DEVELOPING ENGLISH SPEAKING SKILLS THROUGH AI-ASSISTED PRONUNCIATION INSTRUCTION: A PILOT STUDY IN MULTILINGUAL INDIAN SECONDARY CLASSROOMS

Abstract

The pilot study aims to explore the possibility of using AI-supported pedagogy in multilingual secondary classrooms in India to facilitate the development of Englishspeaking competence, such as pronunciation, conversational fluency, and speaking confidence, given large class sizes, the lack of individualized oral feedback, and learners' pronunciation anxiety. A single group, pre-test, post-test quantitative phase (embedded mixed method design) was followed by a qualitative phase of data collection. The intervention involved one teacher-led, human-in-the-loop program deployment for 100 students in Class 9 across four intact classes in three schools, using four AI speaking tools (ELSA Speak, ChatGPT Voice Mode, Speechace, and Microsoft Speaker Coach). Classroom observations, semi-structured teacher interviews, student focus groups, reflective journals, and pre- and post-speaking assessments were used to gather data, which were analyzed using Braun and Clarke's (2006) thematic analysis and descriptive and paired-sample comparisons. This study provides pilot evidence of intelligibility in the context of classroom-based, multilingual Indian secondary schools through a multi-tool AI speaking module, a principle that differs from native-speaker imitation, in contrast to previous studies that focused on higher education or single-tool laboratory settings.