AIMSS: An Architecture for Data Driven Simulations in the Social Sciences

This paper presents a prototype implementation of an intelligent assistance architecture for data-driven simulation specialising in qualitative data in the social sciences. The assistant architecture semi-automates an iterative sequence in which an initial simulation is interpreted and compared with real-world observations. The simulation is then adapted so that it more closely fits the observations, while at the same time the data collection may be adjusted to reduce uncertainty. For our prototype, we have developed a simplified agent-based simulation as part of a social science case study involving decisions about housing. Real-world data on the behaviour of actual households is also available. The automation of the data-driven modelling process requires content interpretation of both the simulation and the corresponding real-world data. The paper discusses the use of Association Rule Mining to produce general logical statements about the simulation and data content and the applicability of logical consistency checking to detect observations that refute the simulation predictions.