Victoria Reyes-García y Matthieu Salpeteur
Matthieu Salpeteura y Victoria Reyes-Garcíabc
a Institut de Recherche pour le Développement, UMR PALOC (IRD-MNHN), 43 rue Buffon, 75005, Paris, France.
b Institució Catalana de Recerca i Estudis Avançats (ICREA), 08010, Barcelona, Spain.
c Institut de Ciència i Tecnologia Ambientals (ICTA), Universitat Autònoma de Barcelona, 08193 Bellatera, Barcelona, Spain.
1. Introduction
Within the Simulpast project, Case Study 1 (CS1) focused on Northern Gujarat, a semi-arid area of India, where archaeological remains show that hunter-gatherers were present from 10.000 to 4.000 BP, and coexisted during the second half of this time-span with communities practicing agropastoralism. The objectives of CS1 were to understand the respective resilience of these subsistence strategies to environmental changes and to decipher the dynamics of their coexistence – until the disappearance of hunter-gatherer populations. The method chosen for this analysis was agent-based modelling (Madella et al., 2014).
The team involved in CS1 gathered researchers and students coming from different disciplinary backgrounds including archaeology, computing sciences, and social anthropology. These researchers were hosted in a range of research institutions (CSIC, ICTA-UAB, UPF, BSC), with very heterogeneous knowledge and experience with computer simulation: some were experts, with years of training in the topic, while most of us were actual novices.
The team started by conducting simulations based on climate models and archeological data related to the area to understand potential livelihood options. Climatic data included rainfall and temperature records at different geographical scales (from the whole Indian subcontinent to the Kutch-Saurashtra area) and archaeological data came from four main archaeological sites. As anthropologists, our role in the team was to bring insights about the social dimensions that might be introduced in the envisioned models. Our work was based on the assumption that, despite many changes, some of the livelihood activities conducted nowadays in the area bear some resemblance with activities conducted in the past. Thus, the aim was for us to bring in information that could help making more reality-based, empirically-grounded models. The data expected from us included, on one hand, secondary information from the available ethnographical and historical literature and, on the other hand, first hand observation from fieldwork to be conducted with contemporary populations of mobile pastoralists in Gujarat.
Over the duration of the project, such highly interdisciplinary collaboration led to works that have informed theory in archeology and anthropology alike. For example, the team developed two models, one focusing on hunter-gatherers (Balbo et al., 2014) and another one focusing on agropastoralists (Lancelotti et al., 2014). These models allowed us to test hypotheses about the resilience of such livelihoods when facing different kinds of climatic variability. Moreover, these models contributed to the use of quantitative methods in general, and modeling in particular, in archaeology. Work done by the team collaborating in CS1 also brought results pertaining to computer simulation design and methods, as the models developed were also used to test several modelling options, for example about agents decision-making and the use of artificial intelligence (e.g., Francès et al., 2015). During the whole project, developments were brought to the scientific community through dedicated sessions during conferences (such as the Society of American Archeology, Memphis, TN, 2012).
Besides results directly related to models, work conducted under CS1 also informed current anthropological debates using fieldwork data collected in Gujarat. For example, qualitative and quantitative data collected with contemporary pastoralists has informed the mechanisms of transmission of traditional knowledge (Salpeteur et al., 2015, 2016) or changes in contemporary pastoral systems (Salpeteur et al., 2017).
But, beyond the team’s academic results, we would like to highlight here some lessons learned through the journey together, as such lessons may benefit a wider academic audience involved with similar approaches or broadly involved in interdisciplinary research.
2. Model development as a solid ground for interdisciplinary research
A famous anthropologist once wrote that “the methods for collecting and analyzing data belong to all of us” (Bernard, 2006, vii). Implicit in this sentence is that methods could help bridge knowledge from different disciplines, fostering interdisciplinary discussions. This was, indeed, our experience as we learnt the basics of Agent-Based Models (ABM – see Epstein & Axtell, 1996), the selected modeling approach in CS1.
The development of ABM requires to follow some specific rules. This includes introducing variables in the model with parsimony; defining with great details each variable, its associated numerical values -if any-, and its behaviour; and having a good understanding of the potential interactions between variables once the model will be running. These conditions define the potential usefulness of a model, and the future readability of its outcomes (e.g., Grimm et al., 2010). In theory, i.e., written in a methods’ manual or in a project’s guidelines, these rules seem obvious enough to be easily fulfilled. In practice, however, materializing a model with a social component that follows these rules requires intense team work, hours and days of collective thinking, and the ability to understand each-other across disciplines.
To achieve the development of working and meaningful models within CS1, we had to engage in in-depth conversations across archaeology, computer science and anthropology. We had to think together about the variables we thought were needed in the model and a myriad of details that needed to be considered or discarded. The process, guided by a shared goal, forced participants to go beyond the comfort zone of their own discipline, to engage with the other disciplines and to reach agreement on key model dimensions.
A good example of the kind of discussions we had relates to the notion of “social relations”, something the team discussed in the first stage of model design. All members of the team considered that social relations was a key aspect to be included in the model, as social relations help account for basic social organisation, cooperative behaviour, and potential inter-individual exchanges of food (all factors that affect the likelihood of success of the different livelihood strategies examined, hunter gatherers and agropastoralist). However, many ways to define and model social relations can be imagined, and only one needed to be selected. Would agents belong to specific groups, and interact only with members of their group ?, or, on the contrary, would interactions be opportunistic, based on random encounters? If agents were assigned to groups, what would be the size of such groups? Something equivalent to a household ? to an extended family ? or rather, to a residency cluster? What about trade relations, sometimes on long-distance, as illustrated by archaeological research? Would it be relevant to include them?
After some discussions, the group reached a consensus by choosing the household as the basis of our model, in other words, the «agent» in our ABM. As anthropologists, trained to look for and to value variation in livelihood strategies, we had a hard time seing the «household» as the best unit of analysis, as there is much variation of what a household can be in ethnographical literature, and the model could not take into account all the details we have been trained to value (e.g., polygamy, household life cycle, aliances). Computer scientists argued that such complexity would make calculations hard and were not convinced that it mattered at large. It took several meetings to define a household that was complex enough to capture the variations highlighted by anthropologists and simple enough to allow computation.
Most of the model design revolved around this type of discussions, as for examples issues related to food gathering and consumption (modelled through energy, not through cultural practices), demographics (modeled through hypothetical birth rates and creation of new households, not through rites such as marriage), or ability to move in the land (modeled considering physical constraints, not individuals own will). The process can roughly be summed up as synthesizing the complexity of social behaviour into rules such as “When agent A meets agent B, then…”. And people from all disciplines need to agree that such rule makes sense and it is useful in the model.
The process was quite demanding from all participants, as it required a great level of open-mindedness, and ability to take distance from own disciplinary background. The result, however, was rewarding as the models that were created were not linked to any particular discipline. Rather, as they emerged from our interactions, they could be placed somewhere in-between our respective expertises.
3. What learnings can we highlight from this experience?
We derive four main lessons from this process. First, ABM revealed themselves as adequate tools for interdisciplinary research. Interdisciplinary projects usually require researchers with different disciplinary backgrounds to be able to understand each other while contributing to reach a common goal with insights from their own epistemological and methodological toolkit. In our case, the co-design of ABM was a process that ultimately allowed us to generate new knowledge. Moreover, this new knowledge is interdisciplinary in nature, as it requires participants to actually define and agree on every detail. The design of the model can be pictured as an arena, where researchers will bring concepts, terminology, hypotheses and understandings from their own disciplinary background. In the design of an ABM, these concepts interplay with concepts from other disciplines to create new knowledge. Importantly, in this process, much of the assumptions that usually remain implicit when working within one’s own discipline have to be made explicit and discussed across all the involved disciplines. Some basic terms and notions that have a well-defined meaning within one field may become unclear when confronted with other disciplines, as illustrated by the example on social relations discussed above. As such, the development of ABM fosters knowledge exchange and mutual understanding across involved researchers, in a particularly efficient way.
The second lesson learned stems from the first: being involved in ABM development is a big challenge, particularly for researchers coming from empirical research fields or backgrounds, where methods to simulate the past or the future are not common. Designing a model to simulate human behaviour is, indeed, a research process different from usual research in anthropology, which is based on detailed and qualitative data, i.e., thick descriptions (Geertz, 1973), paying much attention to the meaning given by people to their actions, emphasizing the complexity and historically-grounded dynamics of social collectives (e.g., Ellen, 1999). Far from the type of data anthropologists are trained to collect, ABM development asks for “simplified” social dynamics: only a limited number of dimensions (reduced to variables) can be introduced in the model, in a very controlled way. Thus, the analytical work that is required in ABM design is quite different from usual research in anthropology, and some scholars may not feel comfortable or be convinced by such an approach. ABM development may, indeed, not be suited for all.
The third lesson learned is that such interdisciplinary work requires, not only participation from several disciplines from its very beginning, but also time. Indeed, in our particular case, the collaboration started by sharing readings and knowledge about models and our respective disciplines. A dedicated seminar, called Ecotono, was organised from the first months of the project, during which every participant was invited to propose key papers in his/her research field, but on topics that could be of interest to the project. This seminar generated rich discussions about social change, cultural transmission, hunter-gatherer ecologies, computer simulation in social sciences, and others. Through the discussion of seminal papers, we shared conceptual frames, epistemologies, methodologies, and research interests, thus creating a ground that was at the same time common to the team, and new to each of us – as it was different from our own disciplinary basis. This common ground was a solid basis to start constructing the model together. Indeed, it was only after this first period, that we were able to move forward to the actual discussions about the models to be developed. In sum, it took time (i.e., several months) to construct the mutual understanding that was needed to work together. This is a key, but often neglected, aspect of such interdisciplinary projects. Future applicants or funding agencies aiming at fostering this kind of research should definitely take this aspect into account.
The last lesson we learned through these interactions is that, when such strong new epistemological and methodological bases are put in place, collaborations extend beyond the project. In our case, this is reflected in collaborations in publications beyond CS1 (e.g., Balbo et al., 2016, Reyes-García et al., 2013, 2016) and in the organization of workshops and conferences, such as the international conference “Small-scale societies and environmental transformations: Co-evolutionary dynamics” held in Barcelona, in December 2014.
4. Conclusion
Interdisciplinarity requires a shared knowledge basis, mutual understanding across disciplines, and the co-construction of research questions and methods. The process that enables such exchanges requires time, proper spaces dedicated to discussions and knowledge sharing, and open-mindedness from participants. As we experienced it during the Simulpast project, ABM design appears as a particularly interesting tool in this regard. We believe this explains why the Simulpast project was such a rich research experience, for us, social anthropologists.
References
Balbo, A. L., X. Rubio-Campillo, B. Rondelli, M. Ramírez, C. Lancelotti, A. Torrano, M. Salpeteur, N. Lipovetzky, V. Reyes-García, C. Montañola, and M. Madella. (2014). “Agent-Based Simulation of Holocene Monsoon Precipitation Patterns and Hunter-Gatherer Population Dynamics in Semi-Arid Environments.” Journal of Archaeological Method and Theory 21(2), pp. 426–46.
Balbo, A., E. Gómez-Baggethun, M. Salpeteur, A. Puy, S. Biagetti, and J. Scheffran. (2016). “Resilience of Small-Scale Societies: A View from Drylands.” Ecology and Society 21(2).
Bernard, H. Russell. (2006). Research Methods in Anthropology: Qualitative and Quantitative Approaches. Fourth edition. Walnut Creek, CA: Altamira Press.
Ellen, R. (1999). “Forest Knowledge, Forest Transformation: Political Contingency, Historical Ecology and the Renegotiation of Nature in Central Seram.” Pp. 131–57 in Transforming the Indonesian Uplands: Marginality, Power and Production, edited by T. Li. Netherlands: Harwood Academic Publishers.
Epstein, J. M., & Axtell, R. (1996). Growing artificial societies: social science from the bottom up. Cambridge & Washington, DC: MIT Press & Brookings Institution Press.
Francès, G., X. Rubio-Campillo, C. Lancelotti, and M. Madella. (2015). “Decision Making in Agent-Based Models.” pp. 370–78 in Multi-Agent Systems. Vol. 8953, Lecture Notes in Computer Science, edited by N. Bulling. Cham: Springer International Publishing.
Geertz, C. (1973). “Thick Description: Toward an Interpretive Theory of Culture”, in Geertz C., The Interpretation of Cultures: Selected Essays, New York: Basic Books, pp. 3–30.
Grimm, V., U. Berger, D. L. DeAngelis, J. G. Polhill, J. Giske, and S. F. Railsback. (2010). “The ODD Protocol: A Review and First Update.” Ecological Modelling 221(23), pp. 2760–68.
Lancelotti C., X. Rubio-Campillo, M. Salpeteur, A.L. Balbo and M. Madella. (2014). “Testing the resilience of agro-pastoralists communities in arid margins through ABM”. In Amblard M., Barcello and Madella, Advances in Computational Social Science and Social Simulation, Barcelona: Universitat Autonoma de Barcelona. http://ddd.uab.cat/record/125597.
Madella, M., B. Rondelli, C. Lancelotti, A. Balbo, D. Zurro, X. Rubio-Campillo, S. Stride. (2014). “Introduction to simulating the past”. Journal of Archaeological Method and Theory 21(2), 251-257.
Reyes-García V., M. Gueze, A. C. Luz, M. Macia, M. Orta-Martínez, J. Paneque-Gálvez, J. Pino, X. Rubio Campillo. (2013). “Evidence of traditional knowledge loss among a contemporary indigenous society”. Evolution and Human Behaviour 34, pp. 249-257.
Reyes-García, V., A. L. Balbo, E. Gómez-Baggethun, M. Gueze, A. Mesoudi, P. Richerson, X. Rubio-Campillo, I. Ruiz-Mallén, S. Shennan. (2016). Multilevel processes and cultural adaptation: Examples from past and present small-scale societies. Ecology and Society. 21 (4), 2. [online] URL: http://www.ecologyandsociety.org/vol21/iss4/art2/. (Special Issue: Small Scale Societies and Environmental Transformations: Co-evolutionary Dynamics. Madella, M. and Reyes-García V., ed.).
Salpeteur, M., H.R. Patel, A. L. Balbo, X. Rubio-Campillo, M. Madella, Ajithprasad P., and V. Reyes-García. (2015)9. “When Knowledge Follows Blood. Kin Groups and the Distribution of Traditional Ecological Knowledge in a Community of Seminomadic Pastoralists, Gujarat (India).” Current Anthropology 56(3), 471–83.
Salpeteur, M., H.R. Patel, J.-L. Molina, A.L. Balbo, X. Rubio-Campillo, V. Reyes-García, and M. Madella. (2016). “Comigrants and Friends: Informal Networks and the Transmission of Traditional Ecological Knowledge among Seminomadic Pastoralists of Gujarat, India.” Ecology and Society 21(2), 20.
Salpeteur, M., M. Madella, H.R. Patel, and V. Reyes-García. (2017). “Adaptation, Access to Resources and Mobility: From Contemporary Pastoral Systems to Ancient Societies.” Nomadic Peoples 21(2), 191–213.
