Polymers filled with either organic inclusions (carbon black, carbon fiber), or mineral inclusions (silica, kaolin, BaTiO3) constitute one of the most exciting fields of research in materials science. The description of these multiscale systems is part of the physics of finely divided matter. The specific properties (percolation, fractal structure, elasticity network) of these materials, along with the hierarchy of length scales they contain, have been widely investigated, either theoretically or experimentally, and are actually used in a broad range of technologies (current limiters, radar absorbers, self-heating cables, ...). One aim of the present study is to highlight many important questions which remain to be answered in order that the modelling of filled polymers can be effective in practice both for the understanding of the basic underlying physics and the optimization of industrial processes. In this article, we show that: (1) the use of a numerical simulation code with several assumptions (quasistatic approximation, periodic assembly of inclusions within a host matrix, no inclusions/matrix interactions) indicates that the shape and volume fraction of inclusions have a strong influence on the permittivity of these materials; (2) the equilibrium sorption kinetics in an excess of solvent of an epoxy resin filled with carbon black gives information about the porosity and the tortuosity of the mesostructure; (3) while the mesostructure and the EPR (electron paramagnetic resonance) response are very different in epoxy resin samples filled either with carbon black or carbon fiber, the permittivity measured in the microwave region can be well described on the basis of Jonscher's model, i.e. by power laws having similar exponents for the different types of inclusions. Gaining a better knowledge of the first principles governing the engineering of filled polymers requires to characterize the physical and physico-chemical properties of these materials along with the detailed determination of their mesostructure. Within this context, controling the interfaces at the nanometre scale is required to make any progress from a fundamental standpoint.