With the increasing information-intensiveness of products, users are challenged with expanding options and possible ways to interact. Rapidly escalating numbers of possible useroperation sequences hinder designers in anticipating all possible (unacceptable) outcomes. Interactively simulating product models with human subjects to explore all options is not practicable. Virtual simulation with computer models of users can open the way towards faster-than-real-time performance and investigation of massive numbers of interaction sequences. This paper reports on opportunities to improve realism of virtual-use simulations by incorporating knowledge about the workings of the human brain We elaborate how, in particular, cognitive-architecture simulations developed by cognitive scientists and error phenotypes identified in human reliability analysis (HRA) can extend a virtual-use simulation approach that we have proposed in foregoing work, by offering the prospective of generating interaction sequences with erroneous user actions unforeseen by the designer. We outline how such an integrated system can be implemented and also discuss validation issues.