<![CDATA[This study investigates the quantitative impact of Gemini AI-enhanced e-portfolio assessment on English speaking performance among ESP students in a vocational education context. Despite the growing integration of artificial intelligence in language education, empirical evidence examining the effectiveness of multimodal AI systems on speaking skill development remains limited, particularly within ESP contexts where students require targeted professional language competencies. Using a one-group pretest-posttest design, 30 second-semester International Business Management students at Politeknik Negeri Bali participated in a 14-week intervention utilizing Gemini AI-integrated e-portfolio assessment. Speaking performance was measured through comprehensive assessments based on the Complexity, Accuracy, and Fluency (CAF) framework, incorporating formal business presentations and structured role-play negotiation scenarios representative of authentic professional discourse demands. Results from paired samples t-test analysis revealed statistically significant improvements following the intervention (t (29) = 15.847, p < 0.001, Cohen's d = 2.89), with mean speaking scores increasing substantially from 62.45 (SD = 8.23) to 78.67 (SD = 6.45). Component-level analysis demonstrated differential yet consistently strong effects across all CAF dimensions: fluency (d = 2.52), complexity (d = 2.27), and accuracy (d = 1.87). The very large effect sizes indicate substantial practical significance, exceeding effect sizes reported in previous technology-enhanced speaking interventions by factors of four to eight. These findings contribute critical empirical evidence supporting AI integration in vocational language education and suggest that Gemini AI-enhanced e-portfolio assessment represents a paradigm shift in ESP pedagogy, with important implications for curriculum redesign and institutional technology adoption decisions in professional preparation programs. Â ]]>
Copyrights © 2026
