MERCURY: an Agent-Based Model of Tableware Trade in the Roman East

A large number of complex hypotheses exists that aim to explain aspects of the Roman economy, consisting of many explanatory factors that are argued to affect each other. Such complex hypotheses cannot be compared or tested through the traditional practice of qualitative argumentation and comparison with selected small sets of written and material sources alone. Moreover, these hypotheses often draw on different conceptual frameworks to abstract the same past phenomenon under study, hampering formal comparison. There is a need in the study of the Roman economy for more formal computational modelling for representing and comparing the many existing conceptual models, and for testing their ability to explain patterns observed in archaeological data where possible. This paper aims to address this need. It argues that communicating the potential contribution of computational modelling to scholars of the Roman economy should focus on providing theoretically well-founded arguments for the selection of the included and excluded variables, the conceptualisation used, and to address those elements of conceptual models that are at the forefront of scholarly debates. This approach is illustrated in this paper through MERCURY (Market Economy and Roman Ceramics Redistribution, after the Roman patron god of commerce), an agent-based model (ABM) of ceramic tableware trade in the Roman East. MERCURY presents a representation of two conflicting conceptual models of the degree of market integration in the Roman Empire, both of which serve as potential explanations for the empirically observed strong differences in the distribution patterns of tablewares. This paper illustrates how concepts derived from network science can be used to abstract both conceptual models, to implement these in an ABM and to formally compare them. The results of experiments with MERCURY suggest that limited degrees of market integration are unlikely to result in wide tableware distributions and strong differences between the tableware distributions. We conclude that in order for the discussion on the functioning of the Roman economy to progress, authors of conceptual models should (a) clearly define the concepts used and discuss exactly how these differ from the concepts used by others, (b) make explicit how these concepts can be represented as data, (c) describe the expected behaviour of the system using the defined concepts, (d) describe the expected data patterns resulting from this behaviour, and (d) define how (if at all) archaeological and historical sources can be used as reflections or proxies of these expected data patterns.