ABSTRACT
Social science research is central to creating computer-mediated systems that teach cross-cultural competencies. In the HSCB CultureCom project, which uses formal microsocial models to improve artificially intelligent software agents, ethnographic and sociolinguistic research refined the formal model and produced annotated, decision-branching dialogs that served as coding input. This paper describes the anthropological methods used to develop and validate project data, and shows how the accumulation of subtle decisions and linguistic interpretations in cross-cultural encounters can lead to dramatically different outcomes.
