Order this book
Timothy A. Kohler
Washington State University
The book’s 11 chapters are divided into four sections. Mark Lake kicks things off with a meticulously referenced chapter on philosophical issues surrounding ABM in archaeology, providing an extended treatment of rationality in agents and what we mean when we say emergence. His detailed discussion of the adequacy of generative sufficiency as a simulation goal (Epstein 1999) is also welcome. This chapter, plus his recent historical overview of the field (Lake 2014), makes a fine introduction to what modelers in archaeology have done and might hope to do, and our varied motivations.
Next comes a brief review by Alan Swedlund and colleagues of one of the first ABMs in archaeology, Artificial Anasazi; they also report some on-going work with a new-generation “Artificial Long House Valley” model. Janssen’s (2009) paper analyzing the original model is briefly mentioned, but I would have appreciated a more detailed discussion of his finding that the widely published good fit between the observed and modeled populations in that work was driven almost solely by adjusting the carrying capacity of the simulated valley, rather than by any dynamics due to the ABM itself. (My view is that Janssen’s critique serves to clarify the original model more than to detract from its value, since there is excellent evidence from elsewhere in the Southwest that population sizes do respond to agricultural productivity by both migration, and birth rate.)
Part II, Methods, contains four chapters, mostly by computer scientists, characterizing ABMs in general and in archaeology in particular, and discussing the importance of topics such as reproducibility, verification, and validation. Andreas Koch, a geographer, introduces us to geosimulation, comparing GIS, ABM, and cellular automata, discussing Schelling’s classic model of residential segregation along the way. Xavier Rubio-Campillo describes how high performance computing (HPC) environments might help archaeologist-modelers by making it possible to quickly search much larger parameter spaces, to investigate whether there are relationships between (large) spatial or demographic scales and aspects of model behavior, and also to “solve the problem of self-fulfilling prophecies.” Here he urges us to model agent goals rather than behaviors, letting the agents decide (in a sense) how to achieve those goals, perhaps through the use of Markov Decision Processes. The CPU-intensive demands of such approaches should, as he points out, be easily within the reach of HPC environments.
Part III provides four case studies, beginning with a project by Kerstin Kowarik and colleagues to model mining in the archaeologically famous Hallstatt salt mines of upper Austria. Their models are beginning to refine understanding of questions such as how long the mines might have been in use, how many people could have worked them simultaneously (not very many, it turns out) and the division of labor that would have been optimal for the operations.
Models vary along several dimensions, but perhaps the most important is an axis from abstract and theoretical to realistic, detailed, and descriptive. The next chapter, by Enrico Crema, is the best example in this volume of the first type. Crema explores how group fission and fusion dynamics (for which variables such as frequency of decision-making, disturbance, carrying capacity, and spatial range of interaction serve as immediate causes) result in changes in settlement rank-size distributions—one of the most famous descriptive models in geography but relevant as well to a newer generation of scaling studies (e.g., Kohler 2012; Ortman et al. 2015). Among his many intriguing results is the sensitivity of primate (concave) rank-size distributions to degree of connectivity in the system; they cannot appear without high connectivity, but high connectivity also increases their instability. Presumably, in the real world, considerations of property rights and systems of inheritance, and other constraints on mobility due to cultural barriers, and sociopolitical factors including the possibility of networking returns from creativity and cooperation in large settlements, would tend to stabilize large aggregates somewhat more than is seen in these simulations—which at a minimum demonstrates how simulations that do not match our view of the world can simultaneously cause us to question our “knowledge” while also throwing into relief the importance of those factors possibly causing the real world to be different from the simulations.
The other two chapters in this section report on somewhat less theoretical but more historically situated attempts to understand the late Iron Age economy surrounding the oppidum of Staré Hradisko (Czech Republic) by Alžbĕta Danielsová and colleagues, and the roles of (among other things) labor exchange (cooperation in hunting) and mobility on identity formation (e.g., ethnicity) among hunter-gatherers in Patagonia (by Joan A. Barceló and colleagues). Danielsová and colleagues conclude, among other things, that population sizes reached in this oppidum (occupied for a little over 200 years) could plausibly have outstripped the carrying capacity of its catchment, contributing to its eventual decline. Barceló and colleagues’ paper suggests that considerations such as the expected benefits of identity similarity and of cooperative labor in specific settings likely affect the emergence of territoriality, as they do in the model.
Finally, André Costopoulos provides a conclusion by tracing the genealogies of the realist-particularist and the abstract-generalist approaches that can be seen in this volume (for example, the Kowarik and Danielsová teams on the one hand, and Crema on the other). This is not a classic discussant chapter—he does not engage the preceding chapters—but he does make provocative connections between their contrasting approaches and larger movements in the sciences.
As a convicted realist-particularist, though, I think he misses one powerful argument in favor of such approaches. Abstract-generalist models cannot really be examined against empirical data since they are not “fit” to any chunk of time or space whose archaeological record could be compared with model behavior. General models (for example, for the development of leadership) can however be fit to specific chunks of time or space using the realist-particularist framework, allowing their degree of fit to that context to be assessed. (Whether that examination is convincing is to some extent a function of the resolution of the archaeological record in question.) As archaeology matures in its use of models—a process to which this volume contributes—I can imagine modelers beginning with very stylized approaches to verify model behavior and build theory concerning the interactions among key processes and constraints. This theory can then be instantiated in several different places, perhaps by different teams, all employing what realism is required to fit the general model to their particular contexts. In this way modeling will become the new norm for both constructing and assessing theory. This will be called model-based archaeology.
JANSSEN, M. A. (2009). Understanding Artificial Anasazi. Journal of Artificial Societies and Social Simulation, 12 (4) 13: http://jasss.soc.surrey.ac.uk/12/4/13.html.
KOHLER, T. A. (2012). Complex Systems and Archaeology. In Archaeological Theory Today, 2nd edition, Ian Hodder (ed.), pp. 93-123. Polity Press, Cambridge.
LAKE, M. W. (2014). Trends in Archaeological Simulation. Journal of Archaeological Method and Theory, 21: 258–287.
ORTMAN, S. G., Cabaniss, A. H. F., Sturm, J. O. and Bettencourt, L. M. A. (2015). Settlement scaling and increasing returns in an ancient society. Science Advances 2015(1): e1400066.
Return to Contents of this issue
© Copyright JASSS, 2015