Human-AI Interaction in Language Acquisition: Evaluating LLM as a Language Partner

Abstract

This single-case study investigates the potential of ChatGPT’s advanced voice interface as a self-guided language partner for a B2.1-level German learner. Given the growing interest in AI’s role in education and the need for flexible, accessible language practice, this research explores a hybrid learning model that combines AI-mediated conversation with conventional textbook study [1], [2]. The participant in this study was also the author, providing a unique insider perspective into the self-directed learning process with AI.

Over 30 structured sessions, the learner engaged in spoken interactions with ChatGPT using its voice function, receiving real-time audio responses while simultaneously working through the Sicher! B2.1 textbook. This mixed-methods study assessed learning outcomes using standardized tests, vocabulary tracking, PANAS (Positive and Negative Affect Schedule) surveys, reflective journals, and self-interviews. The contribution of each learning method was qualitatively assessed through detailed self-reflection and cross-referenced with quantitative data, allowing for an understanding of how each component influenced specific aspects of language acquisition. The qualitative data from journals and interviews were analyzed using thematic analysis to identify recurring patterns and insights.

Results showed a significant increase in vocabulary, particularly in noun acquisition, with 775 words encountered via ChatGPT compared to 319 through the textbook. Standardized test scores improved from 56% to 92%, and workbook assessments remained consistently high. Word cloud analyses across three intervals demonstrated a progression from foundational topics (e.g., relationships) to more complex domains like workplace discourse and media critique. PANAS scores revealed substantial increases in positive affect post-session, while negative affect decreased modestly, indicating reduced anxiety and enhanced motivation [1], [2].

Qualitative data from journals and interviews further supported these trends, highlighting growing confidence, appreciation for the AI’s flexibility, and reduced stress due to the non-judgmental practice environment. However, critical limitations were observed: the AI’s feedback was often overly optimistic, lacked deeper pedagogical scaffolding, and its “pedagogical authority” was constrained by its design as a general-purpose conversational model rather than a specialized language tutor [1]. Technical interruptions and occasional misinterpretations also disrupted the learning flow, pointing to the epistemological constraints of AI in certain learning contexts.

This study concludes that while ChatGPT’s voice interface is an effective tool for promoting speaking fluency and engagement—particularly by providing a low-anxiety practice environment—it functions best when integrated with structured materials and human feedback for comprehensive language development. Hybrid learning models, blending conversational AI with traditional curricula, offer a promising approach for self-guided language learners. This research aligns with recent discussions on the utility of AI conversational partners in language learning [1]–[3].

References

[1] H. Ji, I. Han, and Y. Ko, “A Systematic Review of Conversational AI in Language Education,” Journal of Research on Technology in Education, vol. 55, no. 1, pp. 48–63, 2023.

[2] H. Qiao and A. Zhao, “Artificial Intelligence-Based Language Learning,” Frontiers in Psychology, vol. 14, p. 1255594, 2023.

[3] J. Belda-Medina and J. R. Calvo-Ferrer, “Using Chatbots as AI Conversational Partners in Language Learning,” Applied Sciences, vol. 12, no. 17, p. 8427, 2022.