1 of the main challenges to using computational laboratories to model physical reality is the paradoxical task of taking uncertainty into accurate quantitative account

In spite of continued growth in computational resources the idea of a computational laboratory serving as surrogate to physical reality still faces conceptual and technical challenges. Chief among these is the characterization of physical reality itself, under conditions of incomplete knowledge and information, reflected in observed variability and fluctuations. The quantification of this uncertainty has, in recent years, grown from a collection of scientific ideas into a sub-discipline of computational science that attempts to provide a quantitative description of incomplete knowledge for use in conjunction with model-based computational resources, algorithms, and software.

Uncertainty exists in all branches of science and engineering. Accordingly, in recent reports and initiatives on scientific computing, uncertainty quantification (UQ) has been recognized as a critical element necessary for continued advancement in prediction science, life-cycle design, and societal sustainability. The topic of uncertainty, in general, remains rather nebulous and susceptible to philosophical arguments. Significant progress has been made in recent years within a subset of related problems in science and engineering, namely those for which the behavior can be suitably modeled with conservation or variational laws containing stochastic coefficients. Also in recent years, there has been an increasing awareness of complexity as an essential theoretical challenge in many problems of great societal relevance the hallmarks of which are interacting phenomena, nonlinearities and emergent behavior. Network Science has been developing in response to these challenges and has gained both in mathematical maturity and scope of applicability. In turn, developments in network science have spurred significant research activity in computational social sciences and in particular social networks.

This Workshop on Opportunities and Challenges in Uncertainty Quantification for Complex Interacting Systems, will provide a forum where issues of uncertainty quantification and model validation in predictive science will be addressed. The Workshop will bring together leading scientists from physics-based modeling, network science and social networks to explore the fundamental similarities and differences in the challenges facing them. Challenges and opportunities will be identified and a community of researchers and collaborators seeded.

Challenges with characterizing and propagating uncertainty, and validating predictions permeate network science in general and social networks in particular. Metrics for validation and mathematical constructs that are useful for describing uncertainties are lacking, together with a practical interpretation of uncertainty and predictability. It is widely believed that challenges in the verification and validation of network science and social networks presents a serious impediment to the utility of these methodologies as incisive tools for decision support and societal well-being.

The Workshop will revolve around survey lectures and break-out sessions and will culminate in a report that summarizes participants perspective on challenges and opportunities in developing a rational path forward for Uncertainty Quantification and Model Validation in the context of complex interacting systems.

The Workshop will be chaired by Roger Ghanem and Demetri Spanos of USC, and will be hosted in the Davidson Center on the Campus of the University of Southern California.