Accidental complexity in multilevel modeling revisited

Multilevel modeling (MLM) conceptualizes software models as layered architectures of sub-models that are inter-related by the instance-of relation. Conceptually, MLM provides benefits in terms of ontological classification. Pragmatically, based on arguments in knowledge engineering, MLM meaningfully reduces accidental complexity. In this paper, the problem of accidental complexity in MLM is revisited. The paper focuses on the role of the context of type-instance structures on MLM architectures. We analyze factors of accidental complexity in multilevel models, suggest quantitative metrics for these factors, and show how they can be used for guiding MLM rearchitecture transformations. The relevance of the proposed factors and metrics is shown in an experimental study of type-instance contexts in multiple real-world models.