The Politics of Simulation

I’ve been quiet for a few weeks now because I’ve been diligently studying for my second area exam.  This one will be about the research that has been done within the social sciences on environmental modeling.  Most of this research has focused on the General Circulation Models (GCMs) that are used to understand and predict the effects of anthropogenic climate change, however, there are those who deal with water quality modeling and flood modeling.  What I’ve found interesting in these accounts is the extent to which – at times unacknowledged by the researchers themselves – politics plays a formative role in the development and distribution of computer simulations.

As an example, I’ll take Matthias Heymann’s essay “Constructing Evidence and Trust” in the volume The Social Life of Climate Change Models (edited by Kirsten Hastrup and Martin Skrydstrup) which looks at the different ways that models gain confidence within the scientific community (the author purposefully brackets off scientific confidence from public confidence in order to avoid confusion).  In the article, Heymann identifies four sources – I would almost say orders – of confidence. The first order of confidence comes from the emergent features that are produced in the simulation runs.  That these emergent features resemble patterns one would expect to see in the actual climate (but were not part of the initial programming – thus “emergent”), provides a degree of confidence that the models are capturing the basic physical laws that govern the climate.  The second order of confidence comes from the quantitative fit between simulation results and observed data.  That these models can successfully back-forecast and produce reasonably similar results offers a further confidence in their capabilities.  The third order of confidence results from the similarity of behavior between different models.  Because the models tend to agree despite differences in parameters and code suggests that the models are effectively modeling the same general phenomena.  Finally, Heymann argues that the fourth source of confidence comes from a rhetorical strategy that makes the uncertainties in simulation results invisible.  There is no quantitative method for measuring the uncertainty of these models – this would take multiple model runs which consumes a lot of time and resources.  At the same time, model results are depicted with graphs and the weight of statistical certainty.  Although the uncertainty of the models is always mentioned, it cannot compete with the weight of graphic depictions provided by the simulation runs.


This last order of certainty is the only point in Heymann’s four sources where politics seems to enter.  There are clear political reasons (in excess of the practical reasons) why scientists might downplay the uncertainty of their models. One reason is that modelers are competing for funding, and they have a vested interest in portraying their models in the best possible light.  Another reason is that there is a general sense in the scientific community (and I have seen this firsthand in my own encounters with modelers) that any sign of uncertainty will become fodder for climate skeptics and deniers or the media to attack.  A third reason, not mentioned by Heymann, but discussed at length in another article by Lahsen called “Seductive Simulations” is that modelers themselves may be psychologically attached to and overly confident in the reality of their models.

Clearly, politics comes in at the tail end of the modeling performance. Once the models have been run, and all of the other sources of confidence have been achieved, it is the presentation of models that takes on a political aspect.  However, Heymann points to another point where politics enters the equation from the very beginning.  In fact, although he doesn’t recognize it as such, politics could be said to be the real first order source of confidence in modeling.  Heymann describes the history of climate modeling in two phases.  The first phase is the development of basic climate models starting in the 1950s and going into the early 1970s.  This era is characterized by a general lack of confidence in modeling, and frequent warnings in scientific papers about the inadequacy of the model results.  The second phase begins in the 1970s (in A Vast Machine, Edwards marks a similar break occurring around the same time, but for slightly different reasons), and this is the beginning of confidence and the use of climate models to influence public opinion and policy.  This phase begins with the ascension of a new generation of climate modelers (Schneider, Kellog, and Hansen are the most frequently mentioned) who are driven by the urgency of climate change.  They argue in scientific papers that the that urgency of climate change demands that we use the best tools we have to understand the causes and effects as best as we can so that we might be able to intervene rapidly – a first-order political validation for the use of climate models. In other words, we must have confidence in the models because they are the best tools we have and the urgency of the problem demands that we use them.  Heymann provides the following quote from Schneider:

My view is that once we know reasonably well how an individual climatic process works and how it is affected by human activities (e.g. CO2-radiation effect), we are obliged to use our present models to determine whether the changes induced by these human activities could be large enough to be important to society.”

Had this new generation of modelers – coming of age in the era of the environmental movement – not taken such a political stance, the models might still have been used eventually.  However, it is questionable whether they would have developed the required confidence (by way of the other four methods mentioned above) for many more years without the initial impetus and push of confidence provided by this political urgency.  What this means is that politics runs through the practice of modeling from the very beginning, and that, contrary to the general sense that politics must be kept separate from science, in this case political urgency actually provides a degree of initial confidence and a basis for the development of further confidence without which climate modeling might have remained stagnant for a long time.

Friction – Work – Mangle

I’ve been preparing lately for a conference on Monday about the use of multiple models in evaluating water quality on the Chesapeake.  I’m part of a panel that will be discussing the social and cultural implications of multiple modeling, and the gist of what I want to present is that models don’t simply represent or help us understand the complexity of the Bay (or whatever they happen to be modeling), they also add to that complexity (the subtitle of my talk is “Making a Mess with Models” paraphrasing the title of a paper by John Law on methods – pdf).  In order to make this case, I introduce three concepts that I’ve come to see as tied together – Friction, Work, and the Mangle.

Friction – I borrow this concept from two sources: Anna Tsing’s Friction: An Ethnography of Global Connection, and Paul Edward’s A Vast Machine.  Edwards discusses the concept of computational friction and data friction where the problem is integrating different data sets with one another in order to create a more comprehensive data set (e.g. global climate data).  Tsing describes friction as “the awkward, unequal, unstable, and creative qualities of interconnection across difference” and uses this concept to explore the phenomenon of globalization – specifically, the global lumber trade that has shaped the Indonesian landscape.  So, very simply, friction for me is a metaphor for the resistance to interconnection across some difference.  This difference keeps two or more things from connecting, but also offers an almost infinite possibility if the two things are made to connect.

Work – This is a concept I’ve talked a lot about already, but in this context, it’s the process of overcoming the resistance produced by friction.  It’s through work that the possibilities contained within the space of difference are actualized.  In the process of bringing things into relation with one another, a new world is created, and different ways of working to build relationships produce different kinds of worlds.

Mangle – Another borrowed concept – this time from Andrew Pickering.  This is merely to remind us that everything is mangled in the encounter between work and friction.  People are mangled (altered and affected) as much by the technologies, knowledges, organisms and other beings they work with as those beings are mangled by us (the result of the efficacy of beings).  In other words, humans are not engineers of reality – safely reshaping the world from a distance – we are active parts of the world and are continually reshaped by it.

It is my hope that these three concepts, in the context of my presentation at least, will be able to get the modelers, policy makers, and others at the workshop to think differently about their practices.  My goal is to think about the ways that different modeling practices (e.g. participatory modeling, open access models, etc.) can remake the reality of the Chesapeake – not just for the people, but for all of the beings involved.