Experiences of Speaking with Conversational AI in Language Education

About the thesis:

Speaking a second or third language is challenging to both teach and learn. Spoken dialogue systems incorporating embodied conversational agents provide low-anxiety environments for practising speaking. However, there is limited research exploring the application of digital tools to enhance speaking skills. This thesis aims to explore and gain a nuanced and critical understanding of how conversational AI is experienced by lower-secondary students during their institutional language education when practising speaking skills through face-to-face interaction. The analysis focuses on teachers’ views and experiences of teaching speaking and students’ self-reported experiences in two spoken dialogue systems.

The findings reveal that teachers are willing to incorporate conversational AI as a complementary tool when teaching speaking skills. Students reported sustained educational experiences, finding it easy, fun, safe and confidence-building. However, disengagement occurred when scenarios lacked sufficient relevance or challenge. Communication breakdowns resulted in frustration, while the extent to which students related socially with the conversational agent influenced their overall experience. Students’ educational experiences align with principles of effective language learning, providing valuable supplementary opportunities for practising speaking in social interaction. Still, teachers and students need to be aware of the limitations and challenges experienced by individuals, hence pedagogical framing is recommended.

Faculty opponent:

Professor Ylva Lindberg, Department of Language, Aesthetic Learning and Literature, Jönköping University

Grading committee:

• Senior Lecturer Ola Knutsson, Department of Computer and Systems Sciences, Stockholm University
• Senior Associate Professor Mattias Arvola, Department of Computer and Information Science, Linköping University
• Professor Alexandra Weilenmann, Department of Applied Information Technology, University of Gothenburg

Link to full text version of the thesis