New Perspectives in Science Education

This study explores how a generative AI-based self-practice system can enhance English for Specific Purposes learning within engineering education[1], particularly in fostering learner autonomy, self-regulated learning, and both academic and professional communication in engineering contexts. The participants were Japanese university students enrolled in an Engineering English course and are expected to develop not only academic vocabulary and discourse competence but also professional communication skills necessary for interaction in multicultural workplaces. They used a generative AI-driven platform integrating automatic speech recognition and automated feedback to train their English speaking skills. The system’s evaluation criteria were aligned with the TOEFL speaking rubric and further mapped to the Common European Framework of Reference for Languages (CEFR) to ensure international consistency and transparency in assessment. Learners completed speaking tasks derived from textbooks, handbooks, manuals, and visual materials such as diagrams and charts, representing authentic CLIL-type input for professional communication[2]. After each practice attempt, they reflected on what they noticed about the system’s feedback and what strategies they used to improve their performance. The log data, which included automated evaluations, estimated TOEFL scores and CEFR levels, number of trials, total word counts, and reflection entries, were analyzed to track performance development over time. The study provides empirical evidence that generative AI-assisted systems can foster engineering English learning and support both academic and workplace-oriented communication in higher education.