Development and application of a generative-AI-driven virtual simulation teaching system for international business negotiation scenarios

Abstract

Abstract Conventional international business negotiation courses struggle to deliver authentic cross-cultural interaction, immediate individualized feedback, and scenario variety at scale. This study designs and deploys a virtual simulation teaching system that positions generative artificial intelligence as a culturally differentiated counterpart rather than as a tutor. The system rests on a five-layer architecture that couples a multi-agent role generation framework, a retrieval-augmented case knowledge base for dynamic scenario construction, and an intelligent assessment module producing real-time narrative feedback anchored in a twelve-indicator rubric. A single-institution controlled experiment enrolled 124 third-year international trade undergraduates and compared four negotiation scenarios across a fifteen-week semester; primary outcomes were derived from blind human rubric scoring rather than from the system’s own indicators. After adjusting for baseline covariates and total practice time, the experimental group still outperformed the control group on every competence dimension, with the largest adjusted effects observed for cross-cultural sensitivity (Cohen’s d = 1.84, 95% CI [1.42, 2.26]) and strategy diversity (d = 1.69, 95% CI [1.28, 2.10]). Engagement telemetry recorded a 41% increase in voluntary practice sessions above the assigned schedule, and ablation studies located retrieval grounding as the single largest module contributor to output quality. Learner satisfaction reached 4.34 on a five-point scale, with self-reported retention holding at six-month follow-up in the experimental group. These initial findings—obtained in one cohort of Chinese undergraduates—suggest that generative AI may extend the constructivist promise of virtual simulation into interactional domains previously bottlenecked by partner availability, with gains concentrated where conventional pedagogy tends to underperform; broader validation across institutions, disciplines, and cultural contexts is still required before stronger claims can be made.