Modelling Migration in the Apocalypse

By Christian Espinosa Schatz (Anthropology and Environmental studies PhD student), Yale University 

Climate models predict the future through algorithmic extrapolation. What they lack in imaginative depth they make up in parametrizable flexibility, the ability to output a wide range of “scenarios” by changing the numerical inputs. What happens if we don’t cut emissions until 2025, or do nothing at all? One need merely parametrize the model’s algorithms accordingly to see possible future worlds unfold. The future, however, involves more than climate. It involves politics, culture, and society, variables that resist the quantification of climate models (Hulme 2011). Take for example human mobility. How does one predict the policies, identities, and livelihoods that create and are created by migration? Perhaps in such a complex social issue as migration the model has found its match and must give its predictive authority over to the humanists.  The imaginative work of “cli-fi” may serve society better in predicting climate change’s effects on migration than the quantitative work of climate models.

This, of course, has not happened. Models of climate migration have proliferated in the last decade (Beine and Jeusette 2018). They have this last year enjoyed greater influence in the public sphere through a joint project between ProPublica and the New York Times (Lustgarten 2020a; 2020b). Heralding a new era of the “Great Climate Migration,” Abrahm Lustgarten sets out to answer the question “where will everyone go?” using cutting-edge modelling techniques to focus on Central America (Figure 1). Though he recognizes the difficulty of modelling migration’s social complexity, Lustgarten places the critics of migration models (Boas et al. 2019) within the camp of climate skeptics. The stakes are too high, the possible future too bleak, to not attempt to quantify the future of migration so that policymakers can avert apocalypse, claims Lustgarten.

I have so far assumed a dichotomy between imaginative and quantitative prediction. But perhaps, as many have suggested (Gelfert 2017), a model serves as a “prop for the imagination” and hence is not all that different than science fiction. Both putatively extrapolate from the real world into the future. Both, whether with words or numbers, create captivating possible future worlds.

These assumptions about the relation of models and science fiction to the future, however, do not hold upon inspection. Ursula Le Guin, in the preface to The Left Hand of Darkness, states that “Science fiction is not predictive; it is descriptive”. Science fiction authors “don’t tell you what you will see and hear. All they can tell you is what they have seen and heard … what the weather is now (Le Guin 1969).” I propose that the modelling of climate migration is akin to science fiction not in its ability to predict the future, but in its fabulation of dystopic social worlds (Jasanoff 2015, 1) as a form of description, if only a description of the modeler’s own imagination.

The question is then not merely what the model fails to accomplish. Rather the question is what such a model accomplishes, what real or possible worlds it describes. I briefly offer one approach to answering this question, grounded in the semiotic method of linguistic anthropology (Peirce 1992; Kockelman 2007), with an analysis of how Lustgarten’s model generates the social category of the “climate migrant” through the algorithmic parametrization of possible worlds.

Climate migrants and the parametrization of possible worlds

Climate models are one of the many algorithms that influence contemporary social life. A growing body of research describes how such algorithms do not merely rely on computation but also interpretation (Kockelman 2013). Kockelman describes the various interpretive relations occurring in a model in (Figure 2) as a series of relations between the model, data, and the object being modeled (Kockelman 2020). Of interest to this essay is Kockelman’s assertion that models act as a semiotic “ground” for interpreting the object of the model. Take the following simplistic example: In an unusually hot summer across Europe the temperatures are taken by many Europeans as an instance of the increased temperature predicted by anthropogenic climate change models. Thus many conclude that the unusually hot summer is the result of climate change based on their knowledge of climate models.

Defining the object of Lustgarten’s model, however, involves a dilemma. Unlike temperature the “climate migrant” is not an entity which is readily apparent in the world. Rather, the grounds for interpreting an individual person as a “climate migrant” are undergoing negotiation (McNamara and Farbotko 2017), and modelers are at the negotiation table.

Lustgarten states the model employs a “counterfactual” method for defining and predicting climate migrants across space and time. As it is commonly used, a “counterfactual” is just one of many ways language allows us to talk about more than just the world we inhabit. We can, as linguists say, “displace” the world in all sorts of ways. One such method is the counterfactual, illustrated in the following example: “If the storm system hadn’t been deflected by the jet stream, it would have been snowing in Cambridge” (Heim and Kratzer 1998). This statement is not about snow in the world you or I inhabit, but some world as it would have been should some climatic factor differed. The counterfactual is thus one form of a “possible world” compared with the real world.

This is not the sense in which the climate migration model utilizes a counterfactual. Rather, the model employs a “counterfactual scenario” to “disaggregate climate induced migration from non-climate induced migration.” The mathematics of this disaggregation are relatively straightforward. For each five-year interval the modelers run a series of algorithms that integrate a wide range of climatic, environmental, socioeconomic, and demographic data to produce values unique to hundreds of thousands of cells in a spatial matrix that correspond to 1 kilometer grids of space across Central America. They label the values in this matrix “population potential” which they interpret as a proxy for the number of individuals that migrate to a given cell compared to the previous time interval. They then run the exact same series of algorithms, except keeping constant all variables related to climate and its predicted change. The first set of values presumably refers to all migrants at the selected time interval, whether due to climatic or other variables. The second presumably calculates all migrants due to non-climatic drivers. The simple difference between these two matrices of values results in what the modelers label as the quantity of migration due to climate change for each grid in space.

This “disaggregation” is not a comparison between the real and a possible world, as is usually the case in counterfactuals. Rather it is the comparison of two possible worlds. In plain language, this disaggregation is akin to me saying. “In my dream, there was a sunny day next year in which I went on a run. If that day in my dream had been rainy, I would not have gone on a run.”

The difference between my dream and Lustgarten’s model is that the model’s possible worlds are the product of algorithms rather than restless sleep. As algorithms, they can be parametrized with numerical data. It is between two parametrizations of the model that the modelers generate a new status in society – the climate migrant. Taking up an old anthropological concern (Linton 1936), statuses are the set of culturally contingent positions a person might inhabit in the pattern of social life (e.g. employer, employee, working class, socialist, mother, etc.). This generated status is then taken for the object of the model upon a further set of parametrizations across time.

Towards a counter-apocalyptic science of possible worlds

In the New York Times article, Lustgarten uses his own model as a ground for interpreting the experiences of contemporary Central American Migrants as the first of many climate migrants to come. No other ground but the possible worlds of the model gives reason to ascribe the status of “climate migrant” to Lustgarten’s interlocutors.

Nothing inherent to the model points to climate migration as an apocalypse, though Lustgarten assumes that the future is one in which “misery reigns”. The mapped output of the model communicates this assumption of apocalyptic future, with migration represented through fiery red peaks and ashy black valleys reminiscent of hell. Thus assuming a doomed fate for poor Central American farmers is not internal to the model, but imputed by the modelers’ poverty of imagination and lack of engagement with the very people the modelers’ assume fit the status of “climate migrant”.

 The above analysis thus offers a needed opening for the interpretive social sciences in the practice of modelling possible worlds. The job is not merely one of debunking the model’s unsurprisingly simplistic and overly generalized assumptions. Rather the opportunity is one of improving, collaborating, and co-opting the models to serve the needs and desires of migrants who themselves imagine possible worlds of climate and migration Lustgarten could never expect. The science of possible worlds should not be left to natural scientists alone.

References

Beine, Michel, and Lionel Jeusette. 2018. “A Meta-Analysis of the Literature on Climate Change and Migration.” 18–05. CREA Discussion Paper Series. CREA Discussion Paper Series. Center for Research in Economic Analysis, University of Luxembourg. https://ideas.repec.org/p/luc/wpaper/18-05.html.

Boas, Ingrid, Carol Farbotko, Helen Adams, Harald Sterly, Simon Bush, Kees van der Geest, Hanne Wiegel, et al. 2019. “Climate Migration Myths.” Nature Climate Change 9 (12): 901–3. https://doi.org/10.1038/s41558-019-0633-3.

Gelfert, Axel. 2017. “The Ontology of Models.” In Springer Handbook of Model-Based Science, edited by Lorenzo Magnani and Tommaso Bertolotti, 5–23. Springer Handbooks. Cham: Springer International Publishing. https://doi.org/10.1007/978-3-319-30526-4_1.

Heim, Irene, and Angelika Kratzer. 1998. Semantics in Generative Grammar. Blackwell Textbooks in Linguistics 13. Malden, MA: Blackwell.

Hulme, Mike. 2011. “Reducing the Future to Climate: A Story of Climate Determinism and Reductionism.” Osiris 26 (1): 245–66. https://doi.org/10.1086/661274.

Jasanoff, Sheila. 2015. “Future Imperfect: Science, Technology, and the Imaginations of Modernity.” In Dreamscapes of Modernity: Sociotechnical Imaginaries and the Fabrication of Power. The University of Chicago Press.

Kockelman, Paul. 2007. “Agency: The Relation between Meaning, Power, and Knowledge.” Current Anthropology 48 (3): 375–401. https://doi.org/10.1086/512998.

———. 2013. “The Anthropology of an Equation: Sieves, Spam Filters, Agentive Algorithms, and Ontologies of Transformation.” HAU: Journal of Ethnographic Theory 3 (3): 33–61. https://doi.org/10.14318/hau3.3.003.

———. 2020. “The Epistemic and Performative Dynamics of Machine Learning Praxis.” Signs and Society 8 (2): 319–55. https://doi.org/10.1086/708249.

Le Guin, Ursula K. 1969. The Left Hand of Darkness. 50th anniversary edition. New York: Ace Books.

Linton, Ralph. 1936. The Study of Man: An Introduction. D. Appleton-Century.

Lustgarten, Abrahm. 2020a. “The Great Climate Migration Has Begun.” The New York Times, July 23, 2020, sec. Magazine. https://www.nytimes.com/interactive/2020/07/23/magazine/climate-migration.html.

———. 2020b. “Where Will Everyone Go?” ProPublica. July 23, 2020. https://features.propublica.org/climate-migration/model-how-climate-refugees-move-across-continents/.

McNamara, Karen E., and Carol Farbotko. 2017. “Resisting a ‘Doomed’ Fate: An Analysis of the Pacific Climate Warriors.” Australian Geographer 48 (1): 17–26. https://doi.org/10.1080/00049182.2016.1266631.

Peirce, Charles S. 1992. The Essential Peirce: Selected Philosophical Writings. Edited by Nathan Houser and Christian J. W. Kloesel. Bloomington: Indiana University Press.