Governance by Medical Models: Eight Semiotic Principles

By Gary Genosko

Governance by medical modelling is the attempt to make reality conform to predictions and projections. There is a political imperative in now familiar exhortations, that require broad legislative and regulatory changes that modify social, political and economic relationships in the name of furthering this forced conformity. Modellers and their tools are emerging from their hitherto hidden side of governance, the very persons and tools whose work can only be glimpsed in passing, behind the veil of public statements, exhortations, and further reinforcements of distancing, restrictions, closures, permissible congregation sizes, cancellations and delays.

Pressure has built to release COVID-19 models to the public. But once these models are released, how are they understood? The early results show that rigorous mitigation efforts supported by robust social control measures seem, on the one hand, to push lower negative outcomes toward a best-case scenario of minimal deaths and flattening infection transmission rates; whereas, on the other hand, no mitigation efforts, and no or unenforceable social controls, will result in the worst-case scenario, uncontrolled infection rates and a spiralling death count. The continuum between these two options is wide. The amount of real estate between the two is itself daunting because it is so poorly defined.

What is a model, anyway? The following eight principles constitute a semiotic primer for modelling.

The activity of modelling is today studied in terms of a fairly stable set of parameters. Typically, these include: simplification – a model is always less than that which it models while marking an indirect route to the phenomenon in question; some parts are emphasized and others are downplayed or ignored. Too much simplicity detracts from the modelling process and thus leads to a failure to grasp the complexity of the phenomenon at issue; whereas a highly sophisticated and advanced model may be uncritically adopted and misapplied. Simplicity works well with parsimony as plausibility may be achieved through a spare and compact design.
Pertinence speaks to the relevance of a model’s components. The evaluation of these components needs to take place in relation to the theories and facts that informed the model’s construction. The evaluation of a model’s pertinence returns one to the question of whose needs it meets – those of the modeller, of her political masters, whose intensions and foci it serves; those of the user, maybe the decision-making citizen, into whose problem solving and risk-taking tasks it provides insight.
Visualization of data, primarily by means of computer graphics (from old school mesh plots to 4D animation), is an important issue regardless of whether a model is abstract or physical, as it can be put together clunkily or elegantly. Close attention needs to be paid to the choice of colouration, shapes, pattern, shading, and scale in image building. Efforts at conquering the unseen need to be grasped in the context of their audiences, professional, popular, or both. This had already become more significant in the cases of virus scares like H1N1 with its posters full of lurid colour washes.
Models offer explanations of phenomena. They operate beyond mere description and attempt to account for something in relation to its causes or the laws governing it. Of course, models may find out something new, or help discover something whose existence has been hitherto, hypothetical. Even false models can still explain something by providing a valuable starting point. Models are often considered to be heuristic devices that aid a general sense of discovery. They are, however, of varying degrees of interpretively tightness and looseness.
Modelling is a theoretical activity that involves a fairly wide range of relations with phenomena; conversely, models instantiate theories. This activity is central to scientific practice and theories are both resources for model construction and theory development. But scientific practice is a handmaiden at this moment of politics in which theory is replaced by hasty policy formation in a time of crisis.
Modelling has moved beyond its traditional domain of physical construction with springs and wire into that of the use of computational simulations which is now canonical, but which emerged unevenly. Simulations are useful for real world interventions, valuable in vocational training, and as instruments for real world actions. This is what economists tells us. While the histories of some sciences like meteorology are marked by a split between early predictive and later explanatory models, there is a shared sense that models do more than represent, rather, they theoretically and empirically reconstruct the phenomenon at issue in subtle ways across the levels of hypotheses, measurements, and uses of prediction.
Models contain many borrowed parts. Borrowing across disciplines (linguistics and communication once borrowed heavily from physics) is commonplace. Borrowed parts may be replaced, adjusted, calibrated. There is tolerance of imperfection, misfit, falsity, incompleteness and idealization in modelling.
Multiple models can coexist in a given scientific endeavor because of a multiplicity of purposes, as in El Nino research and all climate predictions. There is no best or perfect model for a hypercomplex phenomenon. Yet models lacking pertinence for a given task can in fact persist, not because they are especially useful, but simply because they enjoy the legitimacy of having been developed in the first place. They are at on hand. Simple existence and availability are sometimes a matter of convenience. The best example I can think of is the one that impinges upon us today, WO Kermack and AG McKendrick’s 1927 mathematical model of epidemics, the so-called S-I-R model https://royalsocietypublishing.org/doi/10.1098/rspa.1927.0118. This tripartite model describes a world of three populations: susceptible; infected; recovered. It has a set of assumptions (equal susceptibility in a homogeneous population; a single attack grants immunity; limited moment of infection transmissibility; the role of population density as a threshold beyond which infection rate becomes much greater). No attention was paid to the capacity of hospitals to treat infections. Thus, subsequent elaborations and refinements of the SIR model introduced factors like population heterogeneity, availability of ICU beds, while downplaying the ideological willingness to govern by means of medical models that effectively translate into changing the fundamentals of social and economic relationships so that they address the predictions.

Why is a critical grasp of modelling processes so important? The creation and application of models is simultaneously the creation of new social and political subjectivities. Since reality is being made to conform to models by placing normalizing limits on social congregation, imposing standard identities in the name of quarantine and isolation, and justifying the criminalization of a wide variety of behaviours, critical discussion of all models remains vital to everyone. Knowing how models are built, tinkered with, fall apart, and still, somehow succeed, despite all of their potential failings, and the political desire to make the world conform to what they suggest, is a most pressing question for everyone.

Yet one part bewilderment and another part disappointment will likely be the result of the release of COVID-19 models by governments around the world. The release of data generated by means of epidemiological models may not, in fact, shine much light on the models themselves. Everyone should now know that models are not crystal balls, as Canada’s Chief Public Medical Officer Dr. Theresa Tam said recently. This is helpful if we want to know what models are not. The complex mathematics upon which the wildly divergent predictions are made are challenging and has not been discussed in public; the programming code is not likely to see the light of day anytime soon due to proprietary interests.

When these models are discussed, then, one of the best outcomes will be an increase in model literacy, a leap in the formulation of probing questions – what was the effect of updating the model on the government’s actions? How many assumptions have thus far been proved wrong? Remember that models are dynamic simulations that rely on data that is hard to get during a crisis, and arrives in fits and starts, and leaps of faith are regularly made about the data that is available.

For skilled modellers and those tasked with communicating to the public about them, it may be useful to remember that many modellers do not construct their own models. Instead, they use either standard or even off-the-shelf models. Identifying the appropriate model is more important here than explicitly constructing a new model that would break new pathways. Building in a few modifications, if possible, is the norm, when managing a standard model. Modified and updated versions of SIRs can be tagged, but they are embedded in a tradition of modelling, with all its warts and bristles intact (i.e., the Imperial College model, https://www.nature.com/articles/d41586-020-01003-6).

Models are dynamic tools that express a power to inform and misinform political decision-making. Epidemic models have been rushed into service and run the risk of pushing forward with new policies and regulations that change the way we live that seem just as speculative and uneven as the models themselves. Making social, political and economic reality conform to models is simply to acknowledge that government sits unenviably, twisting its neck, in-between a rapidly changing pandemic and theoretical models that only hint at the future to come.

The salient question remains: can COVID-19 be modelled with an inherited, patched together set of ideas from the 1920s, adapted on the fly out of necessity to the fast-developing, yet uneven, global situation?

Gary Genosko

Professor of Communication and Digital Media, OnTechU.