On the Relationship between Model Abstraction and Causality: Variance of Causal Ordering under Abstraction Operations

This paper examines the relationship between causal ordering and two temporal abstraction techniques, namely generation of mixed models from dynamic models and aggregation of nearly decomposable models. The paper defines two notions of causal equivalence between an original model and its abstraction or its elaboration. Then, it shows that when the constraint for a uniform temporal grain size for a model is enforced, both elaborating and abstracting a model can significantly alter the causal ordering.