Limits of empirical validation: A review of arguments with respect to social simulation

Output comparison between simulation model and real world reference system is commonly regarded to be the acid test of model credibility. As sound as the comparison-based approach may seem, serious epistemological and methodological qualifications have been made concerning the foundations of the concept, its applicability, and its dependence from the chosen philosophical perspective. The article reviews and reassesses technical and philosophical arguments on the limits of empirical validation with respect to social simulation. The paper is intended to reposition empirical validation for social simulations that are theory-free and non-predictive. The proposed shift is inspired by the recent critical reassessment of significance tests in applied statistics. According to this shift, it is transparency which becomes paramount for the single social simulation project, whereas empirical validation on the macro level is crucial only after meta-analysis of rival simulation models has shown robust findings despite different sets of assumptions.