An Agent-Based Model for Simulating Inter-Settlement Trade in Past Societies

Social and computational archaeology focuses largely on the study of past societies and the evolution of human behaviour. At the same time, agent-basedmodels (ABMs) allow the e icientmodeling of human agency, and the quantitative representation and exploration of specific properties and patterns in archaeological information. In this work we put forward a novel agent-based trading model, for simulating the exchange anddistributionof resources across settlements inpast societies. Themodel is part of a broader ABMpopulated with autonomous, utility-seeking agents corresponding to households; with the ability to employ any spatial interaction model of choice. As such, it allows the study of the settlements’ trading ability and power, given their geo-location and their position within the trading network, and the structural properties of the network itself. As a case study we use the Minoan society during the Bronze Age, in the wider area of “Knossos” on the island of Crete, Greece. We instantiate two well-known spatial interaction sub-models, XTENT and Gravity, and conduct a systematic evaluation of the dynamic trading network that is formed over time. Our simulations assess the sustainability of the artificial Minoan society in terms of population size, number and distribution of agent communities, with respect to the available archaeological data and spatial interactionmodel employed; and, further, evaluate the resulting trading network’s structure (centrality, clustering, etc.) and how it a ects inter-settlement organization, providing in the process insights and support for archaeological hypotheses on the settlement organization in place at the time. Our results show that when the trading network is modeled using Gravity, which focuses on the settlements’ “importance” rather than proximity to each other, settlement numbers’ evolution patterns emerge that are similar to the ones that exist in the archaeological record. It can also be inferred by our simulations that a rather dense trading network, without a strict settlement hierarchy, could have emerged during the LateMinoan period, a er the Theran volcanic eruption, a well documented historic catastrophic event. Moreover, it appears that the trading network’s structure and interaction patterns are reversed a er the Theran eruption, when compared to those in e ect in earlier periods.