Modules in Modeling

Abstract

In the modular approach we do not intend to design a unique general model. Instead, the goal is to offer a framework that can be easily extended and is flexible to be modified. A module that performs best in one case may not be sufficient in another. The goals and scale of a particular study may require a completely different set of modules that will be invoked and further translated into a working model. There is a certain disparity between the software developer and the researcher views upon models and modules. For a software developer, a module is an entity, a black box, which should be as independent as possible, and should be as easy as possible to combine with other modules. This is especially true for the federated approach to modular modeling and is well demonstrated by the web-based modeling systems. For a researcher a model is predominantly a tool for understanding the system. By plugging together a number of black boxes, for which specifics and behavior is obscure and hardly understood, we do not significantly increase our knowledge about the system. The results generated are difficult to interpret when there is not enough understanding of the processes that are actually modeled. The decomposition of such systems requires careful analysis of spatial and temporal scales of processes considered and is very closely related to specific goals of the model built. In this context the modular approach can be useful if the focus is shifted from reusability and ‘plug-and-play’, to transparency, analysis and hierarchical description of various processes and system components. With the modules being transparent and open for experiment and analysis, the researcher can better understand the specifics of the model formalism that is inherited. It is then easier to decide whether a module is suitable or if it should be modified and tuned to the specific goals of a particular study. Modular systems thinking is the way to achieve a number of critical goals: (1) we understand a system better since we can look inside of the system over multiple levels of encapsulation and conceptual abstraction; (2) we foster an economy based on modular systems, allowing exchange of objects via a web-based ‘module repository’; and (3) we are gradually making modules capable of being interacted with via special interfaces (i.e., collaborative modeling).