Computerized training systems may be the only viable solution to accommodate the quickly evolving learning needs, the short lead time, and the large number of students that must prepare for cross-cultural interactions. However, current systems place emphasis on language training, which is only part of the equation in effective communication. Knowledge about culturally dependent social norms directly impacts the interaction outcome. A dynamic, customizable, computationally tractable model for the code of conduct, or etiquette, can alert us of the specific communication protocols of groups of interest, and can prevent over-generalizations which can be detrimental to the end goal. We may not consciously care, or even be aware of etiquette, but it impacts our trust and affect for individuals, and ultimately shapes our subsequent actions. We have adapted a socio-linguistic model of human-human interaction for use in language and culture training. This computational model increases the social intelligence of virtual actors by allowing them to act and react to human students in ways that are more aligned with the humans they simulate. Capabilities for virtual agents to reason about and select actions not only enhances interaction fidelity, but can also result in time and cost savings in training scenario generation and maintenance.