In my line of work, there is often a management expectation that simulation is a perfect reflection of reality. That is, there is a one-to-one correspondence between everything (e.g., variables) that make up reality and the simulation. In "reality" (pun intended), we merely chose those variables that scientific study has revealed as being the most important (or most sensitive to perturbations), and "ignore" other factors.
In a perfect simulation world (unlimited resources), we might not do this. But in "real" simulation, there is a trade-space for how much you can have, for how much your willing to pay, and for how much time you have to build the simulation (e.g., performance, cost and schedule). In complex simulations, we may not know all of the variables, so "knowledge" may be part of the trade-space as well. Managers seem to understand trade-space in the holistic project management realm, but sometimes fail to see its application when simulation is part of the project.
In simulation, bigger (e.g., more variables) is not always better. If we aim for higher fidelity on a subset of variables, we often get better results. All models of reality are inadequate, if one expects perfect correspondence with reality. However, "good-enough" models (and simulations of them) are useful.
No comments:
Post a Comment