Modeling is not about the future

Sometimes being an anthropologist surrounded by archaeologists has its benefits. Engaging with them conceptually has certainly helped me to think through the materialisms that I have encountered in my theoretical explorations, and given me a deeper appreciation for infrastructure. But another thing it has done for me is to give me a better understanding of the meaning of the past and memory in our lives. This is something I think I can also apply to models as projections of the future. 

There is an abundance of literature in archaeology on memory and the social construction of the past (see e.g. Shanks and Tilley’s Social Theory in Archaeology). The underlying premise – put very simply – is that the past is not an objective fact, but is constructed and continually reconstructed in the present. In that sense, the role of the past and memory in our lives and social worlds is not to project an image backward, but to compose relations between people and objects here and now. We seek to make sense of the past, and this in many ways shapes the way our world is structured today. The implication for archaeology – the study of the past through material culture – is that, rather than attempting to uncover the “real” past that will tell us how things really were, the purpose of archaeology is to explore new ways of constructing the past in the present. This has resulted in practices like critical archaeology, the archaeology of race and gender, public and collaborative archaeology, etc. 

My research deals only marginally with the past. There is the history of the models, the history of the ecological systems that they are meant to represent, and the histories of the institutions that have developed around modeling and managing those ecologies. These pasts are relevant and important but not the major concerns of my work. However, my work deals significantly with the future – models have become a kind of oracle through which we can project current conditions and processes out into the future. This is a new kind of science. Field and lab research couldn’t tell us what the future would be like, they could only tell us what could be observed in the present. Models bring a dimension of prophecy to the world of science. 

It’s like the Tarot or I Ching – we call to mind some problem, cast the oracle, and the oracle tells us what might come of the future. Obviously, I don’t believe in magic or the transcendent forces that underlie these methods and lend them legitimacy, but I have always been fascinated with practices of divination. I think it was reading Joseph Campbell and subsequently Carl Jung that led me to accept that these practices might have some value even when those transcendent forces were removed. The power of divination lies not in the power of the cards, sticks, stones, tea leaves, etc. to predict the future, but rather in their symbolic ability to redefine the present and encourage us to think differently about it. The media function as catalysts for reinterpreting the current conditions by way of a prediction of the future so that we might become more aware of the various possibilities open to us and then act to manifest those possibilities in our present lives. 

This is the same thing that archaeological theory tells us about the past and memory, so there is in many respects an analogy between the role of the past and the role of the future in our lives. If the past is a construct that enables us to make sense of and restructure the present, then so too is the future, and methods of divination are the processes and practices by which we construct the present by way of the future. 

As I’ve suggested, modeling too is a method of divination or oracle consultation. It differs from the others, of course, in that it is based in empirical observation and mathematical representation rather than symbolic projection and the unconscious, but there are many similarities nevertheless. There is always a dimension of uncertainty around modeling. No matter how well calibrated the model is (i.e. how good it is at representing the past – another way that my research touches on the role of history) future conditions are always fundamentally unpredictable. We can project what we know forward, but there is always the possibility that some unknown factor will come into play and change the whole system making all of our predictions irrelevant. With that in mind, the idea that models can predict the future is just as spurious as the idea that Tarot cards can tell us when we will die or with whom we will fall in love. It may be that the model provides a better approximation of future events than the Tarot, but the uncertainty always remains. 

Following the implications of archaeological theory and the non-transcendent understanding of the value of divination, however, it becomes apparent that predicting the future is not the function of models at all. Modeling is not about the future. Like divination and archaeology, modeling constructs the future in the present, and provides a catalyst for reinterpreting present conditions. This allows us to be more conscious of the various possibilities open to us so that we can act to manifest those possibilities in our present lives. Whether or not a particular possibility will manifest is fundamentally unknowable – there are simply too many factors at play – and this is why we must continually reconsult the oracle. As with the past, the future must be continually reconstructed in the present in order to structure and restructure our world. 

Again, following archaeology, but this time in terms of praxis rather than theory, this raises a lot of implications for how modeling is performed and how the future is constructed. As I mentioned, in archaeology the result has been a shift to critical approaches, as well as collaborative methods. I think the implications are similar for modeling, but first we have to rethink the role of modeling along the same lines as divination. 

Models have to be deprived of their seemingly transcendent authority – as more-or-less accurate representations of a pre-existing and eternal Nature. This is not to say that models shouldn’t be based on the best empirical knowledge we have at the time, but that we should recognize the limitations of that knowledge and understand that the future is fundamentally unknowable. Modeling must come to be recognized as a tool for understanding and restructuring the present by way of a prediction of the future just as divination methods have done and as archaeology has come to define itself in relation to the past. 

Performative Models

One of the most difficult things for me to convey in describing my research is the idea that I am not primarily interested in the representational aspects of modeling. In other words, questions about what the models include, how well those things are represented, the accuracy of the models, and what is excluded – these are not my first questions. I think that the bias for seeing my project this way stems from a tendency to see models as a form of cognition. Models are generally considered extensions of our minds. While I think that’s true, I also think they are much more than that. Models are also technological structures with elaborate architectures, powered by material processes. They require human and non-human labor to produce, maintain, translate, and implement.

Conceptualizing modeling in this way means that I am interested in the performative dimensions of modeling – what models do – more than in what they represent. I am interested in the activities that go into producing the models and using them to decide how to manage environmental problems. From there, I am interested in how those activities produce not just models but also social relationships, institutions, management structures, and so on. 

This doesn’t mean that I am not at all interested in the representational aspects of modeling. I am in the sense that representations are, themselves, performative. The act of deciding how to model certain dynamics, how to best represent the flow of water and the effects of nutrients on the system – these representations have important consequences for the way that water quality is managed. As a result, I would argue that the classic division between the model (or map) and the territory breaks down. The model may not be the territory, but it is certainly part of the territory that it attempts to represent. The question is – rephrasing it again – how do we confront the model as both representation (cognition) and material activity (performance)? How might this view of modeling change the way we engage in modeling? My research so far suggests that there are a lot of factors to consider beyond representational issues, and I’m hoping to develop a more systematic way to confront those issues. 

Hacking Modeling

I am cross-posting this at my primary blog (Struggle Forever!) because it relates to some issues that are of interest to me beyond this project, but the ideas grow out of the work I’m doing for my dissertation research.

It is interesting to me that models can be seen as both a cognition – simulation and representation – and as infrastructure – a system of computational structures embedded within institutional organizations. In other words, the model is not the territory, but becomes part of the territory it seeks to represent, and increasingly so the more influential and widely distributed it becomes. Over at Synthetic_Zero, Edmund Berger describes modeling for environmental management as a form of repurposing of systems designed within and for military-industrial structures for uses that undermine those very structures. In other words, environmental modeling is a kind of “hacking.” It’s an idea I’m interested in and hope to explore to some extent in my research, though there are a number of other angles I’m exploring as well.

Interestingly, Berger suggests that environmental models always, to some extent, represent this kind of hacking, since environmental values are generally opposed to the neoliberal values of the systems they were designed for. Perhaps I’m still skeptical, but I’m finding that models are powerful tools, and that, regardless of the scale and type of environmental model, they have important performative effects that can be considered reformative if not revolutionary. The underlying question of my research – which I’ve already stated, but will restate again and again in many different ways – is under what conditions can we best foster these performative effects in order to promote their revolutionary potential?

I’m not completely convinced yet that all environmental modeling is equal. Some projects seem to reinforce state hierarchies, though it is these hierarchies that are often able to confront neoliberal institutions head-on. Models – especially big, complex ones – make that possible. On the other hand, there are other projects that have the potential to undermine state hierarchies and the division between expert and layperson that underlie them. However, these projects tend to be smaller scale, and rely on smaller, simpler, and potentially less accurate modeling systems. This combination makes these projects potentially less capable of confronting large scale neoliberal interests, at least on the short term. On the long-term, as more of these kinds of projects accumulate and as people come to expect this kind of collaboration in scientific practice, it seems to me that it might be possible to generate a different subjectivity that sees collective action and mutual aid as an effective means of resistance against social neoliberalization.

This is all speculation at the moment, and I’m not able to back it up with any evidence – particularly the long-term speculations. I’m also not opposed to any approach to modeling per se. At this stage, I think it is important to develop a kind of situational awareness of what the obstacles are to environmental protection and restoration in a given case, and what kind of modeling methods might be best suited to addressing those obstacles. I’ve seen little of that situational awareness being fostered – instead the focus is on improving the accuracy and validity of models. I hope my research will shed some light on the relationship between modeling methods and practices and the social relationships in which they take place so that researchers can make more informed decisions about what approach to modeling best suits the conditions of each particular situation.

Identifying Cases

My project will consist of a series of case studies of modeling on the Chesapeake Bay with varying purposes, institutional circumstances, and degrees of participation from stakeholders. Drawing on these case studies, I will do a comparative analysis to understand the different performances of modeling and get a sense of the potential outcomes that can be expected from them. At the moment, I am in the process of collecting data – doing interviews and observations – with three cases of modeling in the Chesapeake Bay region, and I have a fourth potential case study that I’m trying to get involved with as well. Let me give a quick run-down of those.

1) The Chesapeake Bay Modeling System (CBMS) – This is by far the largest and most influential modeling effort in the Chesapeake Bay. It might even be one of the biggest modeling programs in the country. It’s a watershed-scale model, meaning that it covers parts of six states with a variety of political, economic, and ecological conditions. A suite of models are linked together to form the overarching Modeling System. This includes a watershed model, an airshed model, an estuary model, a land change model, and a scenario builder which uses input from the models to construct different possible scenarios for the Bay. The CBMS consists of a massive amount of code and requires a lot of computing power to run. It is housed with the Chesapeake Bay Program, but is produced with assistance from many partner institutions including academic researchers. The model is also a very controversial one because it influences and shapes almost all of the policy and practice that takes place in the region that affects water quality. As a result it has been the focus of a number of reviews and lawsuits claiming that the modeling is wrong. This makes it a very interesting case, and it will serve as the baseline for comparison with other modeling projects in the region.

2) UMCES Water Quality and Eel Grass Model – This model is the only one that doesn’t model effects on the Bay. Instead it is meant to model water quality on the ocean side of the Eastern Shore. It is a much smaller scale and simpler model that looks at the effects of management practices on water quality and, ultimately, on submerged aquatic vegetation (Eel Grass). It’s designed to be web-based so that it can be easily used by various researchers and management personnel in the area to understand the effects that different practices might have on the water in the area. In addition, it has been produced through a participatory process in which those same researchers and management personnel took part so that they could help inform the model-making process and develop a model that best suits their needs and goals.

3) UMD Economic/Ecological Model – I don’t know much about this model yet, because I haven’t had a chance to talk in detail with any of the people involved, and the modeling hasn’t quite gotten under way yet. My understanding is that it will be a linked model that attempts to simulate best management practices and their effects on both the economy and the ecology of the Bay. I don’t know the scale of the model, and I don’t know if that’s been decided yet. There is, as far as I know, no interest in having a participatory or collaborative component to the process, though the modelers are very interested in combining social data with ecological data.

4) NOAA Collective Impact Modeling – This is another model or set of models that I don’t know much about. “Collective Impact” is an approach to collaborative problem solving and decision making developed by researchers at Stanford. This approach is being applied by NOAA and other groups in the Choptank River watershed as an experiment in collaborative environmental management. The group either uses existing models or has developed models for the specific use of the collective impact group that help them to understand the ecological processes and make better informed decisions. I am in the process of contacting the people involved and trying to get a sense of the role that modeling plays.

I am still looking for perhaps one or two more case studies to fill out my research. In particular I am interested in fully collaborative modeling projects in which stakeholders take part in the actual process of modeling rather than simply serving as informants. My goal is to have a spectrum of different kinds of modeling represented so that I can make reasonable comparisons between them. I think the set that I have already provides an interesting picture of what modeling is like on the Bay – it will be interesting to see what else I come up with.

The Beginning…

This isn’t the beginning of my research on modeling. I am a little late getting this blog up-and-running, in part because I’ve already been writing about my work on my other blog. But I’ve recently realized that keeping the two projects separated might be the better way to go, so here I am starting yet another blog! The goal here is to focus on the research, describing the process that I have been going through and will continue to go through as I study computational modeling and write my dissertation. Let me start by explaining a bit about how I got here.

The conception for the project began in the last days of my Master’s degree as I was transitioning to the PhD. My advisor, Dr. Michael Paolisso, approached me about a possible project doing ethnography of the Chesapeake Bay Modeling System (CBMS) in collaboration with two modelers. Given my interest in science and technology studies, I jumped at the opportunity, and over the next few months began writing a proposal with them for National Science Foundation (NSF) funding. At first the goal was just to do a traditional ethnography, but the project has evolved over time.

The first NSF proposal was rejected, but my interest in the project did not wane. I began doing some background research, and applying for other funding. Over the course of this research and writing, my interests changed. It became clear that an ethnography of modeling wouldn’t be enough – it would just replicate all of the other “ethnography of X scientific practice” that have come before. As a result, I became interested in alternative modeling practices – particularly “participatory modeling” in which stakeholders, managers, and modelers work together to produce a model and develop research and management goals. A lot has been written about participatory modeling – I plan to write a brief “resources” page shortly to list some of this work, but for now I’ll just mention the work of Alexey Voinov, Erica Gaddis, and Sarah Whatmore. However, there has not been, as far as I have seen, a comparative investigation of the different modeling practices and the way they affect social and ecological relations. That, in brief, is the purpose of my project.

More broadly, I am interested in the performative dimensions of socio-ecological systems. When we talk about socio-ecological systems and processes, I think we often get caught up in what Graham Harman refers to as “over-mining” – focusing on the activities and agency of the systems as a whole at the expense of the parts. Theories of performativity were developed to address precisely this problem in the social sciences – allowing us to understand the agency of individuals without denying systemic forces. However, it hasn’t been influential in talking about the relationship between humans and the environment. The reason is that we tend to privilege natural scientific explanations in those contexts without an infusion of social theory. Thus we talk about “resilience” and “systems theories” without recognizing that they suffer from the same agency-reducing effects of the structuralist social theories that performativity was meant to challenge. My goal is to introduce an element of performativity into socio-ecological systems theory so that we can think about the processes by which systems are constructed and potential alternative constructions without naturalizing the limits and closing off the possibility for change before we even begin.

The basic assumption of this research is that socio-ecological systems are performative – that they are constructed through the performances of both human and non-humans as they interact with one another. In addition, it is assumed that the performances of both humans and non-humans are in some ways shaped by the overarching system in which they operate. As a result, scientific research is both social and ecological, and we must recognize it as such and ask what effects our work has on the broader socio-ecological system. Modeling is one example of that, and, given that modeling has become increasingly influential in environmental management, it is an excellent place to begin to ask those questions. The question that drives this research is, what are the effects of the production and use of computational models on the broader socio-ecological systems in which they are embedded?

With this in mind, I continued to apply for funding for my research and finally received an NSF Dissertation Research Improvement Grant. With this funding, I will undertake a comparative project to look at the effects of different modeling practices. I will be conducting several case studies on projects currently under way in the Bay watershed, and drawing comparisons between them. Each project will exemplify different approaches and degrees of collaboration between modelers, stakeholders, managers, and others, and I will be particularly interested in the way these different approaches affect the relationships between these groups. In the end, I hope that this research will help modelers and other scientists see that their research methods have effects beyond the immediate production of data, and that this will encourage them to think about the way that their research fits into and performs the broader socio-ecological system to address not just ecological problems but socio-economic ones as well.

Finally, I apply the same performative framework to my own research. I see this work as performative in the sense that it serves as an intervention into a struggling socio-ecological system. Part of the goal of this work is to improve relationships between modelers, managers, stakeholders, natural scientists, social scientists, and others so that we can work together to perform a cleaner, more healthy Chesapeake Bay. Obviously, I don’t expect this small project to have a dramatic effect, but I will do what I can wherever I can and hope that others recognize the value of this kind of work.