Separating Deities from Mortals: Model-Based Decisions under Uncertainty
CEES Extra seminar by Samuel Subbey
Most systems (physical or biological) have complex dynamics. A mathematical model is a representation of our perception of such complex dynamics using structures, equations, or algorithms. The models are usually defined in terms of parameters, which are optimized/tuned using real, imprecise observations. In resource management, predictions from such models underpin management decisions. This talk will discuss the appropriateness (or otherwise) of inference based on predictions from single, optimized models, especially for biological and ecological systems. It would be argued that especially for such systems, quantifying uncertainty is more important than identifying perfect models. The talk will discuss issues pertaining to model complexity (parameterization), regularization, and the implication of variability in the dynamical system to model performance. Simple illustration will be presented to exemplify key points.