Reframing online English-speaking instruction for low-proficiency learners: a skill acquisition theory-informed, AI-mediated instructional model

Online English-speaking instruction for low-proficiency learners is hampered by pedagogical constraints: overreliance on self-regulated preparation, loosely structured practice, product-oriented feedback, and delivery-oriented uses of technology. This Innovative Practice paper proposes an instructional model that systematically combines Skill Acquisition Theory (SAT) with the affordances of Text-to-Speech (TTS) and Automatic Speech Recognition (ASR) technologies to address these challenges. In this model, learner-generated speech drafts refined through individualized teacher scaffolding anchor declarative knowledge construction; proceduralization is enacted through an iterative TTS–ASR practice loop, consolidated by teacher-led intelligibility checks; and audio-based tasks and designated tests are designed to foster intra- and inter-lesson automatization. Practitioner evidence from heterogeneous online learners illustrates observable gains in draft complexity and coherence across learning units, with learners also demonstrating intelligibility in audio-based tasks and adaptability in test-like tasks. The model's use of widely available tools, its modular design, and its language-agnostic nature ensure accessibility and scalability across diverse instructional contexts. By positioning SAT as the organizing framework and AI-mediated tools as structural components of the skill acquisition process – rather than peripheral supplements – this model offers a theory-driven, pedagogically viable approach to online speaking instruction for low-proficiency learners.