In many simulations in fluid and solid mechanics but even in molecular simulations there are many sources of uncertainty, anomalous dynamics and memory effects, e.g. associated with material properties, boundary conditions and strong heterogeneity or confinement. These phenomena cannot be captured with the standard tools of computational mechanics. Such effects may contribute to large errors in simulation, often much larger than the spatio-tempral errors, leading to erroneous dynamics or performance predictions. We will present new stochastic modeling approaches as well as deterministic fractional models that provide more flexibility and possibly greater rigor in quantifying and predicting such phenomena in large-scale multi-physics simulations.