This paper presents results of extended experimentation with selected models of social phenomena widely used by the system dynamicists in their studies on deterministic chaos. The models selected include various versions of a simple model of migratory dynamics and a model of resource allocation in a firm, and a simple model of long-term economic fluctuations. Chaotic modes seem to appear in each of the experimented model, either due to non-robust or unrealistic rate formulation, or from unrealistic parameter or input specifications or both. Minor changes on the models experimented with, which improve their correspondence to reality, eliminate chaotic modes. The paper raises the issue of the relevance of the chaotic models to real-world phenomena and policy design for system improvement.