CURRICULUM-ALIGNED BOUNDED DIFFERENTIATION IN AI-SUPPORTED INTERCULTURAL LEARNING: A MULTI-SITE STUDY ACROSS THREE EFL HIGHER EDUCATION CONTEXTS

Abstract

This study examines whether a curriculum-aligned methodology for AI-supported intercultural learning can operate across higher education institutions with comparable learning outcomes but different organisational conditions. The intervention was implemented within Country Studies courses at three Uzbek universities: Uzbekistan State World Languages University, Samarkand State Institute of Foreign Languages, and Namangan State Institute of Foreign Languages. The three programmes shared a four-credit disciplinary context and intercultural learning potential but differed in semester placement and independent-learning workload: 72 hours at the first two institutions and 60 hours at the third. A quasi-experimental design involved 316 students, with 158 in the experimental condition and 158 in the control condition. The methodology used bounded differentiation: common curricular coverage and minimum evidence across four dimensions of intercultural competence were retained, while students received required, recommended, and optional task pathways based on diagnostic profiles. Experimental-group post-test means were 79.2, 78.4, and 78.1 across the three institutions, compared with control-group means of 65.2, 64.8, and 64.7. Experimental gains were 18.4, 18.1, and 17.7 points, whereas control gains were 4.8, 4.7, and 4.6. The consistency of the pattern across 72-hour and 60-hour independent-learning arrangements suggests that curriculum alignment and bounded differentiation can support methodological portability without forcing institutional homogenisation. The study proposes a transferable design logic for AI-supported EFL learning in which technology remains subordinate to curricular outcomes, learner responsibility, diagnostic differentiation, and teacher facilitation.