The Effectiveness of the Domestic AI Model “Doubao" in Correcting Japanese Business Email Errors: An Empirical Study at Shanghai Polytechnic University
DOI:
https://doi.org/10.54691/9v09fp97Keywords:
AI-assisted learning, Japanese business correspondence, writing instruction, Keigo, large language models.Abstract
This study looks at some persistent headaches in teaching Japanese business email writing—specifically, why students keep struggling with Keigo (honorific language) and the unspoken rules of business correspondence. Lately, AI tools like Doubao have started showing up in language classrooms, and they're surprisingly good at giving students fast, tailored feedback. So I wanted to see: can AI actually make business Japanese writing instruction work better? I ran a quasi-experiment at Shanghai Polytechnic University. Two groups of undergrads, same course, same material. One group got to use AI to revise their email drafts; the other stuck with traditional teacher feedback only. I collected writing samples from both groups and compared error rates on keigo and sentence structure. The results? The AI-assisted group made noticeably fewer mistakes. But here's the interesting part: it wasn't the AI doing all the work. Students still needed their teacher's input—the sweet spot seemed to be AI suggestions plus human guidance, layered together. They picked up on standard business conventions faster, wrote more confidently, and stopped making those awkward pragmatic errors that usually take months to unlearn.
Downloads
References
[1] Chen, Z. Y. (2020). Japanese honorifics and feminine language in business activities. Journal of Japanese Language Study and Research, 2, 9–16.
[2] Qu, S. (2017). Use of Japanese honorifics in business correspondence and pedagogical reflections. Industrial & Science Tribune, 16, 181–182.
[3] Wu, Y. X. (2015). Theoretical basis for problems in Japanese honorific acquisition among Chinese students. Foreign Language Education, 36, 57–60.
[4] Jian, Z. Z. (2015). Key points and techniques in writing Japanese business documents. Xueyuan, 34, 74–75.
[5] Kasneci, E., Seßler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gül, C., Günther, U., Hüllermeier, E., Krusche, S., Kutrubes, M., Lüders, H., & Reichert, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274. https://doi.org/10.1016/j.lindif.2023.102274
[6] Zhou, X. P. (2018). Features of honorific usage in business Japanese correspondence. English Square, 3, 67–69.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Scientific Journal Of Humanities and Social Sciences

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.





