Bridging the Gap in Oral Proficiency: Pedagogical Integration of Conversational Agents

Developing speaking skills in foreign languages remains a challenge for both students and teachers. Many students experience anxiety when speaking due to limited opportunities for real-life practice and fear of making mistakes, while the transient nature of spoken language makes it difficult for teachers to provide individualised feedback. AI-driven conversational agents are emerging as a promising tool to enhance oral proficiency and interaction, yet research on their pedagogical integration in real classroom settings remains limited. Building on previous findings, including Ericsson’s (2023) doctoral thesis, this study extends the inquiry by examining a large-scale implementation in secondary education and addressing the role of teachers in AI-supported language learning.

Using a mixed-methods approach, this study investigates how 96 Swedish secondary school students practiced speaking French, German, or Spanish with conversational agents embedded in immersive 360-degree video simulations. The pedagogical design aligns with communicative and action-oriented language teaching, enabling dialogues in authentic everyday-life scenarios, such as ordering at a café or bakery. Findings confirm that conversational AI increases speaking time, reduces anxiety, and fosters engagement, particularly for students reluctant to participate in traditional classroom discussions. However, challenges such as technical limitations, a lack of spontaneous speech capabilities, and the need for pedagogical integration highlight the necessity of a teacher-facilitated approach rather than full automation.

This study contributes to the field by demonstrating how AI can enhance accessibility and engagement in oral language practice, offering insights into scalable, pedagogically embedded conversational AI. It also raises critical questions about future trends, particularly regarding AI-mediated feedback, adaptive learning, and the evolving role of educators in AI-assisted education. These findings emphasise the need for a balanced, human-AI collaboration to overcome traditional barriers in language education and foster confident language learners.