The X-MAS SYSTEM: Toward Simulation Systems for Cross-model-validation in Multiagent-Based Simulations

Abstract

The huge number of simulation models available in different scientific communities shows the prominent role that simulations are playing in the study of complex social systems. However, model validation is not an established practice among communities, but it is indispensable for assuring reliable models and results when studying a certain social phenomenon. In order to help researchers in validation processes, this paper proposes the Cross-model validation for MultiAgent-based Simulation (X-MAS) System, a toolkit developed for supporting validation of models and facilitating the implementation of complex social system simulations by addressing the following three aspects: (1) rich object-oriented library for cross-model validation, simultaneously providing X-MAS with verification and validation capabilities. For validation purpose, X-MAS supplies with an agent structure embedding several kind of elements, such as different learning mechanisms and knowledge representation schemes, (2) high-level programming skills are not required for rapid prototyping, and (3) framework facilities for the promotion of an effective cumulative scientific process, making it possible to evaluate and verify different models, permitting their exchange from different scientific communities, and stimulating the replication of results and their further verification and validation. The effectiveness of X-MAS is shown by investigating a bargaining model, a well-study model in game theory.