AI-Supported Oral Language Programmes: A Function-Allocation Framework for Understanding Instructional Authority and Intercultural Competence Outcomes

Abstract

Artificial intelligence (AI)-supported language learning systems are increasingly used in educational contexts, yet the allocation of instructional authority between automated systems and teachers remains insufficiently understood. This study examined eight institutional implementations of AI-supported oral language programmes across the United States, Japan, and China, involving 7,350 students. A convergent parallel mixed-methods design was used to classify implementations through a two-dimensional function-allocation framework encompassing inferential depth and curricular integration. Intercultural communicative competence outcomes were assessed using the Intercultural Development Inventory in six institutions and alternative validated instruments in two institutions, with pre-to-post change standardised as Cohen’s d. The qualitative component drew on programme documentation and interviews with programme directors. Across cases, effect sizes ranged from 0.18 to 0.43. Programmes with Embedded curricular integration showed larger median gains (d = 0.41) than Supplementary programmes (d = 0.21), while Integrated programmes showed intermediate gains (d = 0.32). Inferential depth did not display a comparable monotonic pattern. Qualitative findings indicated that the instructional value of AI-supported practice depended on curricular embedding, assessment alignment, validation practices, and continued human oversight, particularly for consequential judgements. These findings support a function-allocation perspective in which instructional outcomes are associated not simply with AI capability but with how automated functions are integrated into the curriculum and governed alongside teacher authority.