Optimizing AI-Based Adaptive English Language Learning for Students with Special Educational Needs

Abstract

This study investigates the design and implementation of an AI-based adaptive English as a Foreign Language (EFL) model for students with special educational needs (SEN) in a resource-constrained Indonesian special school. The research aimed to align adaptive technology with heterogeneous learner profiles and priority English competencies to enhance accessibility, engagement, and learning outcomes. A descriptive qualitative design with supportive quantitative elements was employed, combining classroom observations, semi-structured interviews with teachers and teaching assistants, analysis of individual education plans and curriculum documents, and learning analytics from an adaptive AI platform, collected across multiple SEN classes (N = 4 students and 3 teachers/assistants) over 16 weeks. The findings from the needs analysis confirmed substantial variation in cognitive levels, sensory modalities, and communication preferences, leading to the specification of target competencies in foundational vocabulary, basic oral interaction, and early literacy. The selected AI platform, configured according to Universal Design for Learning principles, delivered multimodal materials, adjustable difficulty, and locally meaningful scenarios within blended and station-rotation lesson structures. This architecture, organised into warm-up, AI-driven core tasks, and consolidation phases, was associated with increased on-task behaviour, more frequent oral participation, and higher completion rates of adaptive exercises. Learning analytics indicated progressive gains in accuracy and reductions in time-on-task, while enabling data-informed differentiation and strengthening teachers’ and assistants’ instructional efficacy. These results suggest that carefully localised and pedagogically orchestrated AI-based adaptive systems can support more inclusive and motivating EFL environments for SEN learners in low-resource, non-English-dominant contexts without displacing teacher professional judgement.