AI-driven vs. Traditional language assessment: effects on Iranian EFL learners’ motivation, anxiety, and proficiency in a high-stakes exam context

Abstract

Iran’s Konkur exam (national university entrance test) assesses EFL proficiency solely through multiple-choice items, neglecting writing/speaking despite their academic importance. This study compares Intelligent Computer-Assisted Language Assessment (ICALA) and traditional assessments to address this gap. This 12-week mixed-methods study examined how ICALA affected motivation, anxiety, and proficiency in 120 intermediate Iranian EFL learners (CEFR B1–B2). The experimental group (n = 60) used ICALA via DeepSeek, while the control group (n = 60) received traditional instructor-led assessments with identical tasks (250-word essays, 2-min oral responses). Quantitative data from standardized measures (motivation, anxiety, and proficiency scales) and qualitative data from interviews and reflective journals were analyzed. ICALA demonstrated stronger benefits for motivation, anxiety reduction, and proficiency gains compared to traditional assessments, particularly among upper-intermediate (B2) learners. Qualitative analysis revealed three dominant themes: (1) enhanced competence through specific feedback, (2) reduced evaluation pressure, and (3) systematic skill improvement. While B2 learners thrived with ICALA’s detailed feedback (e.g., cohesion suggestions), some B1 learners required simplified guidance due to cognitive load. Although Konkur omits productive skills, ICALA improves writing and speaking proficiency, bridging the gap between exam preparation and academic needs. Simplified feedback for B1 learners, along with balanced speaking tasks, could further enhance outcomes. These findings inform EFL instruction reform in Konkur-driven contexts and contribute to Asia–Pacific and global AI assessment research.