How does your model represent the system? A note on model fidelity, underspecification, and uncertainty

Model-driven engineering (MDE) was elaborated over two decades ago based on a very limited set of seminal concepts: system, model, metamodel, and seminal relationships (i.e., representation of, and conforms to), as shown in Fig. 1.Since the early development of MDE, the modeling community has produced a large body of knowledge about conformity.This also allows metamodeling techniques to be used as definitions of the modeling languages themselves.Beyond a pure conformity relation between model and metamodel, various detailed "typing" relationships have been explored to support a better decoupling and modularity of models, and thus a certain degree of reuse.Some flexibility in conformity has been explored to support different levels of formality during modeling activities, hence supporting the process from informal to formal modeling.However, very little literature can be found on the representativity of the model with regard to the system under study, raising the question: "How well does a given model actually represent the system?"Indeed, the core modeling activity has been conceptually defined, e.g., by Stachowiack in terms of characteristics of a model: a model is an abstraction of an original (i.e., the "system"), and fulfill a specific purpose with regard to this original.Several relaxations have been identified.For example, the system might still be under construction and does not exist yet, a model may have many