Pecha, a language practice peer: Guiding language learning interactions through large language models

Abstract

The interaction hypothesis of second language acquisition (Long, 1981) states that negotiated interaction is necessary for language development. In many language learning contexts, educators and stakeholders seek to provide opportunities for learners to engage in meaningful real-life interactions that help them build linguistic, semantic, and rhetorical competence. However, the opportunities provided for interaction can vary in their degree of effectiveness and may only sometimes lead to increased language ability. If these interactions are scaffolded correctly, they can be tuned to maximize their benefits (Loewen & Sato, 2018). Unfortunately, this is not always practical from a temporal and economic perspective. Artificial intelligence (AI) could be the solution for providing learners with individualized, comprehensive assistance during their learning interactions. Accordingly, the authors developed a bespoke application that employs advanced natural language processing and large language model AI technologies to support learner interactions. The application was developed to support students in two different contexts: Kanda University of International Studies in Japan, where students study English, and Massachusetts Institute of Technology in the United States, where learners study Japanese. The rationale for creating the application and selecting its major features is discussed. This is followed by a discussion of how the application functions and how it will be used. The authors will then discuss their plans for implementation into both informal and formal learning contexts at the two universities. They conclude by discussing potential limitations and plans for improving the application.

Keywords

• Computer-Assisted Language Learning (CALL)
• artificial intelligence
• AI
• interaction
• large language models
• LLMs