The prediction of the earthquake shaking that could occur at a site from a nearby earthquake is a vital part of seismic hazard assessment. Both the median ground motion and the variation about this median (aleatory variability) need to be assessed. Of equal importance is an estimate of how confident we are about these individual models (epistemic uncertainty). Significant improvements in the models used to make these estimates have been made in the past decade. There remain, however, areas ripe for developments. For example, improved techniques for ground-motion model development; the estimation of ground motions in low seismicity areas: from model development to adjustments of existing models; the assessment of the epistemic uncertainty in ground-motion prediction and its reduction; the robust separation of aleatory ground motion variability into its constitutive parts; and the adjustment of ground-motion estimates to site-specific conditions. In light of increasing computer power and improved numerical modelling techniques, the balance between robust estimates and the use of physics-based simulations is an important topic that also needs to be addressed. We encourage submissions on all aspects of ground-motion prediction, especially those that seek to go beyond the state-of-the-art (even if the ideas are not yet fully developed).