© Copyright JASSS

  JASSS
logo

Dwight W. Read (1999)

Introduction to the special issue on Computer Simulation in Anthropology

Journal of Artificial Societies and Social Simulation vol. 2, no. 3, <http://jasss.soc.surrey.ac.uk/2/3/10.html>

To cite articles published in the Journal of Artificial Societies and Social Simulation, please reference the above information and include paragraph numbers if necessary


1.1
The articles in this thematic issue of JASSS exemplify some of the ways that simulation is used in anthropology. Though simulation in anthropology goes back to the 1960's (Fischer 1994: 184), there is, as yet, no consensus about the direction that simulation ought to take or is likely to take in the future. In many ways simulation in anthropology reflects the heterogeneity of the field, both theoretically and methodologically. Nonetheless, most examples of simulation in anthropology share a common concern with addressing research questions found to be methodologically difficult, if not intractable, when approached through more conventional modeling. Two early examples that illustrate this aspect of simulation in anthropology are the demographic simulations undertaken by Nancy Howell (Howell and Lehotay 1978) to better understand population dynamics at the scale of hunting-gathering groups and the simulation done by David Thomas (1973) aimed at modeling the distribution of paleo-indian sites in the Great Basin of the western portion of the United States.

1.2
The first of these two simulations - a demographic simulation - developed out of the Harvard Bushman (San) Research Program begun in 1963 among the !Kung san of Botswana and directed by Richard Lee and Irven DeVore (see, for example, Lee and DeVore 1976). The project was concerned with biological, demographic, ecological, social and cultural aspects of the !Kung san. The simulation, described in Howell and Lehotay (1978), enabled construction of a demographic profile of the !Kung san, taking into account the sampling issues that arise from both the small size of the population (several hundred individuals) and lack of direct information on the age of individuals (calendrical age is not of concern to the !Kung san). The simulation was individual based and kept track of individuals in the simulation as they aged, gave birth and died. But as only a relatively small number of anthropologists were familiar with computing methods at that time, the simulation software was not widely used beyond analysis of the demographic structure of the !Kung san (see Howell 1979), though that pioneering research has had extensive impact through its detailed, demographic study of a small-scale, non-western society.

1.3
Thomas's simulation had a different kind of impact. Whereas Howell's simulation addressed primarily problems that arise from stochastic effects in small populations coupled with incomplete demographic data, Thomas's simulation attempted to account for the distribution of artifactual material and the spatial location of paleo-indian sites in the Great Basin of the United States through application of then current theories about factors affecting the spatial location of hunting and gathering groups as they underwent their yearly round. The verification problem faced by the archaeologist when positing an argument for site location is substantial. The archaeologist has only the artifactual remains of groups as indicators for the locus of activities. The nature of the activities and hence of the general character of the archaeological site (e.g., permanent village, temporary camp, hunting site, etc.) is necessarily inferential. Even more problematic are attempts to infer the reasons why those activities took place at a particular geographic locality. Thomas saw simulation as providing a way to model the distribution of sites across space using an environmental and ecologically framed model for site location. In addition, the model also made assumptions about the likelihood that artifactual materials would be discarded rather than carried away when the group moved to another locality. Whereas the intent of the !Kung san demographic simulation was to aid in constructing a "best possible" demographic profile, the importance of Thomas's simulation lay not so much in its ability to accurately model site location and spatial density of artifacts as in providing data that could be used to assess theories archaeologists were making about site location and site formation processes.

1.4
These two themes - the use of simulation to work out the consequences of processes already reasonably well modeled versus simulation that attempts to clarify what might be those processes and their consequences - are exemplified by the three simulations presented in this issue of the JASSS. The two simulations done by Piazza and Pearthree and by White, respectively, though quite different in content, are concerned with establishing whether or not a posited process accounts for a set of empirical observations. Piazza and Pearthree want to know if the process of down-the-line migration as posited by the archaeologist Patrick Kirch is consistent with the archaeological data on the migration and colonization by the Lapita peoples in the southwest portion of the Pacific that began about 3500 years ago and took place over a time period of about 500 years (Piazza and Pearthree, this issue). For Piazza and Pearthree the validity of the migration model, per se, is not at issue; only whether the model predicts results consistent with the archaeological data. They constructed the simulation to be in accordance with the posited migration process and used a range of parameter values (growth rate, migration rate, and initial population size) estimated from historical data that bounded plausible values for these parameters. They find that the historical pattern of colonization, measured in terms of number and population size of colonies over the time period in question, could not be duplicated except with unrealistic parameters such as a 90% migration rate. The simulation results led them to suggest a different model for the migration that removes the sequentialness of colony formation inherent in the down-the-line migration model. The sequentialness implies that colony formation is only weakly affected by earlier, distant colonies; e.g., with a 10% migration rate from a colony, only 0.1% of the initial population would directly become part of migrants making up the third colony with down-the-line migrations. In place of the down-the-line migration they suggest "sea-nomadism" as a way to account for both the rapid development of colonies and the archaeological evidence that suggests early settlement and approximately equal population sizes throughout the region colonized by the Lapita peoples.

1.5
In the simulation undertaken by White the topic of concern relates to establishing whether or not a social network established through a series of marriages has a pattern due solely to stochastic processes biased by demographic factors, or whether, after removing stochastic and demographic biasing effects, the marriage structure is still patterned. In some cases there might be an a priori expectation of patterning; e.g., patterning arising from marriage rules that form what Lévi-Strauss calls elementary kinship systems. In elementary kinship systems marriage rules (whether preferential or proscribed) are expressed in terms of kinship categories. As discussed by White, previous work by Hammel (1976) raised the question of whether such systems did, in fact, induce patterning, Demographic simulations done by Hammel suggested that the incidence of prescribed marriages was no more greater than would be expected by chance alone, once demographic factors were taken into account.

1.6
For other societies where marriage is guided only by prohibitions such as incest taboos, patterning in marriages might arise from what White refers to as "strategies," namely marriage decisions that take into account a variety of factors such as inheritance, prior family alliances, and the like. In both cases the methodological problem identified by White is one of disaggregating the social aspect of marriage choice from the demographic aspect. The methodological problem being addressed, then, is to determine how the "marriage structure can be statistically disaggregated into demographic and social choice components" (White, this issue). White does so by constructing a baseline of random marriage choices that takes into account the actual frequency of marriage at a specified level in a marriage structure. Statistical comparison can be made between the actual marriage frequency for some specified criterion (e.g., spouses have similar age) with the frequency of that kind of marriage in a distribution of marriages produced through random permutation of actual marriage partners made in accordance with whatever structural criteria is used to define the comparison level (e.g., all marriages, ignoring age). According to White "we would expect this subset [of marriages satisfying a specified criterion] to have non-random characteristics in comparison to a uniform-probability model at that level. Marriages of similar-age spouses, for example, appear non-random against a uniform-probability model across all age groups" (White, this issue).

1.7
The method obviates the need for modeling the demographic history of the population being examined - a difficult task at best - by using random permutations of actual marriage partners to generate a baseline frequency distribution against which comparison of actual marriage frequencies can be made. The baseline model captures the patterning due to both the historical, demographic facts of the population and to criteria affecting marriage choice, much in the way a regression model separates variation due to the regression model from other effects, thereby allowing for assessment of any patterning in the residuals above and beyond what is expressed in the regression model. This also provides a method, as White discusses, that is sensitive to the distinction made by Lévi-Strauss between elementary and complex kinship systems.

1.8
The third article by Cathy Small, uses a multiagent approach to implement ethnographic data on the fahu custom in Polynesia. Fahu refers to the custom of deference through which prestige and goods would flow to the descent line of a woman and her children from the descent line of her brother and his sons due to her superior status in relationship to her brother. Under normal conditions, Small observes, a woman would marry hypergamously, hence consolidating power in high ranking descent lines. Under conditions of warfare, though, it might be expected that lower status chiefs who had greater wealth and power would want to eliminate the fahu custom since it legitimized higher status and one could not "directly appropriate rank and prestige through power and wealth" (Small, this issue). The question she addresses through the simulation is: Why, in fact, is there no historical evidence for elimination of the fahu custom under conditions of warfare? She posits an answer based on the pattern of marriages produced by the simulation when there is no warfare versus the pattern that occurs when there is warfare. In this case it is not a proposed process that is being examined, but change in patterning in an on going system triggered by a change in one of the parameters, namely amount of available land. The latter apparently served as a catalyst for warfare.

1.9
The fourth article, by Klüver and Schmidt, though not by anthropologists, takes up a theme - the emergence of social dynamics considered at a global level from local level social interactions of agents and the rules specific to these interactions - likely to be of growing importance in theorizing about social systems, whether by sociologists or by anthropologists. This theme has guided a recently published simulation in anthropology by David Kronenfeld (Kronenfeld and Kaus 1993). In his simulation Kronenfeld considers the means by which social solidarity, as discussed by Durkheim, might arise. Kronenfeld uses the flocking property of starling birds as a way to exemplify the idea that it is not humanness, per se, that is critical for the emergence of social solidarity, but the rules for the interactions of agents. Kronenfeld's use of a simulation based on the flocking behavior of starlings also brings to the fore the problem of disentangling in our theories of human social dynamics those dynamics that only arise in the human context from social dynamics that arise via processes shared with non-human species. To the extent that the former are attributable to culture/language as the primary, distinctive feature of human societies, it is of importance not only to argue for processes that account for the dynamics of human societies, but why, if these dynamics are claimed to be uniquely human, the same properties cannot arise in non-human societies (see Read 1987 for a more detailed discussion of this theme).

1.10
Klüver and Schmidt indirectly provide a partial answer by introducing the concept of the geometry in which the actors are embedded. The geometry, according to Klüver and Schmidt, determines what kind of interaction is likely to take place. The geometry is not that of physical space but of "[s]ocial 'spaces' such as institutions, organizations or social networks" as these "are constituted by nothing other than social rules of interactions" (Klüver and Schmidt, this issue). It is that geometry, along with the social rules of interaction, that indicates when the resulting social dynamics are uniquely human.

1.11
Klüver and Schmidt formally define a topology and metric for the social space through the notion of adjacency of social actors. They use the formal representation to extend the notion of control parameters for systems based on Boolean networks or cellular automata to "geometrical" control parameters that "can be expressed using graph theoretical concepts such as the density of graphs or geodetical properties." The formalism they introduce allows them to explore the relationship between the range of values taken on by control parameters and social complexity. They summarize their conclusions in the form of a "Theorem of Social Inequality: A social system characterized by a rule system whose properties produce social inequality will generate only simple dynamics; the more the system contains properties of social equality, the more complex the dynamics of the system."

1.12
Underlying the argument is their "conviction that social reality is to be understood only as the product of social actions and interactions governed by specific social rules.... that is why the classical methods of the natural sciences only achieved limited success when applied to the social sciences. Therefore one has to look for formal systems which can serve as models of local interactions and then investigate their particular features" (Klüver and Schmidt, this issue). The models used by Klüver and Schmidt are based on Boolean networks as these, they argue, are "elaborate enough to tackle the problems dealt with by the mainstream of sociology."

1.13
Fischer and Read reached much the same conclusion for anthropology when they commented that "traditional analytic models alone cannot address (as yet) the complexity of modeling distributed agency embedded within the complex material and symbolic contexts in which agency is expressed. But incorporating this complexity within models is necessary if we are to succeed in modeling the issues on which most anthropologists work, especially human behavior" (Fischer and Read, 1999). Multi-agent simulation, argue Fischer and Read, provides a means for addressing this modeling problem since " [a]gent-oriented modeling has the potential of serving as a means for retaining the range and diversity of interests found within a society, so that the emergent properties arising from the interaction of agents with diverse interests can be retained within a model." The simulations in this issue also provide examples of the role that simulation can play in this kind of anthropological theorizing.

* References

FISCHER, M. 1994. Applications in Computing for Social Anthropologists. London: Routledge.

FISCHER, M. and D. Read, 1999 Economics and Social Science Research Council Grant Proposal, UK.

HAMMEL, E. 1976. The Matrilateral Implications of Structural Cross-Cousin Marriage. pp. 145-168, in, Ezra B. W. Zubrow, editor, Demographic Anthropology: A Quantitative Approach. Albuquerque: University of New Mexico Press.

HOWELL, N. 1979, Demography of the Dobe !Kung. New York: Academic Press.

HOWELL, N. and V. A. Lehotay 1978. AMBUSH: A Computer Program for Stochastic Microsimulation of Small Human Populations. American Anthropologist 80:905-22.

KRONENFELD, D. and A. Kaus 1993. Starlings and Other Critters: Simulating Societies. Journal of Quantitative Anthropology 4:143-174. LEE, R. and I. DeVore (eds) 1976. Kalahari Hunter-gatherers : Studies of the !Kung San and Their Neighbors. Cambridge: Harvard University Press. READ, D. 1987. Foraging Society Organization: A Simple Model of a Complex Transition. European Journal of Operational Research 30:320-326

THOMAS, D. H. 1973 An Empirical Test for Steward's Model of Great Basin Settlement Patterns American Antiquity 38:155 - 173.

ButtonReturn to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, 1999