AI-mediated dynamic assessment in EFL presentation courses

Abstract

While generative artificial intelligence (AI) tools such as ChatGPT and Gemini have attracted growing attention in language education, empirical research examining their role in oral communication and presentation skills development remains limited. Drawing on Dynamic Assessment (DA) theory and Vygotsky's sociocultural framework, this study investigates how generative AI can function as a mediator in EFL presentation courses. A mixed-methods design was employed with 84 undergraduate students enrolled in two semester-long EFL presentation courses at a private university in Japan. Participants completed an 11-item Likert-scale survey alongside three open-ended questions upon course completion. Course presentation rubric scores (N = 98) were collected at two time points as a secondary performance data source, enabling triangulation of self-reported perceptions with observable competence indicators. Quantitative results revealed strong gains in vocabulary expansion (M = 4.10, SD = 0.94) and structural understanding of the OREO framework (M = 3.95, SD = 1.05), while self-correction (M = 3.87, SD = 0.90) and weakness awareness (M = 3.69, SD = 0.94) items indicated meaningful engagement with DA principles. Thematic analysis of open-ended responses identified three salient themes: AI as a reflective partner, heightened awareness of audience perspective, and confidence building through iterative feedback. These findings suggest that generative AI, when systematically embedded within a DA framework, can serve as an effective pedagogical mediator that complements teacher-led instruction and fosters learner autonomy in EFL presentation contexts.