Instructor-Designed AI Tutors in University Foreign Language Education: A Mixed-Methods Study of Learner Motivation and Reflective Learning Experience Based on Self-Determination Theory
TL;DR
The findings suggest that the educational effectiveness of generative AI in foreign language education is not determined by frequency of use alone but also by the quality of pedagogical design underlying its deployment.
Abstract
This study investigates the educational potential of instructor-designed Generative Pre-trained Transformers (GPTs) in a university-level Japanese course, drawing on self-determination theory (SDT) and the noticing hypothesis. Using a mixed-methods design, we examined how the continuous use of an instructor-developed Artificial Intelligence tutoring system (basic Japanese GPTs) relates to learners’ psychological needs satisfaction, cognitive noticing, and perceptions of Artificial Intelligence (AI)-assisted learning among 74 undergraduate students at a South Korean university. Quantitative data were analyzed using descriptive statistics and Pearson correlation; qualitative data from open-ended items and reflective writing underwent systematic content analysis. The findings revealed three key patterns. First, learners reported relatively high levels of satisfaction across all three SDT needs—autonomy, competence, and relatedness—particularly in relation to self-directed reviews and affective safety. Second, qualitative analysis identified three distinct noticing experiences: AI-supported clarification of linguistic form, noticing through intentional error generation and AI feedback, and metacognitive regulation of learning strategies. Third, the learners perceived the instructor-designed GPTs not merely as a convenience tool but as a structured learning environment that supported output-oriented, interaction-based practice. These findings suggest that the educational effectiveness of generative AI in foreign language education is not determined by frequency of use alone but also by the quality of pedagogical design underlying its deployment. This study contributes a practice-based model for AI integration in general education language courses while acknowledging limitations related to its single-course scope and reliance on self-reported data.