable threats (Aradau and van Munster, 2011: 45). They retain degrees of uncertainty to accommodate the highly uncertain event of a potential pandemic.

However, this requirement to operate under the premise of a limited capacity to know the pandemics to come does not fit well with influential accounts of computer simulation. Especially in the philosophy of science we find the view that simulations are mimetic representations. They are supposed to correlate with an outer reality that they are assumed to reproduce. According to Paul Humphreys, who is a key figure in this field of research, a simulation realizes an external process by computational means: ‘System S provides a core simulation of…process B just in case A is a concrete computational device that produces, via temporal process, solutions to a computational model that correctly represents B…’ (Humphreys, 2007: 110, emphasis added). Such a view, however, is inadequate when it comes to pandemic simulations. It not only overburdens the expectations of what pandemic simulations are able to do: measured against the standard of a ‘correct representation’, simulations can only get it wrong when faced with the problématique of emerging infectious disease. Most notably, Humphreys’ account obscures what pandemic simulations actually put on display and how they are practically used. It is more appropriate to understand them as non-representational devices in the sense that they do not aim to reproduce or imitate an outer reality. ⁷

In order to further substantiate this claim, it suffices to look at the very basic epidemiological processes to be accommodated in a simulation. Although generalizations are hard to make, given the different types of epidemiological models, most of them differentiate the population into three groups: (1) those susceptible; (2) those infected; and (3) those who have recovered. ⁸ With regard to these three categories, epidemiologists are primarily interested in the rate with which people move from one state to another. A first key factor for determining this rate is the number of contacts each person has. What counts as a contact is always related to a particular disease: the term refers to the peculiar form of interaction able to cause an infection. A second factor is the probability that a contact is infected. This probability is not distributed evenly in a population but may vary depending on the social milieu. A third factor is the probability of infection per contact. It not only changes with the number of those infected, it depends on the degree of susceptibility which may be influenced by gender, age, income, cultural habits or even the climate. These factors help to establish the basic reproductive ratio R ₀ (Heesterbeek, 2002). R ₀ indicates the number of individuals infected by one infected individual in a population without immunity.

Even such a rudimentary sketch of the most elementary epidemiological mechanisms to be incorporated by simulation models generates a sense of their non-representational quality. With regard to the aforementioned factors, pandemic simulations have to work with estimates if they are supposed to provide information about pathogens whose properties are uncertain. Especially at an early stage of a pandemic, data is not very robust, if there is sufficient data at all. Due to the lack of incidence data, R ₀ in particular can only be quantified after infection patterns have unfolded (Hefferman et al., 2005: 291). This is the reason why simulations are often based on past disease incidents even if they are supposed to generate insights about a coming pandemic (Mansnerus, 2013: 282). Modelling becomes all the more complicated when it has to include hypothetical scenarios, for instance, about how the health infrastructure operates under stress (Timpka et al., 2009: 206–7) or about how affective turbulences within a population impact upon contact patterns. Such scenarios are evaluated by experts whether they are plausible or not. Subsequently, qualitative data has to be translated into quantitative data.

Given these practical challenges, it amounts to a matter of epistemological realism to see computer simulation marked by ‘hypothetical and heuristic aspects’ and a ‘general inexactitude’ (Pias, 2011: 32; see also Hörl, 2008). According to Claus Pias, who is the Director of the Research Centre on Media Cultures of Computer Simulation (MECS) at Leuphana University (Lüneburg, Germany), in computer simulation ‘performance dissociates from accuracy’ (Pias, 2011: 34) and ‘adequacy replaces proofs’ (p. 35). Pias’ assessment thus corroborates the view that the knowledge produced through the simulation of pandemics is characterized not by correlation but is constitutively infused with approximations, estimations and speculations.

In order to capture this non-correlationist quality, one might adapt a concept put forward by Niklas Luhmann: what pandemic simulations enact is neither something real nor something unreal, it is rather a ‘doubling of reality’ [Realitätsverdopplung] (Luhmann, 2002: 58). In the factual dimension simulations only appear lofty, imprecise or fictitious when conceived of as mimetic representations. The concept of the ‘doubling of reality’ considers them instead as another layer of reality, a layer that might inform or irritate other layers of reality. Under this epistemological premise, the reality of the simulation does not have to converge with the simulated reality to be fully real. Rather, simulations make reality more complex by introducing a new realm of reality within reality. They create second worlds within the world that have to be taken into account. To be sure, technologies for doubling reality are not confined to the realm of computer simulation but have existed throughout modernity. Elena Esposito (2007) identifies both literary novels and probability calculations as key sites in this respect. However, it has been argued that the doubling of reality gains a new quality with digital technologies. Evelyn Ruppert, John Law and Mike Savage (2013: 28) emphasize this point: ‘The digital actualizes relations and connections that are otherwise beyond perception and thus inherent to the very imagining of social relations.’ As digital devices, pandemic simulations do not subtract something from reality by means of virtual representation; they enhance reality by providing an additional access to relations and connections.

It is important to note that such an understanding differs from the probably best-known concept of simulation to be found in the social sciences: the one put forward by Jean Baudrillard (1994). Baudrillard is not so much concerned with specific simulation technologies but with the state of postmodern culture as a whole. His main diagnosis revolves around the loss of reality. In his view, the ‘era of simulation is inaugurated by the liquidation of all referentials’ (1994: 2). The ‘simulacral regimes’ are seen as ways of replacing the world, the reality principle erodes under the postmodern condition. This diagnosis stands in contrast to the argument developed so far. According to the analysis presented in this article, pandemic simulations do not replace the world but inform it; they do not ‘liquidate all referentials’ but fashion a particular referential structure; they do not engender a lack of reality but rather introduce a supplementary stratum of reality. Instead of suspending reality by substituting the real for its model, they enrich reality and increase it in its density.

Epstein and Axtell (1996) have found an apposite metaphor for such a fabrication of second worlds within the world that are staged for the purpose of adding a relation to the world. With their agent-based models Epstein and Axtell purport to ‘grow’ artificial societies. As already hinted at in the introductory section, simulations contain assumptions about actors and rationalities of interaction that are implemented digitally in order to ‘run’ social processes. They ‘grow’ artificial societies that are animated quasi-empirically and put on display by means of graphical visualization. In this way, society can reflect on itself in the simulation of contagious contacts. The particular kind of vision established through such digital observatory is highlighted by the developers of a computational model of smallpox transmission. The group of authors, which also includes Joshua Epstein, aims to analyse the epidemic impact of a bioterrorist attack with anthrax by modelling local contact dynamics:

All of this action can be depicted graphically in real time, as if looking down on the social space from above, watching agents move to and from various social units…changing colors as they progress through the phases of the disease. (Burke et al., 2006: 1142)

From communication to communicability: the materiality of being-with

A sociologist deeply rooted in the tradition of the discipline might, indeed, raise the objection that pandemic simulations primarily show a biological reality. After all, they are about viruses, bacteria and infection. In opposition to this view stands the self-description of agent-based models in terms of ‘social simulations’ (Gilbert, 2004). Famous simulations such as EpiSims and EpiCast, both of which have been developed at the Los Alamos National Laboratories (Barrett et al., 2005; Mniszewski et al., 2008), use census data to generate synthetic populations in which each actor is modelled individually in his or her local behaviour. Every actor is constituted by a set of simple rules that inform his or her actions. For instance, if a ‘person’ is a bus driver, a schoolchild or a teacher, then she has a number of specific contacts during the day. ⁹ Based on factors such as age, income, occupational status, size of household, means of mobility (especially car ownership) or location, a complex distribution of contacts is modelled in order to observe how interactions affect and become affected by the dynamics of contagion (Del Val, 2016). In this way, agent-based models dissolve the social collective into a multiplicity of calculated and calculating agents (Vehlken and Pias, 2014: 168). They are a ‘media technology that offers insights into the fine granulate of societal coherency’ (p. 175).

Yet by referring to the interactional ontology at the core of agent-based models, not much is said about the particular form in which society becomes visible to itself. To determine the specific worldview provided by pandemic simulations, one has to inquire more deeply into the aesthetic experience of the social that they engender. A suitable subject matter for such analysis is the simulation developed by Dirk Brockmann (of the Robert Koch Institute and HU Berlin) and Dirk Helbing (of ETH Zürich). In a collaborative work, they simulate the spread of pandemic disease by linking epidemiological processes with the rhythms of airline mobility (Brockmann and Helbing, 2013) (see Figure 1). The decisive ‘connectors’ are the infrastructures of global traffic that co-constitute the directionality and the temporality of contagion: ‘Our global connectivity has reshaped the way that spreading phenomena evolve on networks that connect us.’ ¹⁰

OPEN IN VIEWER

Figure 1.

Dirk Brockmann and Dirk Helbing: a geometric approach to network-driven contagion phenomena.

With regard to the social dimension of this peculiar form of global self-observation, the motive of pandemic contagion dramatizes a specific form of global co-existence. The simulation depicts a world of dense linkages that do not respect any boundary between inside and outside. In a certain sense, the transmission of pathogens through the system of world traffic pictures a new universality of ‘being-with’ (Nancy, 2000: 28–47). Although the mode of being-with enacted by pandemic simulations hardly meets the communal expectations nourished by the philosopher Jean-Luc Nancy, who introduced the notion to translate Heidegger’s Being and Time into a decidedly social ontology (see Critchley, 1999), the scene of contagious co-existence nonetheless stages a cosmopolitical world society – one, however, that is born out of the collective exposure to a catastrophic incident (Beck and Levy, 2013). ¹¹ Contagion connects globally through physical contacts generated through entities in movement. The only nation that appears in this context is, so to speak, the concatenation of living things, mediated by traffic infrastructures. At the same time, the simulation signals that life – especially life in dense aggregates – is inherently threatened. It is threatened not by an external enemy, but by its immanent potential for disease emergence which appears as red dust (see Figure 2). In his book entitled In the Dust of This Planet, Eugene Thacker (2011: 35) aptly characterizes such contagious phenomena: ‘There is only the strange, immanent and fully distributed “life force”…not being a discrete thing itself.’ Supported by the logistical apparatus of the global mobility regime, the world-spanning associations emerging via computer simulation constitute a truly trans-individual collective.

OPEN IN VIEWER

Figure 2.

‘In the Dust of This Planet’ - GleamViz.

Even if this world picture corresponds in some respects to sociological descriptions of globalization (Held et al., 1999: 16), one should not ignore its particularities. Taking seriously the premise that pandemic simulations are tools for societal self-observation, it is worthwhile reconstructing how they conceive of a social act. In this regard, they point towards a form of global connectivity that does not so much belong to the realm of communicative action but which portrays the basic social operation as communicability. The connotations of communicability re-activate semantic elements of a pre-modern notion of communication. The old communicatio signifies the alliance, the exchange, the commerce and – especially in the Christian tradition – the act of sharing that is constitutive of a community. The communicability enacted by digital means retains the materiality of these processes. In particular, Brockmann and Helbing’s simulation forces its observers to acknowledge how global connectivity is orchestrated by an infrastructural assemblage composed of heterogeneous entities. It fosters pathogenic ties in a sharing economy of a different kind. ‘Communication becomes contamination; transmission becomes contagion’ – what Jean-Luc Nancy (2015: 34) formulates as a cultural diagnostic in the context of his philosophy of being-with becomes the general frame for observing world societal processes.

Of course, this shift towards communicability may not come as a great surprise given that the subject matter of simulation is infectious disease. Yet it has far-reaching implications as soon as the pandemic simulation is analysed as an animated social theory. Most notably, grasping social interconnectedness in terms of communicability marks a radical break with the tradition of interpretative sociology. In a lineage starting with Max Weber, social operations are comprehended as meaningful relations. Pandemic simulations, in contrast, highlight the event of material transaction; their primal scene is not one of mutual understanding but one of metabolic circulation. They present an ecology of the social characterized by close ties between different forms of vibrant matter, both organic and inorganic. This vision corresponds to the ‘emerging infectious disease worldview’ (King, 2002: 767), according to which planetary life exists enfolded into a multiplicity of processes: the virulence of disease is deeply affected by trade lanes, travel routes, technological changes, human-animal relations or climatic conditions. In a similar vein, the ‘old social’ overlaps with the biological and the technological in pandemic simulations – in fact, these distinctions do not seem to take a leading part at all and instead are blurred.

This vision of the social resonates with some developments associated with the current turn towards a ‘new materialism’ (Coole and Frost, 2010). In particular, scholars working in the field of media studies have made the compelling case that our contemporary networked culture exhibits epidemiological dynamics. Authors such as Jussi Parikka (2007) or Tony D. Sampson (2012) reach the far-reaching conclusion of adapting the notion of contagion as the basic social category. Against this background, it is interesting to see that computational models of infectious disease are also adopted to simulate phenomena as diverse as fashion trends, financial transactions or the spread of emotions. Contagion thus conceptually spans across semiotic, affective and somatic registers, turning into a truly transversal category. Furthermore, the ontology of connectivity shimmering through pandemic models shares features with Bruno Latour’s sociology of associations. Society loses its contours under both the conceptual premise of a new ‘associology’ (Latour, 2005: 9) and under the algorithmic premise of modelling global contagion. Just as in new materialist approaches, in pandemic simulation it becomes impossible to distinguish the social from the non-social.

To avoid such a dissolution of a distinctively bounded social, one can stick to the conceptual perspective advanced in this article. The conclusion with regard to the social dimension then reads as follows: through the view on universal communicability generated by pandemic simulations, current world society observes its multiple couplings with processes operating in its environment. This sociological qualification is not necessarily tantamount to rejecting communicability as a mode of operation tout court. It rather grants the possibility of reflecting on how simulations let the world appear within the world. The world picture enacted through pandemic simulations forms a particular horizon of world societal operations: pandemic simulations generate a communicable knowledge about a world of heterogeneous communicabilities.

From flows to jumps: the world as globe and topology

The global ecology of communicability excavated in the last section has already touched upon the question of how pandemic simulations configure the world spatially. To deepen the analysis in the spatial dimension, the simulation platform GleamViz is a good place to start. GleamViz is the result of a research project that has been funded by the EU and the US Department of Health inter alia. In the Global Epidemic and Mobility Model (GLEaM) at the core of the simulation, ‘the world is defined in geographical census areas connected in a network of interactions by human travel fluxes corresponding to transportation and mobility patterns’ (Balcan et al., 2010: 142). The visualization (Figure 3) shows the world as a perfectly rounded globe. It perpetuates the ‘graphical and technical play with totality and its image’ that Peter Sloterdijk (1999: 45) sees as a defining feature of the occidental metaphysics of wholeness and which lingers on in our current notions of globalization. It does so by perpetuating the modern topos of what Luhmann (2005: 67) calls a ‘seamless civilization of traffic’ [durchgängige Verkehrszivilisation].

OPEN IN VIEWER

Figure 3.

The perfectly rounded globe - GleamViz.

The spatial figure of a closed planetary surface on which the world traffic moves freely correlates with epidemiological accounts of pathogenic pathways (Opitz, 2016). The virologist Stephen S. Morse (1992), one of the key figures in the debate on emerging infectious diseases, describes them in terms of ‘global microbial traffic’. In this view of the world, no location can make itself inaccessible to any other location. ‘Vulnerability is universal’– Margaret Chan’s famous opening words in the WHO (2007) report, A Safer Future, translate the spatial implications of such viral globalism into the vocabulary of global health insecurity. The vision of a space criss-crossed all over by the flow of vital traffic is, without any doubt, the ultimate instance of ‘a shift in reference from place to movement’, a shift that represents, for Luhmann (2012: 88), the primary spatial feature of our contemporary world society. At the same time, the world picture generated by the simulation modifies the notion of a ‘space of flows’ (Castells, 2000: 440–59), which has had a huge impact on the sociology of globalization. While the image of the flow remains on the global surface, the Earth-spanning operation of communicability rather appears to be a jump through air space. Whereas the image of the flow implies a movement that is ‘globe-covering’, the world motion picture of GleamViz circumscribes a mode of parabolic ‘globe-hopping’ that skips the space in-between. ¹²

However, pandemic simulations also exhibit another spatial motive that is at least as iconic as the globe: the network. Figure 4, borrowed again from the collaborative work by Brockmann and Helbing, shows exactly the same mobility regime of pathogenic air traffic as Figure 1. In fact, it is based on the same model. Brockmann and Helbing have merely changed the mode of visualizing the data. Now, the network structure is neither moulded into the geography of the world map nor tied around a perfectly rounded globe. It rather takes the shape of a ‘shortest path tree network’ (Brockmann and Helbing, 2013: 1339). Most importantly, the edges of the network are now measured according to the time that a body needs to move from one node to another: ‘Our approach is based on the idea of replacing conventional geographical distance by a measure of effective distance’ (p. 1337). Simply put, London would be significantly closer to Hamburg than Groningen. Although the geographical distance from Hamburg to Groningen is only one-third compared to the distance from Hamburg to London, due to the spatio-temporal characteristics of transportation networks Groningen is twice as far removed from Hamburg when measured in ‘effective distance’.

OPEN IN VIEWER

Figure 4.

Pandemic topology by Dirk Brockmann and Dirk Helbing.

Determining this shift in spatial rationality in more conceptual terms, one can say that the extensive topography of the global is being translated into a global topology. Under the influence of Michel Serres’ work in particular, the notion of topology is being used in the social sciences to reach beyond a Euclidean concept of space (Harris, 1997; Martin and Secor, 2014). The aim is to capture spatial formations not by means of a metrical system that is external to them but according to the immanent rules of association and transformation. In Serres’ own words, the ‘first problem’ is

to find the single space or the set of operators by which these spatial varieties in impractical, inconceivable vicinity will be joined together. To open the route, way, track, path in this incoherent chaos, this tattered cloud, whose dichotomic thicket is reformulated in the common space of transport when it is reconstructed. (1982: 51)

This spatial insight cannot be underestimated. It suggests that it becomes increasingly difficult to capture geographically the vicinities and adjacencies in our current world. We live in a world, in which what is geographically nearby can be far removed and, conversely, in which the geographically distant can be close. Yet perhaps topographical and topological spaces should not be dealt with in an exclusive manner. Looking at the world pictured by pandemic simulations, we rather seem to oscillate between the image of a seamless globe circumscribing the totality of a complete world interior, on the one hand, and a topologically fractured world, on the other.

From prediction to premediation: a virtual spectrum of present futures

The analysis so far has already pointed towards the inextricable link between space and time established by pandemic simulations. On the most basic level, pandemic simulations take the form of ‘temporal objects’ that are characterized by a sequential chain of relational events. ¹³ In order to calculate processes of contagion, they reckon with the rhythms of mobility, the frequency of bodily interactions or the speed in which health provisions are made, and relate them to the features of a particular disease. This amalgam of bio-social temporalities lies at the heart of all epidemiological knowledge that always had to link incubation and infection periods with contact patterns, journey times or the impact of administrative measures, for instance, decisions on quarantine (Harrison, 2013). Simulations assemble these temporalities on the basis of process models and visualize them as diagrams or, preferably, as movement images.

However, with regard to the temporal dimension, the analysis has to move one step further since the simulated time frame stands itself in a temporal relation. Pandemic simulations perform temporalized processes of contagious dynamics that refer to something that is not actual: they are constitutively oriented towards the future. The research team working with EpiSims at Los Alamos Laboratories, for instance, opens its investigation of the different effects administrative measures have on the spread of influenza with the claim that ‘our first defense against pandemic flu is to see it coming’ (Mniszewski et al., 2008: 556). In a similar vein, the developers of GleamViz characterize their simulation model as a technology of risk management. In both cases, the aim is to allow for some kind of anticipatory stance towards disease threats expected in the future. Hence, pandemic simulations are involved in a complex exploration of time. As animated theories, they are tools for actual experiments with non-actual processes.

In order to understand their temporal mode of operation properly, it is important to identify the particular form of anticipation enacted by pandemic simulations. It already follows from the non-representational logic at work in the factual dimension that they do not have the quality of a prediction. A prediction seeks to foretell a future course of events with some accuracy. If it fails to do so, the prediction would be considered erroneous. A paradigmatic example is the weather report: within a relatively narrow range of tolerance, one expects the weather on the afternoon to be as it was forecast in the morning. The heuristic knowledge of approximation that pandemic simulations offer does not function accordingly. This is made explicit in the literature, especially where policy implications are debated. As a Swedish research team put it in an article for the Bulletin of the WHO: ‘The predictive validity of simulations at early stages of a pandemic will be inevitably poor’ (Timpka et al., 2009: 307). It would be a mistake, however, to infer from such a statement that simulations are simply inaccurate or deficient, only because they appear so when measured against the standard of prediction. They embody a different temporal logic vis-à-vis the future. Instead of giving an account of a dated future incident, they chart a series of futures.

A concept introduced by media theorist Richard Grusin helps in specifying this distinctive relation towards the future: as opposed to apparatuses of prediction, pandemic simulations are technologies of ‘premediation’ (Grusin, 2010; see also de Goede, 2008). They function as future-invocative media designed to fathom the potentiality of the future. Premediation does not strive to determine the ‘future present’; it rather operates by multiplying ‘present futures’ (Luhmann, 1976; Opitz and Tellmann, 2015). For example, in the presentation of the GLEaM model analysed above, the authors praise ‘the possibility of generating large numbers of in silico epidemics’ (Balcan et al., 2010: 133). These in silico epidemics are used to estimate the parameters of the model with the aim that the simulation eventually will serve as a ‘risk assessment tool for scenario evaluations of an epidemic emergency’ (Balcan et al., 2010). The simulation is thus supposed to operate as a supplement to scenario planning. Given that the precise event of an emerging pandemic remains irreducibly unknown, the simulation is employed to play through the eventualities of an emergency. This is why simulations are aptly characterized as preparedness technologies (Lakoff, 2008). Bluntly put, the goal is to avoid surprises by creating and testing a series of possible futures. Processing versions of the future generates a sense of what the future might have in store, it renders the unknown familiar and permits acting out different responses to what might happen. Simulations are therefore a means of transforming indeterminate contingency into determinate contingency, paradoxically by generating multiple contingencies: by running scenarios on the basis of their model assumptions, they assemble a portfolio of future courses that are neither necessary nor impossible (Dillon, 2007).

Taking pandemic simulations seriously as animated social theories, one might in fact ponder whether this extraordinary mode of future orientation points to the very limit of Luhmann’s notion of meaning, i.e. to the limit of the concept that has organized the analysis so far. For Luhmann, each meaningful reference, irrespective of whether it occurs in the factual, social, spatial or temporal relation, is characterized by a referential excess: a ‘surplus reference accessible from actually given meaning’ (2012: 21). In the moment of being actualized, meaningful acts always refer to other possibilities, which can be actualized subsequently while, again, referring always already to other possibilities, and so on. Yet due to their temporal characteristics, pandemic simulations reach beyond the distinction between the actual and the possible. Of course, using a paradoxical phrasing, one might say that they render visible actual possibilities of disease: simulations actualize possible futures. Although this formulation is not fully inaccurate, it is at least problematic in two respects.

Adapting an argument presented by Gilles Deleuze (1988: 96–103), the possible is usually distinguished from the real. Something that is possible is not in itself real but can only become real. Furthermore, the ‘rules of actualization’ (1988: 97) are governed by a logic of resemblance. Actualizing the possible is tantamount to merely adding reality to it. The notion of a doubling of reality introduced above and the temporal logic of premediating future contingency run against those two presuppositions. Simulations are more aptly described if one follows Deleuze in distinguishing the actual from the virtual. This distinction holds the advantage of not having to decide between reality and non-reality. The simulation of a disease is virtual in the sense that it can be highly real without being actual. In fact, as a particular premediation of a global threat, virtual disease can be more intensely real than the actual disease. Moreover, the virtual does not stand in a relation of sameness to the actual but is characterized by divergence and difference. Simulations of infectious diseases create a virtual spectrum of futurity that is supposed to initiate a play between different orders of reality for the purpose of being prepared for the next pandemic.

Conclusion: global self-observation in times of catastrophe

Right at the beginning of his theory of society, Luhmann delineates a dizzying circularity. Since no theory of society exists outside society, every theory of society always already performs what it describes. But if each theory of society is necessarily part of the subject matter that it reflects, sociology would have to reflect on society as a self-reflecting entity. According to Luhmann, such ‘autology’ is inescapable. In his words:

The contributions of sociological research to the self-description of society become a topic of sociological theory and a problem for its logic and its methodology, or in other words, the re-entry of the observer into the observed re-enters the observer. (1997: 77)

This article has undertaken a sociological reflection of pandemic simulations as modes of world societal self-observation. It has aimed at pushing this exercise in each of the four sub-sections up to a point at which the analysis may feed back into sociological theory. Taking pandemic simulations seriously as animated social theories, the following characteristics were elaborated: first, simulations provide not a mimetic representation of the real world but a doubling of reality; they add worlds to the world. Second, simulations depict the social in terms of a universal materiality of being-with; the primary mode of operation is not communication but global communicability. Third, they supplement the figure of the spherical globe with a topological network; pandemic simulations enact a non-Euclidean world in which what is geographically distant can be close and what is geographically close can be distant. Fourth, they do not function as a means of prediction but as a technology of premediation; they produce a series of present futures in order to map the space of contingency. In the four dimensions of meaning, pandemic simulations thus change how we situate ourselves in the world factually, socially, spatially and temporally.

It is worth recalling that, according to Heidegger, the world picture transforms the modern world into something ‘at hand’ by rendering it knowable. Correspondingly, computer simulation must be understood as an ‘epistotechnics’ of the present that allows for a ‘re-addressing of world’ (Hörl, 2008: 103). Specific to pandemic simulations, in this respect, is their catastrophic diagrammatic (Opitz and Tellmann, 2015). They provide an observational form by means of which our current world society confronts itself with the possibility of what biologists have termed a ‘viral storm’ (Wolfe, 2012). Such mass contagion appears as a phenomenon of societal self-endangerment, since it unfolds through operations of global connectivity.

In a way, this observational form responds to a diagnostic remark Luhmann made two decades ago. After having dismissed all modern formulas – ‘the leftover vocabularies of the tradition’ (Luhmann, 1997: 69) – as inadequate to describe our current world society, he merely designates what he considers the problem to be solved: ‘Our problem is to mark off a space in which we can observe the emergence of order and disorder’ (Luhmann, 1997). The analysis provided in this article supports the conclusion that pandemic simulations operate in the problem area demarcated by Luhmann. They mark off a space in which the (dis)order of emergence itself can be observed. Digitally shaping the prospect of a potentially catastrophic future, they relate us to the overwhelming relationality of the world: pandemic simulations establish a synthetic relation to the multitude of socio-material relations involved in global disease transmission.

Further research would therefore have to venture deeper into the pragmatics of pandemic simulations. This is without any doubt a challenging task due to the complexities of modelling and the high level of expertise involved. It is no coincidence that the social science literature on simulation has mostly followed the relatively broad cultural diagnosis put forward by Baudrillard (e.g. de Lint et al., 2007; Bogard, 2012), thereby circumventing the intricacies of the technicalities involved. There are only a few sociologists interested in simulation technologies as an empirical phenomenon (Hoffmann, 2007; 2014). Those scholars working in the field of risk and precaution refer to simulation methods only in passing (e.g. Collier, 2009; Aradau and van Munster, 2011; Amoore, 2013), without focusing their analytical gaze on the phenomenon itself. Accordingly, the literature on pandemic simulation is relatively scarce, notwithstanding some important contributions (Pias, 2011; Gramelsberger and Mansnerus, 2012; Davis et al., 2017). In particular, a deeper understanding of the actual use of simulation methods in the contemporary regime of global health security would be desirable. How are they employed as technologies of preparedness? How exactly does the view generated of the contingencies of infectious disease inform administrative measures? How do they, for instance, help to determine emergency plans with regard to traffic restrictions, vaccination programmes or quarantine policies?

This article does not answer these questions, but it might hopefully help to better pose them. Its analysis ultimately points to something akin to a cybernetic mechanism that pandemic simulations set in motion: on the one hand, simulation technologies produce an outlook on contagious dynamics that informs the administration of socio-material connectivities; on the other, the data collected from actual disease incidents feeds back into simulation models in order to improve them. It would be interesting for sociologists to find out more about the work of such socio-technical assemblages and their recursive effects in contemporary administration. Hopefully, the reflection on computational modes of societal self-reflection undertaken in the current article can motivate or even be conducive to such research. The sociological research on technologies of pandemic simulation is still in its infancy.