Reviewed by
Jim Doran
University of Essex
The contributors to this thought-provoking volume address two major questions concerning agent-based modelling in archaeology: is such modelling going to play a substantial role in archaeological studies in the future, and, if so, why has it not played much of a role in the forty or so years since it was first proposed (by, amongst others, the present reviewer)? JASSS readers will incline to say "yes" to the first question and to be puzzled by the second. However, the answers offered to these questions by the contributors are diverse and certainly of interest to all working with agent-based modelling in any of the social sciences.The first chapter is an introduction written by Costopoulos, Lake and Gupta. They look back at the history of agent-based modelling in archaeology with its early limitations including the negative impact of the post-processual counter-revolution. They offer a heuristic classification of modelling studies and see an important new landscape in the development of complexity theory which they primarily equate to the concept of emergence. Wobst, a pioneer who carried out some of the earliest successful modelling work in archaeology, then contributes a brief but insightful comment. He sees only a limited potential for agent-based modelling in archaeology and suggests inherent biases in the technique notably a tendency to make human society seem unduly static. Another major problem, he feels, is that archaeologists are rarely at home with the complexity and the numeracy inherent in computer modelling work.
In an excellent third chapter, Lake argues that the gaps are in archaeological theory and method, not in simulation theory and method. He classifies the purposes of archaeological simulation into hypothesis testing, theory building, and the development of methodology and also discusses the different ways in which the success of a modelling a project may be assessed. He sees a need for greater archaeological concern with "multiple levels of causality" and for more attention to uncertainty and to experimental design.
In chapter four Costopoulos looks back to a "vision" of computer simulation in archaeology published (by this reviewer) in 1970 and rightly notes that it envisaged a "general simulation of the archaeological process". He sees this forty year old vision as providing both a "crib" against which subsequent work may be matched, and also as introducing a set of detailed elements that are at last ready to be assembled. Premo contrasts models that emulate with models that explore; the former tend to be overly general but the latter can address "the dynamics of well-defined and well-bounded archaeologically relevant processes". He warns of the interpretive hazards posed by equifinality (i.e. more than one way to a particular outcome) and of non-determinism, and rightly stresses that all models must involves abstraction. Premo also notes the early influence of David Clarke (but cites only Clarke's relatively minor 1973 Antiquity paper).
The remaining two chapters differ. Reynolds, Whallon and colleagues report experiments with an actual agent-based model of alternative hunter-gatherer decision-making strategies, especially alternative resource-sharing strategies, and their impact on population dynamics. Their experimental results support the broad conclusion that information preservation down the generations is the key to success and that "strategies that cultivate expertise" (rather than "egalitarian sharing strategies") usually do better. It is significant that Reynolds has a background in artificial intelligence studies. Finally, Aldenderfer discusses recent important advances in scientific visualisation and geosimulation, arguing that these developments can help overcome ubiquitous difficulty in interpreting an avalanche of results from a modelling experiment. It is unclear whether his proposals encompass interpretation by heuristic algorithms that are non-visual.
All these chapters are interesting and insightful, yet I come away from them feeling that there is little entirely new or fully compelling. Refining a heuristic classification of archaeological modeling studies, for example, can begin to seem like beating about the proverbial bush. Complexity theory certainly promises much, but arguably as yet it lacks relevant substance.
As regards the disappointing history of archaeological modelling, several of the contributors rightly note that a major obstacle has been and continues to be archaeologists' lack of training in the skills needed for computer modelling. Fortunately, limitation of actual computing resources has now abated. A very different problem is a past tendency to fragmentation in the archaeological modelling community with a sometimes surprising lack of mutual awareness. Even more serious, archaeologists too often seem to believe that modelling in their discipline has its own distinctive methodology and modelling objectives, so that the disciplinary boundary, itself largely a social construct, becomes an impermeable barrier.
As regards the future, there is a deep further difficulty that is all too often overlooked. Distinctive human social structures and social processes emerge from distinctive human cognition. But we do not yet know how to model human cognition on a computer in other than relatively superficial and oversimplified ways. Thus we cannot yet experiment with the models that really matter: those that capture more than simple routine cognitive behaviour. Archaeology faces this challenge as do all the social sciences. For help we need to look to developments in artificial intelligence engineering and in cognitive science modelling.
Return to Contents
of this issue