Reviewed by
David L. Sallach
Social Science Research
Computing, The University of Chicago.
Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes is a book of articles[1] edited by Timothy Kohler and George Gumerman (2000). Drawing upon complex adaptive systems (CAS) theory, and focusing upon primate and simple human societies[2], these diverse and insightful studies expand the horizon of agent simulation research.
Attempts to distinguish scholarly periods can be imprecise and arbitrary, yet they can also help identify and characterise progress in a field of study. This collection represents a significant contribution to the agent simulation research program and, in fact, might be regarded as an exemplar of a third stage of agent simulation studies.
The early work of Schelling (1978), Maynard Smith (1982) and Axelrod (1984)[3] provided a first wave of exemplars demonstrating the potential of a new approach to social simulation research.[4] A new generation of agent simulation research, including Epstein and Axtell (1996), Axelrod (1997) and Young (1998), provided a second wave of exemplars. They respectively illustrate, inter alia:
From the standpoint of standard periodisation, it seems premature to identify a new stage in agent simulation research. However, considering the substantive contributions made by these studies, a new level of sophistication is introduced into agent modelling. In particular, a number of chapters in this collection serve as exemplars in the area of empirically grounded agent simulation, investigations that stand in visible contrast to the study of abstract social processes.
Empirically Grounded Agent SimulationOne example is informed by patterns of aggregation and abandonment in Pueblo populations in the Mesa Verde region of Colorado (USA) between 900-1300 C. E. (Kohler et al. 2000). Annual maps of agricultural potential, created and updated by Van West (1994), were used to construct paleo-production landscapes. These landscapes include a base map partitioned into cells of four hectares each, with associated elevation settings and soil data including soil depth, water capacity, plant productivity, agricultural yield and a Palmer Drought Severity Index (PDSI) calculation that integrates the effect of precipitation on available soil moisture. The latter is then re-expressed as potential bean and maize yields.
To the paleo-production model are added all known prehistoric sites, assigned to one of three Pecos Classifications of culture, and associated with a Universal Transverse Mercator (UTM) co-ordinate. The latter are transferred to a Geographical Information System (GIS) application, where they are converted to grid layers.
A similar database of hydrographic information was drawn from United States Geological Survey (USGS) maps, a Colorado Division of Water Resources database and Bureau of Land Management Hydrographic Inventory records. The USGS maps were digitised into three distinct layers representing springs and rivers, perennial streams and intermittent streams. In addition to identifying numerous additional springs, the latter two data sources also provide flow rates and consistencies.
Households make planting decisions based on past harvests, including searching the wider area for plots, and assessing the need for possible relocation. Households are subject to probabilistic fertility, mortality, and children's marriage rules, with the new households having equal probability of remaining near or moving away. Households engage in weeding and harvesting activities, which help define their caloric needs. Continuous farming in specific locations result in a reduction of maize productivity.
Fifteen sets of site location rules were tested relative to two distinct productivity levels based on a FALLOW_FACTOR parameter indicating the intensity of planting. The viability of various simulation models is assessed using the fit between simulated and estimated actual population trajectories, although the article does not employ statistical tests in the assessment.
A second example of grounding an agent simulation in empirical data patterns is found in Long House Valley (Arizona, USA) from 1800 B. C. E. to 1300 C. E. (Dean et al. 2000). Long House Valley (LHV) has an (intensively surveyed) archaeological history that includes seven distinct environmental zones.
Using surficial geomorphology, palynology, dendroclimatology and archaeology, the project uses multiple data sources to reconstruct low- and high-frequency variations in alluvial hydrologic and depositional conditions, effective moisture and climate in unprecedented detail. Measures of environmental variability, including the fluctuation of alluvial groundwater, the deposition and erosion of flood plain sediments and extrapolations from the relationship between PDSI and soil types (Van West 1994), are used to create a dynamic landscape of annual maize production. The process of making such estimates, which is described in much more detail than can be recorded here, should be of significant value to others undertaking comparable projects.
Modelled household (agent) attributes include life span, vision, movement, nutritional requirements, consumption rates, storage capabilities, grain stocks and location. Household agents harvest grain, storing any beyond the annual consumption of 800 kg of maize. However, grain stored for over two years is lost. Households can cease to exist, however, in addition to member death, this may involve being absorbed by other households, or moving out of LHV.
Each simulated household is conceived to be both matrilineal and matrilocal, so rules controlling household formation and movement focus upon female members of the household, including daughter-based fission assumptions. Based on grain stores and current year harvest, the household estimates the amount of grain that will be available in the following year. If the estimate will not satisfy current requirements, the household moves, first determining a productive farmland site within 1600 meters of a water source, and then selecting a settlement location (the location nearest the farmland that contains a water source). Available land is defined as land that is not currently farmed or settled.
Evaluation of the models involves assessing the fit between multiple simulation runs, on the one hand, and estimated archaeological and environmental data, on the other. As with Kohler et al. (2000), the process of assessment is graphical rather than statistical. Estimation procedures for the actual data influence the goodness of fit because, as an example, the ceramic data upon which site occupation is based has quarter-century grain, whereas simulated populations can be (and are) computed annually. Discrepancies between the actual results and simulations are (reasonably) regarded as artefacts of the estimation process by the authors. Nonetheless, since simulated population growth (decline) could be computed using a quarter-century grain, it would seem that making such an adjustment would add clarity to the resulting fit assessment.
Small (2000) provides a third example of empirically grounded simulation. But, whereas the first two studies examine the evolution of settlement patterns within particular ecological niches, Small's topic is more specialised: the impact of high-status marriage rules upon social structure in Polynesia.
Chiefly status dynamics in Polynesia are complex, with marriage being a core political process. Prototypically, Polynesian marriages are hypergamous (wives marry up in status). To increase the rank of one's line, one must use wealth to attract a hypergamous marriage. As a result, what Tongans call fahu privileges institutionalise a regular flow of wealth from wife receiving lines to wife giving lines. Status itself has multiple subtly interrelated dimensions, including inherited rank (including primogeniture), the effect of supporting lines, and personal ability.
Small investigates three sets of marriage rules: incest, virgin sister and fokonofo. Although incest taboos prohibiting marriage among first and second cousins are fully enforced on commoner marriages, they are relaxed for chiefly lines to the extent that for chiefly marriages of sufficiently high status, sibling marriages are permitted and honoured. For lower status lines, incest rules disallow sibling or parallel cousin marriages, but do permit cross-cousin marriages. The virgin sister rule prohibits the marriage of the eldest (sacred) sister of the highest line. One interpretation of this rule, that Small investigates, is that it prevents a high-ranked sister from passing her status to another line, and thus protects the relative status of her natal line. The fokonofo rule is a constraint upon the form of polygamous marriages. In particular, in a marriage between two high chiefly lines, the wife may bring additional wives, typically younger sisters or inferior cousins, into the home, and these fokonofo wives would be her husband's only other wives. The convention preserves the status of their eldest son by preventing the dissipation of status originating in the wife's line.
The model handles the dynamics of up to fifty (hierarchically related) chiefly lines as they marry, procreate, and generate and distribute agricultural wealth. Wealth generation is influenced by land ownership, labour (kin size), and leadership ability. Wealth distribution includes fahu obligations, annual first fruit tribute to the highest ranked line as well as immediately superior lines and, finally, the obligations of a chief to his commoner retinue. While most of the agricultural wealth is redistributed to the commoners that produce it, chiefly lines vary in their ability to do so. Low redistribution rates may erode commoner loyalty to the line, and result in defections to other lines.
Small's simulations suggest that marriage rule changes have significant effects on the internal stratification developed and sustained within the Polynesian chiefly system. When chiefs cannot marry their relatives, chiefly statuses converge. The highest chiefly line was more than twice as stable when the virgin sister parameter was active. The fokonofo convention was more indirect and context dependent than the first two marriage rules, increasing stratification only in mature systems relying primarily on cross-cousin marriages. However, Small observes, this context approximates the circumstances in which the fokonofo rule originated. Small anticipates further studies of the relationship between kinship dynamics and social structure in virtual Polynesia.
Another study that is both specialised and empirically grounded is provided by Lansing's (2000) study of the subak system of Bali. Subaks are "egalitarian, co-operative farmers' associations that manage the flow of irrigation water into rice terraces." Subaks are small, averaging less than 50 hectares and 100 members. Subaks are interdependent in the sense that a subak cropping pattern actively modifies the ecological conditions (pest populations and irrigation water flows) of its neighbours.
Empirical planting patterns show (counterintuitive) patterns of synchronous planting, even though such co-ordination creates simultaneous shortages of water (at the start of the planting cycle) and labour (at the start and end of the planting cycle), and also a surfeit of water a few weeks into the planting cycle. The ecological advantage of the synchronous planting cycle is that rice pests (including rats, insects and insect-carried diseases) are deprived of an alternative habitat, causing their presence to decline. Notwithstanding the fact that subaks do not intentionally create a complex cropping schedule that maximises water usage and minimises pest damage, simulation studies show they achieve a near optimal solution along the entire watershed. The purpose of Lansing's study is to understand how this outcome occurs.
Lansing begins by defining upstream and downstream subaks as players in a two-person game, in which the former care about pest damage but not water stress, while the concerns of the latter are inverted. Using an extensive survey of farmers from ten subaks, Lansing validated the structure of these preferences. Lansing then simulates this spatial game on a lattice, where subaks seek to improve their harvests by monitoring four immediate neighbours (north, south, east, west), and imitating the cropping pattern of the more successful of these neighbours. When displayed graphically, the simulation results show a close relationship with the actual planting patterns in Bali. The model further predicts that harvest yields will rise while variance will decline. This prediction appears to correspond to empirical results as well.
Lansing subsequently investigates the impact of varying the number and location of neighbours monitored by subaks in the model. The adaptability and stability of the result appear to be quite sensitive to these factors. Related research discusses a cultural system of agricultural rites and water temple networks that institutionalise subak patterns of co-operation (Lansing 1993).
Reynolds (2000) addresses a substantively compelling issue in a technically innovative way. A decision tree structure is created to identify how raiding influences settlement patterns (and state formation) in the Valley of Oaxaca (Mexico) between 1400 B. C. E. and 500 C. E. In each region and time period, the settlement pattern at over 2700 sites is described by over 100 variables describing the environment, architecture, and economy, including agriculture, craft production and trade. A dozen environment conditions helped to classify raiding patterns, including the ability to identify periods within, for example, the defensive positioning.
The decision tree form of knowledge capture seems appropriate for case assessment or diagnostics, more than for identifying generalisations implicit in the process under investigation. The most serious issue is that, by sequentially utilising the best available indicator in making its binary classification, the decision tree does not capture interaction effects. It might be preferable to consider using Multivariate Adaptive Regression Splines (MARS), a comparable type of model that does capture interaction effects (Friedman 1991).
While many artificial society studies graphically portray the activity space as a mathematical object (such as a lattice or torus), the present discussion has observed in passing that empirically dense social simulations frequently use GIS software to display agents and households in the setting of historically appropriate maps. The present collection is thus enriched by the presence of a study by Lake (2000) that combines an archeologically grounded model of mesolithic foraging with a consideration of how a GIS system is used to incorporate, display and analyse such data.
Lake introduces two software packages: the Multi-Agent Geographically Informed Computer AnaLysis (MAGICAL) simulation package and the Geographical Resources Analysis Support System (GRASS) GIS. GRASS is an open source GIS package distributed under the terms of the GNU public license. It contains a set of programs and scripts for operating on raster and/or vector maps of spatial phenomena. Because of the openness of the GRASS system, it is easily extensible. The MAGICAL package consists of three new GRASS functions that operate as standard GRASS components.
The MAGICAL package can be viewed in two ways. For those interested in constructing models of hunting and gathering societies, MAGICAL provides a pre-constructed framework that facilitates the implementation of a simulation, particularly in the areas of mobility, subsistence and agent decision-making. As Lake (2000), p. 109) illustrates:
... the software allows each individual to rationally calculate the benefits of moving to a particular location in the landscape and to have an energetic state that may be decremented due to the cost of moving, or incremented as a result of resource capture.
Lake provides a fairly in-depth view of the type of genotype (script) that users may develop to specify agent decisions and actions. Alternatively, for researchers who are interested in using GRASS to provide GIS support for different types of simulations, the same detailed description provides an overview of how GRASS may be customised and extended to achieve particular simulation goals by adding compatible modules.
The mesolithic foraging model, presented primarily as a means of demonstrating the MAGICAL software, concerns the extent to which site settlement and exploration on the Scottish island of Islay might be accounted for by initial landing locations followed by subsequent movements motivated by foraging for hazelnuts. The presence of boot camps and other site locations is inferred from the distribution of flint artifacts. The model of hazelnut availability incorporates climate, soil type and the characteristics of the species of tree (ash versus birch versus hazel). The plausibility of the map thus generated is assessed using pollen evidence.
The foraging model incorporates the seasonal cycle, changes in the location of base camps, exchanges of information between group members, and agent decision-making concerning the reduction of risks implied by optimal foraging theory. Agent spatial knowledge was explicitly modelled. Assumptions were varied to explore the likelihood of various possible landing sites. All runs were composed of 216,000 time increments, each representing approximately 10 years of foraging.
The longest contribution to the volume (78 pages) provides an in-depth narrative exploration of a (possible) autonomous transition from a simple to a complex society (Lehner 2000). The design components for the resulting model of centralised nation-state are based upon pharaonic Egypt, more specifically, Giza between 2134 B. C. E. and 905 C. E. While the details of this case are far beyond the scope of the present discussion, its presence further illustrates the commitment of this volume to the deep empirical grounding of social simulation.
Empirical grounding is a strategy that addresses several important challenges facing social simulation as a nascent endeavour. The technical nature of agent simulation has caused the focus to be on software development issues, and remain strongly within the assumptions of methodological individualism (Padgett 2000). Such an emphasis also tends to result in highly creative approaches to social simulation nonetheless failing to fully engage issues that motivate the substantive disciplines. However, the dense models generated in the contributions described above provide an approach that has the potential to overcome such limitations.
Theoretical And Conceptual ContributionsIf one strength of the collection under review resides in its empirically grounded simulation studies, as we have seen, there are also contributions that are more theoretical and conceptual in nature. It is to these that we now turn.
Gilbert (2000), who undertakes to report on and generalise from artificial society research, primarily from Europe, provides a potentially pivotal theoretical contribution. The domain of Gilbert's review is not just the results of completed experiments, but also the space of all possible societies. The resulting methodology is, thus, eidetic in nature. "The assumption is that there are some 'laws' which apply to all societies within this space, and the task of the theorist is to identify these laws" (2000, p. 358).
Within this domain, Gilbert identifies five abstract social processes that have "emergent effects when carried out by any collection of interacting agents, whether computational or human (2000, p. 359). The five processes are:
For each of these processes except the last, Gilbert summarises common consequences and discusses supporting examples.
The identification of abstract laws that may be emergent fromthe social relations of both humans and computational agents is an important contribution, indeed, and the five that Gilbert describes appear to be promising candidates. However, reflection upon these candidates causes this reviewer, at least, to desire more than their narrative description. It would be desirable for example to analyse the empirically grounded simulations described above, in order to assess the presence/absence and effects of such abstract processes. Such a task would be greatly facilitated if they were presented in a notation, or formalism, that specified their form. Such a notation would not only simplify their recognition, but also their comparison, and/or analysis of their interaction, in diverse contexts.
It would also be desirable to identify agent actions that are associated with the operation of such 'laws' including, especially, the (inter)actions that agents (human or otherwise) must take in order to create a new instance of the abstract social process in question. Such a further specification, which would also benefit from appropriate notation, would facilitate the use of these abstract social processes in simulation research and, thereby, potentially provide greater linkages between research and conceptual activities in agent simulation.
One potential benefit that using a formalism to express theoretical generalisations, and as a means of addressing both of these desiderata, is that such a step would at least implicitly acknowledge the issues of knowledge representation that are inherent in social simulation. Major contributions have been made by John Sowa in the specification of processes, contexts and agents (2000, especially pp. 206-347). Following Sowa's lead would suggest that conceptual graphs may play a special role, along with predicate calculus and the Knowledge Interchange Format (KIF), in expressing entities, processes and insights associated with social simulation (2000, pp. 467-491).
Other contributions to the volume under review do introduce formalisms to convey their insights. In one of the few chapters that addresses primate society, te Boekhorst, and Hemelrijk (2000) construct non-linear dynamic systems models of the non-linear dynamics of processes, the simultaneous interaction of units and the effects of local spatial configuration. The authors also draw upon related work in situated robotics, nicely spanning the range of organic and computational agents.
Similarly, Skyrms (2000) introduces game theory to make a subtle point about how inference might have emerged in adaptive settings. The formulation concerns the specificity of species-specific alarm calls. In addition to conveying a general cry of danger, it may be adaptive for animals to convey information about the specific nature of the threat, even when it is possible only to narrow ambiguous possibilities. Inherent in the correlative adaptive response of the receiver is a conceptual basis for the emergence of inference, although such a summary does not adequately convey the elegance of Skyrms' argument.
One contribution to this volume is based upon the (by now) classical form of artificial society research. Pepper and Smuts (2000) experiment with the structure of a two-dimensional grid (torus) in order to investigate the circumstances under which altruistic traits may be selected. More specifically, the size of resource patches and the gap width between such patches is varied in order to explore the circumstances in which group selection may cause altruistic behaviours such as alarm calling and/or feeding restraint to be adaptive. This study, the results of which are promising, is discussed as a conceptual contribution because its setting is defined by conceptual design, rather than the importation of extensive empirical parameters as in the first set of studies. The chapter also serves as an effective example of a study organised around the substantive issues of a discipline, even though the latter is biology rather than social science.
Other than Gilbert, Doran (2000) provides the most broadly theoretical article in the collection. His arguments for moving beyond rationality, and incorporating emotion in agent models are cogent and appropriate. Unfortunately, his concept of emotion is narrow, strongly tending toward sources of misbelief, rather than viewing emotion as a developmental precursor to rationality, and as a heuristic that shapes the way inference is situationally invoked (Brady and Sniderman 1985). Inasmuch as emotion is a vital physiological component of the broader bio-cognitive process (LeDoux 1996, Panksepp 1998, Lane and Nadel 2000), Doran is surely right to argue that well-designed representations of human emotion will be an essential aspect of building effective social simulation models.
Overall, the theoretical and conceptual contributions to the book are solid and insightful. Yet, one objective they do not fulfil, or probably try to fulfil, is to provide a theoretical foundation for the types of empirically grounded study that the remainder of the collection features so prominently. When empirically grounded work is fully absorbed within agent simulation methodology, a new stage of agent simulation research may well have been entered. As suggested above, that absorption probably includes formalising the interchange between artificial society and empirically grounded approaches to social simulation research. This may involve the specification of an intermediate layer that is independent of the implementation, and can be used to "transplant" particular concepts and mechanisms from one social simulation to another. Among these mechanisms maybe agent model design components that express the physiology of emotion in a plausible, if not necessarily detailed, form.
Absorption is also likely to include addressing the new types of validation and docking issues (Burton 1998) posed by the type of empirically grounded research featured here. To the extent that this type of research is heavily customised to the empirical profile of specific cases, the results it can achieve differ sharply from design that is more generalised in nature. These cases can be seen as a test bed relative to which generalised models can be assessed. Alternatively, should innovative models be developed within empirically grounded settings, their generalised assessment provides another focus for subsequent research.
ConclusionAs must be obvious, this book contains a wide range of substantive and insight-provoking articles. It should be on the shelf of any scholar interested in the emergence of social simulation as a distinct approach to the study of human society. In addition to a number of relevant conceptual contributions, it provides an in-depth view of how agent simulation methods can be adapted to the study of empirically rich cases.
One of the editors holds that small societies have unique features that allow agent simulation to be feasible (Kohler 2000, p. 8). However, the simplicity of small societies may be emulated by effective abstraction of complex domains. For example, computational organisation theory has identified three component based and five category based models[5] that may sufficiently abstract the modern workplace domain to support the use of empirically grounded methods as developed here. It seems likely that other complex domains may support comparable types of abstraction. To the extent that such abstraction is viewed as a knowledge representation task, including the utilisation of standard formalisms, the generalisability of such endeavours will be enhanced.
Returning to the question that opened this review, whether this collection exemplifies a new stage in agent simulation research, the answer is probably, "Not yet." But it does give us a glimpse of what that next stage may look like and, in so doing, helps define the tasks that lie ahead.
Notes1 In order to do justice to the range of topics and the number of contributors (25), each article is listed separately in the references.
2 Despite the book title, only one or two of its articles focus upon primate society. The rationale for the emphasis upon small societies is summarised by Kohler (2000, p. 8): "It is felt by many anthropologists that although our ability to model societies is very primitive, that what we are able to do now, and what we can envision being able to do in the near future, approximates much more closely the operation of small-scale societies than it does modern industrial civilizations."
3 For a critique of Axelrod's work from a game theoretic perpective, see Binmore (1998).
4 The waves or generations of agent simulation exemplars identified here are drawn primarily from two areas of social simulation: complex adaptive systems and evolutionary game theory. Parallel developments were occurring in distributed artificial intelligence (multi-agent systems, see (Weiss 1999), demography (microsimulation, see Wachter, Blackwell and Hammel 1997), ecological modelling (individual-based simulation, see (DeAngelis and Gross 1992), and computational organisation theory (see (Carley and Prietula 1994a). Development in each of these areas followed a different pattern. Early and continuing contributions in distributed AI, for example, were primarily in a variety of technical and problem-solving domains (see Bond and Gasser 1988); only later did multi-agent insights begin to be applied in the area of social simulation (Castelfranchi and Werner 1994). It is of inherent interest how the same computational capabilities gave rise to similar innovations in various specialised areas of research. A study providing an overview of the emergence, technical taxonomies and interaction in and among all such areas would be useful in establishing the broad origins of this increasingly significant approach to social research.
5 The three component model is organised around agent, situation and task (Carley and Prietula 1994b). The five category model (COMIT) includes agent, task, action, technology and output (Kaplan and Carley 1998).
ReferencesDE ANGELIS D. L. and L. J. Gross 1992. Individual-Based Models and Approaches in Ecology: Populations, Communities and Ecosystems. Chapman and Hall, New York, NY.
AXELROD R. 1984. The Evolution of Co-operation. Basic Books, New York, NY.
AXELROD R. 1997. The Complexity of Co-operation. Princeton University Press, Princeton, NJ.
BINMORE K. 1998. Review of The complexity of cooperation: Agent-based models of competition and cooperation by Robert Axelrod. Journal of Artificial Societies and Social Simulation, 1:<https://www.jasss.org/JASSS/1/1/review1.html>.
TE BOEKHORST I. J. A. and C. K. Hemelrijk 2000. Nonlinear and synthetic models for primate societies. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
BOND A. H. and L. Gasser, editors, 1988. Readings in Distributed Artificial Intelligence. Morgan Kauffman, San Mateo, CA.
BRADY H. E. and P. M. Sniderman 1985. Attitude attribution: A group basis for political reasoning. American Political Science Review, 79:1061-1076.
BURTON R. 1998. Validating and docking: An overview, summary and challenge. In M. J. Prietula, K. M. Carley and L. Gasser, editors, Simulating Organizations: Computational Models of Institutions and Groups, The M. I. T. Press, Cambridge, MA.
CARLEY K. M. and M. J. Prietula, editors, 1994a. Computational Organization Theory. Lawrence Erlbaum, Hillsdale, NJ.
CARLEY K. M. and M. J. Prietula, 1994b. ACTS Theory: Extending the model of bounded rationality. In K. M. Carley and M. J. Prietula, editors, Computational Organization Theory. Lawrence Erlbaum, Hillsdale, NJ.
CASTELFRANCHI C. and E. Werner, editors, 1994. Artificial Social Systems: Fourth European workshop on Modelling Autonomous Agents. Springer-Verlag, Berlin.
DEAN J. S., G. J. Gumerman, J. M. Epstein, R. L. Axtell, A. C. Swedlund, M. T. Parker and S. McCarroll 2000. Understanding Anasazi culture change through agent-based modeling. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
DORAN J. E. 2000. Trajectories to complexity in artificial societies: Rationality, belief, and emotions. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
EPSTEIN J. M. and R. Axtell 1996. Growing Artificial Societies: Social Science from the Bottom Up. The M. I. T. Press, Cambridge, MA.
FRIEDMAN J. H. 1991. Multivariate adaptive regression splines. The Annals of Statistics, 19:1-141.
GILBERT N. 2000. Modeling sociality: The view from Europe. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
KAPLAN D. J. and K. M. Carley 1998. An approach to modeling communication and information technology in organizations. In M. J. Prietula, K. M. Carley and L. Gasser, editors, Simulating Organizations: Computational Models of Institutions and Groups, The M. I. T. Press, Cambridge, MA.
KOHLER T. A. 2000. Putting social sciences together again: An introduction to the volume. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
KOHLER T. A. and G. J. Gumerman, editors, 2000. Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
KOHLER T. A., J. Kresl, C. Van West, E. Carr and R. Wilshusen 2000. Be there then: A modeling approach to settlement determinants and spatial efficiency among late ancestral Pueblo populations of the Mesa Verde region, U. S. Southwest. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
LAKE M. W. 2000. MAGICAL computer simulation of mesolithic foraging. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
LANE R. D. and L. Nadel 2000. Cognitive Neuroscience of Emotion . Oxford University Press, New York, NY.
LANSING J. S. 1993. Emergent properties of Balinese water temple networks: Co-adaptation on a rugged fitness landscape. In C. G. Langton, editor, Artificial Life III, Addison-Wesley, Reading, MA.
LANSING J. S. 2000. Anti-chaos, common property, and the emergence of cooperation. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
LEDOUX J. 1996. The Emotional Brain: The Mysterious Underpinnings of Emotional Life. Touchstone, New York, NY.
LEHNER M. 2000. Fractal house of Pharaoh: Ancient Egypt as a complex adaptive system, a trial formulation. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
MAYNARD SMITH J. 1982. Evolution and the Theory of Games. Cambridge University Press, New York, NY.
PADGETT J. F. 2000. Closing panel: Results and prospects. In C. M. Macal and D. Sallach, editors, Proceedings of the Workshop on Agent Simulation: Applications, Models and Tools, Argonne National Laboratory, Argonne, IL.
PANKSEPP J. 1998. Affective Neuroscience: The Foundations of Human and Animal Emotions. Oxford University Press, New York, NY.
PEPPER J. W. and B. B. Smuts 2000. The evolution of cooperation in an ecological context: An agent-based model. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
REYNOLDS R. G. 2000. The impact of raiding on settlement patterns in the northern valley of Oaxaca: An approach using decision trees. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
SCHELLING T. C. 1978. Micromotives and Macrobehavior. W. W. Norton, New York, NY.
SKYRMS B. 2000. Evolution of inference. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
SMALL C. A. 2000. The political impact of marriage in a virtual Polynesian society. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
SOWA J. F. 2000. Knowledge Representation: Logical, Philosophical, and Computational Foundations. Brooks/Cole, Pacific Grove, CA.
WACHTER K. W., D. Blackwell and E. A. Hammel 1997. Testing the validity of kinship microsimulation. Journal of Mathematical and Computer Modeling, 26:89-104.
WEISS G., editor, 1999. Multi-agent Systems: A Modern Approach to Distributed Artificial Intelligence. The M. I. T. Press, Cambridge, MA.
VAN WEST C. 1994. Modeling prehistoric agricultural productivity in South Western Colorado: A GIS Approach. Department of Anthropology Reports of Investigations Number 67, Washington State University Pullman, WA.
WRIGHT H. T. 2000. Agent-based modeling of small-scale societies: State of the art and future prospects. In T. A. Kohler and G. J. Gumerman, editors, Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford University Press, New York, NY.
YOUNG H. P. 1998. Individual Strategy and Social Structure: An Evolutionary Theory of Institutions. Princeton University Press, Princeton, NJ.
Return to Contents
of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2001