In this paper, we explore an agent-based simulation model of dressed stone exchange between quarry owners, merchants and construction sites (castles and churches) during the medieval age (X-XIV centuries). The aim of the simulation model is twofold: to build an integrative view of this system and to confront a set of hypotheses, from the least to most specific one. Exploring models of various complexity level enables us to investigate how far do we have to describe the system to observe substansive changes in the market behaviour.Using an incremental process, we compare:– a random only driven exchange market (the selection of a particular dressed stone producer is defined by a uniform probability law)– a spatial interaction system (the selection of a particular dressed stone producer is constrained by distance to the construction site and producing capability of the quarry)– an expert model (the selection of a particular dressed stone producer is constrained by a set of rules proposed by historical and archaeological scientists).Our exploration procedure evaluates the impact of several input parameters (number of construction sites across time, number and type of chalk producers, needs and capabilities, environment size and topology, etc.) on two output variables derived from an origin/destination matrix (the total volume of exchanges and the median distance travelled by chalks) and on the dynamic of the system. We also measure the interactions between the different mechanisms and the variations introduced by new behaviours (ie. sub-models of higher complexity levels).Exploring our incremental model implies to run, describe, and compare hundreds of thousands of simulations using distributed computing [1], each of which having its own specific set of input and output values computed at each time step. This represents a huge amount of data which needs specific computation methods to compare and understand [2, 3, 4]. REFERENCES 1. Romain Reuillon, Mathieu Leclaire, Sebastien Rey-Coyrehourcq, OpenMOLE, a workflow engine specifically tailored for the distributed exploration of simulation models, Future Generation Computer Systems, Volume 29, Issue 8, 2013, https://doi.org/10.1016/j.future.2013.05.003. 2. Chérel, G., Cottineau, C., & Reuillon, R. (2015). Beyond corroboration: Strengthening model validation by looking for unexpected patterns. PloS one, 10(9), e0138212. 3. Cottineau, C., Chapron, P., & Reuillon, R. (2015). Growing models from the bottom up. An evaluation-based incremental modelling method (EBIMM) applied to the simulation of systems of cities. Journal of Artificial Societies and Social Simulation, 18(4), 9. 4. Grimm, V., & Railsback, S. F. (2012). Pattern-oriented modelling: a ‘multi-scope’ for predictive systems ecology. Philosophical Transactions of the Royal Society B: Biological Sciences, 367(1586), 298-310.