Approaches compared

This page contains a brief outline of its final contents, which will be completed in 2018.

  • In modeling practice, modeling paradigms represent the “metaphor” adopted during model design: for example, process-based, agent-based, stochastic/probabilistic or deterministic models, spatial models, etc. Entire schools of thoughts, books, courses and conferences are dedicated to single modeling paradigms, with specialized software tools for each.
  • While great inroads have been made in all of these areas, it has remained extremely difficult to mix and match models incarnating different paradigms. Their differences usually are extreme, and span the entire lifecycle of an application, from initial design to working simulation. This has often meant that complex problems are handled with one paradigm that perfectly fits one or more components of the modeled system, but must be “tricked” into unnatural behaviors in order to handle the rest.
  • Paradigms essentially consist of choosing specific, constraining semantics for the observables involved in a simulation (process-based, agent-based), the context (dynamic, spatial models) or the mode of observation (stochastic, deterministic, continuous or discrete time and space). In a semantics-first workflow, where semantics drive the design of the modeled system, different paradigms can peacefully coexist and a single paradigm does not need to be chosen a priori. We can model agents moving through a world subject to processes, each using stochastic or deterministic implementations according to what is best for the context.
  • Because required data and models are automatically integrated, each individual observable can be scaled automatically, independently and appropriately in space and time. Multiple-scale models are thus a natural outcome of semantically aware modeling. Scale is, after all, the defining element of the worldview, which by design ensures the avoidance of incompatible statements and clear, well-defined scale mappings and transformations. By allowing each agent in a model to view the world through the lens of their own scale, more accurate models can be built and the effects of observational perspective on results can be explored.
  • Examples.