The paper explores methodological question of how to evaluate the evidence of simulation models of complex systems. This is the process of model validation. Since simulation models provide a tool for investigating social mechanisms, model validation requires an approach that differs from the analysis of statistical significance of variables. The conceptual difference can be explained by distinguishing clear and distinct terms, a terminology of early modern philosophy. It is suggested that model validation benefits from making use of Max Weber’s concept of adequate causality. This can be investigated by means of sensitivity analysis. A mechanism is causally adequate if the model is not robust under variation. This is illustrated by an example. An analysis of this example reveals the significance of extreme events. In complex systems, extreme events provide evidence of the presence of presumed social mechanisms that are implemented in a model. This implies regarding conflicts as extreme events of societal operations that indicate fundamental social mechanisms.