Speaker
Description
While Generative AI (GenAI) is widely used to improve oral proficiency in English Language Teaching (ELT), its role in shaping the learning process remains under-researched. This multiple-case study explores how structured AI-assisted speaking tasks can foster young learners' agency (behavioral, cognitive, affective) and foundational AI competencies. Six Grade 5 Vietnamese EFL learners from Ly Thai To Primary School participated in a two-week pilot of an AI-assisted Speaking Framework (AISF) across four structural stages: Prepare, Interact, Refine, and Reflect. Data regarding their speaking performance and behaviors were triangulated via AI conversation logs, screen recordings, observation field notes, and immediate L1-medium stimulated recall interviews. The findings reveal that the tasks significantly enhanced behavioral agency through student-led, self-paced dialogues and cognitive agency via active meaning negotiation during communication breakdowns. Additionally, the non-judgmental AI interlocutor reduced foreign language anxiety (affective agency), while nurturing basic skills in human-AI collaboration and responsible AI use. This study provides critical empirical refinements to the preliminary framework. It offers primary school educators a validated, practical pedagogical template that can be adapted for small-group instruction, station-rotation models, and blended learning contexts.
Keywords: learner agency · AI competencies · primary EFL · multiple-case study · AI-assisted speaking
