Order this book
Order this book
University of Essex
The volume edited by Sheila Kohring and Stephanie Wynn-Jones is perhaps the less relevant to JASSS readers of the two. Essentially it is a collection of essays about complexity in archaeological discourse, seemingly deriving from a recent workshop held at the University of Cambridge (although this is not explicitly stated). The collection, by a variety of authors, broadly reflects the at times conflicting traditions of thought at Cambridge fostered by Colin Renfrew and by Ian Hodder. There are examinations of various ideas of the nature of complexity, heterarchy and community and the exploratory use of these concepts in particular archaeological and social anthropological contexts drawn from around the world.
In their initial chapter on "Socialising Complexity" Wynn-Jones and Kohring discuss past and present concepts of (social) complexity, which in their view is too often wrongly equated with social hierarchy. They also revisit the old discussion of heterarchy as contrasted with hierarchy. But they never attempt to be mathematically precise. Readers from a hard science background may feel that these concepts are altogether too fluid and ill defined and that the authors are impeded by their own reluctance to commit to precision. Where the authors do approach modelling of complex systems they find, unexpectedly and surely somewhat optimistically, that compared with the "rather mechanistic models" of the past "current approaches are much more sophisticated and flexible in their modelling, capable of incorporating dissident voices, conflict and multiple belief systems" (page 6). Agent-based modelling as such is not on their radar.
Particularly interesting and illustrative of the volume as a whole is the chapter by Kristian Kristiansen (Chapter 5) addressing the early Bronze Age society of southern Scandinavia. He reviews the archaeological evidence, primarily relating to homesteads and burial mounds, and argues that it supports an interpretation of the society as a "networked decentralised chiefdom" (page 73). This makes it possible, he concludes, to reconcile two traditional and competing interpretations of the evidence as indicating either an aristocratic or an egalitarian society and "to integrate the related theoretical concepts of heterarchy/self-regulation and hierarchy/regulation into a single explanatory framework of decentralised complexity" (page 73). Kristiansen makes no use of formal models - rather he is offering a different way of looking and a persuasive one.
Although a great deal in Kohring and Wynn-Jones's volume is interesting and/or thought provoking, I am left feeling that what is really needed is less a new framework in which to think about complexity but rather (having due regard to the essential limitations of the human mind) practical tools to help us handle complexity of evidence within the actual process of archaeological interpretation.
This brings us to the second and more substantial volume, that edited by Tim Kohler and Sander van der Leeuw, and to formal modeling in archaeology especially computer modeling. This second volume comprises chapters by an international range of authors and derives from a workshop held at the Santa Fe Institute in 2004. It focuses on the use of a "complex systems" orientation and of mental or formal models, where by the latter are meant mathematical or computer simulation models including "agent-based" models.
The two chapters of most immediate interest to JASSS readers, and perhaps objectively of most weight, are those of Tim Kohler and colleagues (Chapter 4) and of Tony Wilkinson and colleagues (Chapter 9). Each of these chapters reports the work of a large team and impresses by the sheer scale of the data gathering exercise that has provided the input to an agent-based simulation model and by the complexity of the model itself. In both models, agents correspond to households and make decisions of various kinds especially concerning location and farming strategy. The project of Kohler and colleagues, is the latest of a number of attempts to model settlement dynamics in South-Western USA going back at least as far as the computer simulation study of Ezra Zubrow (1975). Their model is topped by a version of Bob Reynolds' "cultural algorithms" enabling study of aspects of the accumulation and communication of knowledge by agents. By contrast Wilkinson and colleagues present a detailed case study (Tell Beydar, Syria) of human demography, settlement dynamics, and natural systems in 3rd and 2nd BC millennium Mesopotamia. Their "Enkimdu" model is built within DIAS, a powerful and general object-based simulation framework, and draws upon a wide range of evidence including cuneiform texts. Among other things they simulate a stress scenario - an epidemic specifically targeting children - and even (page 181) optimistically foresee handling within individual agents such "slippery" concepts as rhetorical skill, and charisma which they see as factors underpinning political authority and hence necessary for urbanisation. Neither team yet reports much in the way of systematic experimental results, although seemingly such experiments are in progress or imminent. Significantly, Kohler et al are clear that a close match between their model and the past is not to be expected. These authors emphasise that discrepancies between model behaviour and the past as archaeologically reconstructed reveal where the principles underlying the model are insufficient or where there are important omissions from it (page 103).
Both these studies are sophisticated archaeologically, but perhaps less so as regards ABSS method. There seems to be an uncritical assumption that the more detail that is included, the better the model. There is little evidence that the authors are familiar with the attempts of the ABSS community to clarify such matters as validation, sensitivity analysis and parameter space exploration, and the critical task of choosing an appropriate level of abstraction (or granularity) for a model. Unexpectedly neither of these two agent-based models appears to make much use of stochastic variables even though these are surely a standard way of summarising difficult to specify detail.
Those working with agent-based social simulation will note an important feature of archaeological modelling. The focus is necessarily on just one past instance of a socionatural process from evidence of which the model is to be built and which the model is, in some sense, to explain. There is no question of prediction. Thus the central issue is whether the constraints built into the model allow the properties of the past process to be reproduced if the model's parameters are set plausibly (ie in line with the evidence collected.). Clearly there may be several ways in which the observed results might be obtained from what is, typically, a model with very many parameters, but the foregoing authors pay little attention to this complication.
Other projects reported seem to have the potential of agent-based modelling but as yet are in a pre-modelling state. Both Steve Shennan's impressive chapter (Chapter 7) on the spread of farming into Central Europe in the Neolithic and Serge Cleuziou's (Chapter 10) on the Early Bronze Age in Oman marshal and interpret archaeological and natural environment evidence in a way that could be used to support agent-based modelling. Peter Jordan (Chapter 2) looks at cultural diversification in NW California using NeighborNet plots and Cladistic Analysis and other statistical tests, but as yet makes no attempt at simulation.
In Chapter 3, Jean-Francois Berger, Laure Nuninger, and Sander van der Leeuw report an investigation into the "Middle Rhone Valley" from 800 BC to 800 AD a timespan that includes, of course, the Roman period and the barbarian takeover. They are attempting to integrate a group of theories including complex systems theory, with the long-term aim "to model these dynamics formally". Essentially this is archaeological interpretation, using much climatic and agricultural reconstruction and with significant use of complexity concepts (e.g. "cycles", "hypercoherence").
Several of the remaining chapters do an excellent job of combining different types of formal modelling. Thus Eric Alden Smith and Jung-Kyoo Choi (Chapter 5) report an investigation into the emergence of social inequality. They explore two particular theories, and their investigation is noteworthy for deploying game theory alongside the use of two simple scenarios and abstract agent-based simulation. Although these authors discuss the asymmetries that may underlie social inequality, rather surprisingly they do not consider the possibility of cognitive heterogeneity as a motor. Perhaps our current inability to program genuinely cognitive agents within simulations is a factor here (cf Doran and Palmer 1995).
Marco A Janssen and John M Anderies (Chapter 8) are studying "irrigation societies" as control systems. They have used stylised models to look at two areas of vulnerability: (i) where the controller comprises agents in a social network and (ii) investment in infrastructure. As regards the first of these, the authors revisit, extend and experiment with the well-known Lansing-Kremer model of a Bali irrigation system using (a) agents ("subaks" - collections of farmers) that imitate best cropping patterns from the entire area, and (b) subaks able to learn and adapt their strategies. The authors explore the consequences of pest dispersal connections change. They find that imitative subaks outperform adaptive subaks only in a rather special case (page 163). To look at the vulnerabilities associated with management of a multiple resource system (seen as analogous to portfolio management) Janssen and Anderies have studied in detail the Hohokam socionatural system in the American SW (AD 1 to 1450), including the impressive irrigation infrastructure created in the Sedentary period. They deploy a bioeconomic model of extensive and intensive food production comprising two differential equations incorporating simplifying assumptions, and relate model properties to the actual cultural sequence seeing the society being led into fewer and fewer options. Finally, they discuss whether these vulnerabilities and option reductions played a role in the observed Hohokam cultural collapse.
In Chapter 6 Patrick Kirch and colleagues build on extensive previous work to study emergent cultural complexity as it occurred in the Hawaiian ecosystem from 800 AD to 1800 AD (ie to European contact). They focus on two particular study areas (Kahikinui and Kohala) and provide very substantial reconstructions of human and natural system trajectories using environmental and archaeological evidence. There is no agent based modelling, but many minor models of informal and formal types are deployed, again including some differential equation modelling. This is very worthwhile work of wide significance.
The final chapter by Henry T. Wright is in the nature of an enthusiastic review of the remainder of the volume. Wright provides insightful comments drawing upon his own long experience of formal modeling in archaeology (e.g. Wright and Zeder 1977) and emphasises the progress being made.
Impressive, even exciting, though all this work is, readers should take with a grain of salt the suggestion made by the editors in their Introduction that this is a new way of thinking in archaeology (Chapter 1, pages 2-4). Thus David Clarke (who died lamentably young in 1976) proposed most of the relevant ideas in 1972 in his introduction to the volume "Models in Archaeology" (Clarke 1972)(some in his earlier book "Analytical Archaeology" of 1968) and since then the use of formal models in archaeology has repeatedly been advocated with ever more sophisticated concrete studies, e.g. Doran (1970), Doran and Hodson (1975, chapter 11), Hodder (1978), Renfrew and Cooke (1979), Sabloff (1981), Gardin and Peebles (1992), Gilbert and Doran (1994, chapters 1, 8, 9,10), van der Leeuw and McGlade (1997), Kohler and Gumerman (2000). What is striking in this particular volume is the sheer scale of the agent-based models deployed and of the teams that have built them, and the successful integration of a variety of evidence and of model types, reflecting more available data from the past natural environment and from past human activity, enhanced access to computation, wider acceptance of mathematical modelling as a standard, and significantly more developed understanding of complex systems. Thus rather than revolution this volume documents solid progress and will surely stimulate more such work over the next and subsequent decades.
CLARKE DL (Ed.) (1972) Models in Archaeology. London: Methuen.
DORAN JE (1970) Systems Theory, Computer Simulations and Archaeology. World Archaeology, 1(3), 289-298.
DORAN JE and HODSON FR (1975) Mathematics and Computers in Archaeology. Edinburgh: Edinburgh University Press.
DORAN JE and PALMER M (1995) 'The EOS Project: Integrating Two Models of Palaeolithic Social Change'. In Gilbert N and Conte R (Eds.) Artificial Societies, London: UCL Press, pp 103-125.
GARDIN J-C and PEEBLES CS (Eds.) (1992) Representations in Archaeology. Bloomington: Indiana University Press.
GILBERT N and DORAN JE (Eds.) (1994) Simulating Societies: the Computer Simulation of Social Phenomena. London: UCL Press.
HODDER I (Ed.) (1978) Simulation Studies in Archaeology. Cambridge, MA: Cambridge University Press.
KOHLER TA and GUMERMAN GJ (Eds.) (2000) Dynamics in Human and Primate Societies: Agent-Based Modeling of Social and Spatial Processes. Oxford, UK: Oxford University Press.
RENFREW C and COOKE KL (Eds.) (1979) Transformations: Mathematical Approaches to Culture Change. London: Academic Press.
SABLOFF JA (Ed.) (1981) Simulations in Archaeology. Albuquerque: University of New Mexico Press.
van der LEEUW SE and McGLADE J (1997) Time, Process and Structured Transformations in Archaeology. London: Routledge.
WRIGHT HT and ZEDER M (1977) 'The Simulation of a Linear Exchange System under Equilibrium Conditions'. In Earle TK and Ericson JE (Eds.) Exchange Systems in Prehistory. New York: Academic Press, pp. 233-253.
ZUBROW EBW (1975) Prehistoric Carrying Capacity: a Model. Cummings: Menlo Park.
Return to Contents of this issue
© Copyright Journal of Artificial Societies and Social Simulation, 2008