Detail:Individuals from diverse cultural backgrounds are increasingly called upon to work collaboratively online, be it in the classroom or the workplace. However, predicting when such multi-cultural groups might be successful in accomplishing a given task and how cultural factors might affect success remains an open problem. Understanding why certain multi-cultural groups are successful while others are not is critical to devising proper mediation strategies that maximize success. Scholars in sociology and anthropology conduct qualitative studies to determine which cultural factors correlate highly with successful group work; however, such approaches are quite labor-intensive, and also not scalable given the multiplicity of cultures intrinsic to human society. By contrast, we develop an approach where appropriate strategies for cross-cultural communication can be automatically learnt from the sociolinguistic content of online conversational data. We do so by combining established theories of culture by Hofstede (2001), who proposed six dimensions along which multiple cultural backgrounds can be evaluated, and incorporating these dimensions in computational models to predict which groups will be successful using machine learning methods – specifically, Hidden Markov Models.

Publications from this project

C47