Author(s): Isaac Rabago, Tzu-Yu Weng
Mentor(s): Xinru Page, Nancy Fulda
Institution BYU
While online technology connects people worldwide, norms around privacy, appropriate interactions, and information sharing vary significantly from culture to culture. In this study, we explored OpenAI Generative Pre-trained Transformer 4 Omni’s capability in understanding and analyzing cross-cultural conflicts or misunderstandings. Using experiences from 37 interviewees, four researchers from China, Mexico, Costa Rica, and the United States developed 32 scenarios across cultural dimensions of Hofstede’s, Tall’s, and Trompenaars’s cultural models. Half of the scenarios had cross-cultural conflicts, and half were conflict-free. We used three guiding questions to assess the model's performance: Is there a misunderstanding of cross-cultural communication in this scenario? What is the misunderstanding in that scenario? Which cultural dimension does this misunderstanding fit into? In identifying and classifying misunderstandings, the model showed 63% accuracy in conflict scenarios and 52% in non-conflict scenarios.