Navigating Interpretive Uncertainty in AI-Assisted Self-Access Japanese Writing

Anh Nguyen Thi Lan, Hanoi University, Vietnam. https://orcid.org/0000-0002-3700-0368

Cuc Pham Thi Thu, Hanoi University, Vietnam. https://orcid.org/0009-0004-6813-5064

Lien Nguyen Thi Phuong, Hanoi University, Vietnam. https://orcid.org/0009-0003-7148-6685

Anh, N. T. L., Cuc, P. T. T., & Lien, N. T. P. (2026). Navigating interpretive uncertainty in AI-assisted self-access Japanese writing. Studies in Self-Access Learning Journal, 17(2), 203–217. https://doi.org/10.37237/170104

Abstract

This exploratory qualitative study examines how students and teachers perceive AI-assisted self-access writing in a Vietnamese university context. Here, self-access writing refers to writing carried out independently beyond classroom feedback cycles. Data were collected through open-ended and Likert-scale surveys from 49 Japanese language students and 11 writing instructors. Across both student and teacher responses, similar concerns appeared repeatedly. AI had become part of routine revision for many learners, yet this convenience did not remove uncertainty. Participants described difficulty deciding whether AI-generated expressions actually matched the intended situation, and also raised concern about becoming too reliant on AI during writing. Taken together, these responses suggest that AI is changing more than the speed of revision; it is reshaping how learners make sense of feedback and decide what to trust.

Keywords: generative AI; self-access learning; feedback literacy; learner regulation; AI-mediated revision

Across many university Japanese language programs, writing support no longer remains confined to classrooms, teacher commentary, or scheduled feedback cycles. Recent studies have shown that digital writing support, including machine translation and AI-assisted feedback, has become increasingly integrated into second language writing practices both inside and outside formal instructional settings (Godwin-Jones, 2022; Kasneci et al., 2023). Students increasingly revise drafts through continuous interaction with online resources, peer communication platforms, translation systems, and generative AI tools that can rewrite, explain, and evaluate texts in real time. What was once a relatively bounded pedagogical sequence of assignment, correction, and revision has gradually become dispersed across overlapping digital environments extending beyond formal instructional settings.

This transformation is particularly visible in foreign-language university contexts such as Vietnam, where opportunities to use Japanese outside institutional settings are often limited. For many learners, writing development depends heavily on regular feedback because writing in Japanese requires not only grammatical accuracy but also sensitivity to register, interpersonal positioning, discourse organization, and contextually appropriate expression. At the same time, sustained individualized feedback is increasingly difficult to provide through classroom instruction alone. As classes face tighter curricular pressures, revision support is often reduced to selective correction, brief comments, or final evaluation, leaving much of the revision process unfold elsewhere. Students, therefore, continue writing not after instruction ends, but in the spaces around it: through online chat groups, translation platforms, model texts, and increasingly through generative AI systems that can respond immediately to unfinished drafts.

Under such conditions, self-access learning cannot be understood simply as an independent study separate from external support. Recent work has increasingly emphasized that self-access learning is often shaped through networks of social, institutional, and digital mediation rather than through isolated learner activity (Murray et al., 2014; Mynard, 2019). In this study, self-access writing refers specifically to writing activities carried out beyond formal classroom feedback cycles, where learners regulate revision through available social and digital resources. Rather than writing alone, learners often move across overlapping forms of support—checking expressions in online communities, comparing examples found on the internet, asking friends for confirmation, revising through translation tools, or interacting with AI systems that respond instantly to incomplete texts.  From this perspective, self-access learning involves navigating overlapping forms of support distributed across different learning spaces.

The emergence of generative AI intensifies these developments in important ways. Unlike earlier digital writing tools such as dictionaries, machine translation, or grammar checkers, generative AI allows learners to interact continuously with feedback throughout the writing process itself. Learners can request explanations, compare alternative formulations, revise repeatedly, and explore stylistic differences through dialogue-like interaction with AI systems. As a result, AI increasingly functions not simply as a supplementary writing aid, but as part of an ongoing feedback environment embedded within self-access writing activity.

This shift is especially significant in foreign-language writing because feedback is not merely informational. It shapes how learners judge what sounds appropriate, what feels persuasive, and how they position themselves in relation to readers and communicative situations (Hyland & Hyland, 2006). Recent work on feedback literacy has similarly emphasized that learners must not only receive feedback, but also interpret, question, and use it effectively (Carless & Boud, 2018; Winstone & Carless, 2020). Under AI-supported conditions, however, learners increasingly encounter multiple forms of feedback before teachers read their drafts. Revision therefore emerges through interaction among AI systems, peer support, online examples, and learners’ own self-monitoring practices.

While expanded access to feedback may encourage experimentation and sustained revision, it also introduces new forms of uncertainty. In Japanese writing, grammatical accuracy alone rarely guarantees communicative appropriateness. A sentence may sound polished while still shifting interpersonal distance, unintentionally softening or strengthening tone, or introducing expressions that feel pragmatically unstable in context. Learners often recognize that AI-generated revisions “sound better,” yet remain uncertain about why one formulation feels more appropriate than another (Maynard, 1997; Minegishi Cook, 2008). The difficulty, therefore, is no longer simply correcting grammar, but deciding whether fluent revision can be trusted.

Although much of the current discussion on AI-assisted writing has focused on visible concerns such as plagiarism, writing quality, or learner motivation (Godwin-Jones, 2022; Kasneci et al., 2023; Yan, 2023), less attention has been given to how learners actually work through AI-generated feedback during revision itself, especially in writing that continues beyond classroom supervision. This gap becomes equally relevant for teachers, whose role as evaluative authorities may also be shifting as students increasingly revise before submission through AI-mediated interaction.

Against this background, the present study explores learners’ and teachers’ perspectives on AI-assisted Japanese writing in a Vietnamese higher education context. Drawing on open-ended survey responses, the paper focuses on how learners experience AI-mediated feedback during self-access writing and how teachers perceive the changing structure of revision and feedback practices. Specifically, the study addresses three questions: (1) How do learners use generative AI during self-access Japanese writing? (2) What kinds of uncertainty emerge when learners interpret AI-generated revisions? (3) How do teachers understand the changing role of feedback under AI-assisted writing conditions? In doing so, the study contributes to ongoing discussions of learner autonomy, feedback literacy, and the changing structure of self-access language learning in digitally mediated environments.

Self-Access Learning, Feedback, and Generative AI

Self-access language learning has long been associated with spaces intentionally set apart from the classroom: self-access centers filled with graded materials, audio resources, worksheets, and other tools designed to support learners in continuing their study independently (Gardner & Miller, 1999). Behind these early models was the broader idea that learners should gradually take greater responsibility for decisions about what to study, how to study, and how to evaluate their own progress. This understanding of autonomy has deep roots in language education research, particularly in the work of Henri Holec (1981), David Little (1991), and Phil Benson (2011), who emphasized that autonomy does not mean learning without support, but learning through greater control over one’s own learning process.

At the same time, the conditions surrounding self-access learning have changed considerably. Learners today rarely move from classroom instruction into a clearly separate space of “independent study.” Writing, in particular, often continues through digital platforms, online communities, peer exchanges, and algorithmically mediated tools that remain closely connected to formal learning. Recent work has therefore begun to reconsider self-access less as a physical location and more as a shifting network of resources, relationships, and forms of mediation that learners move through while studying (Murray et al., 2014; Mynard et al., 2022).

This shift becomes especially visible in writing. Unlike some forms of language practice that depend on face-to-face interaction, writing has always involved some degree of mediation: dictionaries, model texts, teacher comments, and reference materials have long shaped the writing process. What has changed is the immediacy and responsiveness of that mediation. A learner writing today may move quickly between their own draft, online examples, peer comments, translation tools, and now generative AI, often within the same writing session. The boundary between drafting and revising has therefore become far less stable.

This changing structure also alters how feedback is experienced. In more conventional writing classrooms, feedback often followed submission. A text was completed first, then corrected later. That sequence gave learners time to produce language before confronting external evaluation. More recent work on feedback literacy has questioned the assumption that feedback simply moves in one direction—from teacher to student. Even when feedback is available, learners still have to decide what to do with it, which comments to accept, and how those comments fit their own intentions as writers (Carless & Boud, 2018).

This becomes more complicated once several forms of feedback appear at the same time. A learner may compare teacher comments with online examples, ask peers for confirmation, use machine translation, and consult AI-generated reformulations, all while revising the same paragraph. The difficulty here is not a lack of support, but the opposite: too many possible directions. Revision often becomes a process of moving back and forth among alternatives, sometimes with no clear authority to settle which choice is best.

Generative AI intensifies this condition in a way earlier digital tools did not. Dictionaries and grammar checkers usually responded to specific linguistic problems. Generative AI responds to unfinished language itself. A sentence can be rewritten, softened, expanded, shortened, or reorganized almost immediately. Learners may ask for several alternatives in succession, not because the original sentence is wrong, but because they are still searching for one that feels right.

This matters particularly in Japanese writing, where appropriateness often depends on subtle judgments of interpersonal distance, politeness, and rhetorical stance (Maynard, 1997; Minegishi Cook, 2008). A grammatically correct sentence may still feel awkward depending on who is speaking, to whom, and in what context. In this sense, AI-assisted revision introduces a different kind of problem. The issue is not simply whether the sentence is correct, but whether the learner can recognize what has changed when the AI makes it “better.”

Seen from this perspective, the central question is no longer whether AI improves writing in a measurable sense. A more immediate question is how learners work with AI-generated feedback while revising, and how teachers understand these changing revision practices as they increasingly take place before any classroom feedback occurs. It is this shift in the structure of revision itself that the present study takes as its starting point.

Methodology

Research Design

This study adopted an exploratory qualitative design to examine how learners and teachers perceive the use of generative AI in self-access Japanese writing activities. The study did not aim to measure writing improvement or compare performance outcomes. Instead, it focused on how participants described their experiences of revision, feedback, and decision-making in writing practices that increasingly take place outside formal classroom settings. A qualitative approach was considered appropriate because the study sought to capture emerging patterns of uncertainty, dependence, and feedback interpretation that may not be easily observable through quantitative measures alone.

Participants

The study was conducted within a Japanese language program at a Vietnamese university. A total of 49 undergraduate students and 11 instructors participated in the survey. The student participants were enrolled in intermediate and upper-level Japanese courses and had prior experience with academic writing assignments in Japanese. Most of them were at approximately JLPT N3 to N2 level, where writing tasks required not only grammatical control but also awareness of register and contextual appropriateness. The instructor participants were responsible for writing-related courses across different levels of the program and had experience providing written feedback and evaluating student compositions.

Participants were recruited through departmental distribution during the regular academic semester. Participation was voluntary, and responses were collected anonymously. All participants were informed of the purpose of the study before completing the survey.

Instruments

Separate questionnaires were designed for students and instructors. Both questionnaires combined Likert-scale items with open-ended questions in order to capture both general tendencies and individual reflections.

The student questionnaire focused on writing habits, experiences with AI-assisted revision, perceived advantages and difficulties, and situations in which students felt uncertain while revising Japanese texts. Particular attention was given to how students used generative AI during writing, what kinds of feedback they trusted, and how they evaluated alternative expressions.

The instructor questionnaire focused on current writing instruction, feedback practices, perceptions of students’ writing difficulties, and views on the possible influence of generative AI on revision behavior and writing assessment. The inclusion of open-ended items allowed instructors to describe concerns that might not emerge through fixed-response questions alone.

Data Collection

Data were collected between March and April 2026 through online survey distribution. The surveys were administered using digital forms and completed outside class time. This format was chosen because it allowed participants to respond in their own time and describe their writing practices more freely, particularly since many of these practices occurred outside formal classroom settings.

The student and teacher surveys were distributed separately. In total, 49 valid student responses and 11 valid teacher responses were obtained. The responses included both scaled answers and written comments, which provided the main qualitative data for analysis. To preserve anonymity, student responses are identified in the findings as S1–S49 and teacher responses as T1–T11.

Data Analysis

The qualitative responses were analyzed using thematic analysis following the framework proposed by Virginia Braun and Victoria Clarke (2008). The coding process was primarily inductive, allowing themes to emerge from the participants’ own descriptions rather than being imposed in advance.

The analysis followed several stages. First, all responses were read repeatedly to identify recurring patterns and notable expressions. Second, initial codes were assigned to segments related to revision behavior, feedback interpretation, uncertainty, and perceptions of AI-assisted writing. Third, related codes were grouped into broader thematic categories. These themes were then reviewed and refined in relation to the research questions.

To strengthen interpretive consistency, coding decisions were reviewed multiple times during the analytic process, and contradictory or less typical responses were also considered in order to avoid overgeneralization. The analysis did not aim at statistical generalization, but at identifying meaningful patterns in how participants described the changing experience of writing and revision under AI-assisted conditions.

Findings and Discussion 

Theme 1: AI as Part of Everyday Revision

One of the clearest patterns in the student responses was how regularly generative AI had entered their writing routines. Out of the 49 students, 37 reported using AI during writing at least occasionally, while 21 said they used it frequently when revising sentences or checking expressions. For many of them, AI was no longer treated as an occasional support tool but had become part of the writing process itself.

The most common uses were practical. Students described using AI to check grammar, search for kanji, ask for more natural phrasing, or compare different ways of expressing the same idea. As one student explained: “When I write and feel that a sentence sounds strange, I usually ask ChatGPT to rewrite it. Sometimes I ask several times until it sounds better.” (S12). Another explained: “I often use AI when I know what I want to say but cannot find the right Japanese expression.” (S18).

These responses suggest that AI was often used at moments of hesitation rather than after a draft had been completed. Instead of leaving uncertain parts unresolved, students tended to stop, consult AI, and continue from there. In this sense, revision was often taking place during writing rather than after it.

Teacher responses showed a similar perception. Eight of the eleven instructors said they had noticed changes in how students revised before submission. Several commented that student drafts now appeared more polished at the sentence level than in previous years. One instructor noted: “Compared to before, students seem to revise more before handing in their writing. The sentences are often cleaner, but sometimes it is difficult to know how much of that is their own revision.” (T4).

At the same time, the responses also suggest that most AI-assisted revision remained highly local. Students spoke mostly about fixing wording, improving grammar, or making expressions sound more natural. Much less attention was given to larger concerns such as audience awareness, organization, or rhetorical intention. This unevenness appeared repeatedly in both student and teacher accounts.

Theme 2: Uncertainty and the Problem of Feedback Authority

A second pattern appeared in how students talked about trust. Although many described AI-generated revisions as helpful, a considerable number also admitted that they were not always sure why the revised sentence sounded better. Out of the 49 student responses, 28 mentioned some form of uncertainty when judging AI-generated wording, especially in cases involving politeness, tone, or expressions they had not used before.

One student wrote: “Sometimes the AI version sounds more natural than mine, but I cannot explain what changed. I just feel it sounds better.” (S30). Another commented: “There are times when ChatGPT gives a very polite sentence, but I do not know if it fits the situation I am writing about.” (S20).

These comments appeared most often when students described writing tasks such as emails, reflections, or formal compositions, where small shifts in expression could affect how the text might be read. Several students said they accepted AI-generated revisions because the sentences sounded smoother or more advanced than what they could produce on their own, even when they were not fully confident about the meaning.

Teachers noticed similar patterns. Seven of the eleven instructors raised concern that some student writing had begun to show unevenness after AI-assisted revision. In several cases, certain sentences appeared unusually sophisticated, while surrounding sentences remained much simpler or showed abrupt changes in politeness level. One instructor wrote: “Sometimes one sentence looks almost native-like, but the next sentence returns to the learner’s usual level. The balance feels strange.” (S45).

Another instructor observed: “Students may choose AI-generated expressions because they look correct, but they do not always understand the difference in nuance.” (T9).

What becomes visible in these responses is not simply error, but hesitation. Students were often able to sense that something had changed in the revised sentence, yet they could not always explain whether that change improved the text or shifted it in an unintended direction. This uncertainty appeared repeatedly in both student and teacher responses, particularly in relation to register, politeness, and contextual fit.

Theme 3: Dependence and Reflective Regulation

A third theme concerned the growing dependence on AI during revision and the different ways students positioned themselves in relation to that support. Among the 49 student responses, 19 explicitly mentioned that AI had become something they relied on regularly during writing, while 14 expressed concerns that frequent use might reduce their ability to revise independently over time.

One student wrote: “Before, I tried to fix my sentences by myself first. Now I often ask AI immediately because it is faster.” (S46).

Another reflected more cautiously: “Sometimes I worry that I am improving the text, but not improving my own writing.” (S35)

These comments suggest that for some students, AI had gradually become part of their default writing behavior. The issue was not simply convenience. Several students described a growing difficulty in returning to earlier writing habits once AI-assisted reformulation became easily available, particularly when they lacked confidence in their own Japanese.

At the same time, not all students used AI in the same way. Twelve students described more deliberate or selective use, such as comparing several AI-generated alternatives, modifying suggested sentences, or combining AI-generated wording with their own expressions. One student explained: “I usually compare the AI version with mine. Sometimes I keep only one phrase and change the rest.” (S23). Another wrote: “I ask for different versions and choose the one closest to what I want to say.” (S26).

Teacher responses reflected this difference as well. Six instructors commented that stronger students seemed more likely to question AI-generated suggestions, while weaker students often accepted them quickly. One teacher noted: “Students with stronger language ability still make choices. Weaker students often copy the revision without asking why.” (T5).

Across these responses, the contrast was less about whether students used AI and more about how they worked with it. Some remained actively involved in shaping the final text, while others appeared to hand over more of the revision process to the system itself. This difference appeared repeatedly in both student and teacher accounts and became one of the clearest points of tension in the dataset.

Discussion

The findings show that generative AI is not simply adding another tool to self-access writing. What seems to be changing more fundamentally is the timing and structure of revision itself. In many of the student accounts, revision no longer appeared as a separate stage following a completed draft. Instead, writing and revision had become closely intertwined, often interrupted by repeated consultation with AI-generated suggestions. This changes the rhythm of writing. Learners do not necessarily finish a thought and revise later; they revise while the thought is still forming

This shift matters because it complicates how self-access learning has often been understood. Earlier discussions of self-access emphasized learner control, resource selection, and independent management of learning (Benson, 2011; Gardner & Miller, 1999). The present findings do not contradict that view, but they suggest that learner control now operates under very different conditions. The issue is no longer whether support is available. Support is almost always available. What becomes difficult is deciding which support to trust, when to use it, and how far to follow it.

Seen in this way, generative AI intensifies a problem already present in feedback literacy. As Carless and Boud (2018) argue, receiving feedback has never guaranteed understanding. Learners must still interpret what feedback means and decide how to act on it. What changes under AI-assisted conditions is the speed and frequency of those decisions. Feedback no longer arrives after submission in limited amounts. It appears continuously, often in multiple alternative forms, and demands immediate judgment.

The student responses in this study suggest that this interpretive burden becomes especially visible in Japanese writing. Unlike surface-level grammar correction, decisions about politeness, tone, or interpersonal distance are rarely straightforward. A sentence may be linguistically correct and still feel wrong in context. This helps explain why many students described AI-generated sentences as sounding “better” while remaining unsure whether they were actually appropriate. The uncertainty here is not technical. It is relational. It concerns how language positions the writer toward the reader.

Teacher responses make this tension even clearer. Several instructors pointed to unevenness in student writing, where AI-assisted revision produced highly polished sentences beside much simpler ones. This unevenness suggests that AI may improve local phrasing without necessarily supporting coherence or rhetorical consistency across the text. In that sense, fluency can become misleading. A polished sentence may give the impression of strong control while masking uncertainty underneath.

The findings also complicate common discussions about learner dependence. Dependence here does not appear as simple passivity. Many students remained active during revision, asking for alternatives, comparing versions, and making adjustments. Yet activity alone does not always mean reflection. Some learners appeared to move through revision procedurally, accepting AI-generated reformulation because it reduced uncertainty quickly. Others used AI more selectively and maintained stronger control over final choices. This difference may be one of the most important implications of the study.

Taken together, these patterns suggest that the central challenge of AI-assisted self-access writing is not access to feedback, but the ability to remain critically engaged while feedback becomes faster, smoother, and more persuasive. In this sense, generative AI does not reduce the importance of learner autonomy. It changes what autonomy now requires.

Implications for Self-Access Learning

The findings of this study point to several practical implications for self-access writing support in Japanese language education. The first concerns how learners are prepared to work with feedback. If AI-generated feedback is now part of everyday writing practice, then the question is no longer simply whether students have access to support, but whether they know how to use that support critically.

Several student responses showed that uncertainty often remained even after AI-assisted revision. In many cases, students accepted rewritten sentences because they sounded smoother or more natural, but they were not always confident about why those changes worked. This suggests that self-access support may need to place greater emphasis on helping learners slow down during revision and examine what exactly has changed in an AI-generated reformulation.

One possible response is to incorporate reflective comparison into writing support. Instead of asking students only to correct mistakes, teachers or self-access advisors might encourage them to compare their original sentence with an AI-generated alternative and explain why they chose one over the other. The purpose here is not to reject AI assistance, but to make revision more visible as a decision-making process.

A second implication concerns the role of feedback literacy. The findings suggest that access to immediate revision support can easily create the impression that writing problems have been solved, even when the learner remains unsure about tone, nuance, or contextual appropriateness. This may be particularly important in Japanese writing, where small changes in expression can alter politeness or interpersonal distance in ways that are not always obvious. Helping learners recognize these differences may become an important part of writing instruction.

The teacher responses also suggest a possible shift in pedagogical focus. Several instructors noted that students now often arrive with drafts that have already been revised through AI before submission. This does not necessarily reduce the teacher’s role, but it may change where that role becomes most important. Rather than concentrating only on sentence-level correction, teachers may increasingly need to focus on broader issues such as coherence, rhetorical consistency, and contextual fit—areas that AI-generated revision does not always handle reliably.

For self-access learning more broadly, the findings suggest that support structures may need to expand beyond access to resources alone. Learners may also need spaces—whether in advising sessions, peer discussion, or classroom reflection—where they can talk through uncertainty and examine how they make revision decisions. This may be one practical way to support autonomy under conditions where feedback is always available, but not always easy to interpret.

Conclusion

This study examined how learners and teachers in a Vietnamese university context understand the growing role of generative AI in self-access Japanese writing. The findings show that AI has already become part of ordinary revision practices for many learners, not simply as a tool for correcting grammar, but as a constant source of alternative wording, reformulation, and immediate feedback during writing.

At the same time, the study also shows that easier access to revision support does not necessarily reduce uncertainty. For many students, the difficulty shifted from producing sentences to judging whether AI-generated revisions actually fit the communicative situation they intended. This was especially visible in areas such as politeness, tone, and interpersonal distance, where appropriateness could not always be judged through fluency alone.

Teacher perspectives reinforce this point. While instructors recognized that AI-assisted revision could encourage more frequent rewriting and cleaner sentence-level output, they also pointed to new difficulties in evaluating student writing, particularly when polished AI-generated expressions appeared alongside uneven or inconsistent language. This suggests that AI may be changing not only how students revise, but also how teachers interpret what student writing represents.

Taken together, these findings suggest that the central issue in AI-assisted self-access writing is no longer access to feedback itself, but how learners work with feedback that is immediate, persuasive, and sometimes difficult to evaluate critically. In this sense, the challenge of writing may increasingly lie in interpretation rather than correction.

This study is limited by its small-scale design and its focus on one institutional context. The data also rely on participant accounts rather than direct observation of revision practices as they happened. Future research may therefore benefit from following actual writing processes over time, examining how learners interact with AI during revision, and comparing these patterns across different proficiency levels and learning environments.

Notes on the Contributors

Dr. Nguyen Thi Lan Anh is a lecturer at Hanoi University, Vietnam, where she serves as Head of the Division of Japanese Language and Culture. Her research interests include Japanese Studies, Japanese history and culture, language education, linguistics, and intercultural studies. She has extensive experience in teaching and researching Japanese language, culture, and Vietnam–Japan intercultural exchange. Her recent work focuses on Japanese language education, learner autonomy, intercultural communication, and the pedagogical implications of generative AI in foreign language teaching and learning.

Pham Thi Thu Cuc is a lecturer at Hanoi University, Vietnam. She has more than 20 years of teaching and research experience in the fields of Japanese language, Japanese culture, and tourism. Her research interests include Japanese Studies, Japanese language education, tourism studies, intercultural communication, and Japanese culture. She has extensive experience in curriculum development, textbook writing, and supervising student research. Her recent research focuses on Japanese language education, intercultural communication, tourism studies, and the application of generative AI in foreign language teaching and learning.

Nguyen Thi Phuong Lien, MA, is a lecturer at Hanoi University, Vietnam. Her teaching areas include Japanese language, Japanese writing, Japanese grammar, and Japanese language teaching methodology. Her research interests include applied linguistics, Japanese language education, learner corpus research, grammatical error analysis, second language acquisition, Vietnamese learners of Japanese, and the application of artificial intelligence in teaching and learning Japanese writing. Her current work focuses on grammatical errors in L2 Japanese writing, learner corpus data, transparent error annotation procedures, and the pedagogical use of AI to support the development of Japanese writing skills among Vietnamese university students.

References

Braun, V., & Clarke, V. (2008). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa

Carless, D., & Boud, D. (2018). The development of student feedback literacy: Enabling uptake of feedback. Assessment & Evaluation in Higher Education, 43(8), 1315–1325. https://doi.org/10.1080/02602938.2018.1463354

Gardner, D., & Miller, L. (1999). Establishing self-access: From theory to practice. Cambridge University Press.

Godwin-Jones, R. (2022). Partnering with AI: Intelligent writing assistance and instructed language learning. Language Learning & Technology, 26(2), 5–24. https://doi.org/10.64152/10125/73474 

Holec, H. (1981). Autonomy and foreign language learning. Pergamon.

Hyland, K., & Hyland, F. (2006). Feedback on second language students’ writing. Language Teaching, Vol 39 (2). 83 – 101. https://doi.org/10.1017/S0261444806003399 

Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeiffer, F., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., …Kasneci, 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

Little, D. (1991). Learner autonomy 1: Definitions, issues and problems. Authentik.

Maynard, S. K. (1997). Japanese communication: Language and thought in context. University of Hawai‘i Press.

Minegishi Cook, H. (2008). Socializing identities through speech style: Learners of Japanese as a foreign language. Multilingual Matters

Murray, G., Fujishima, N., & Uzuka, M. (2014). Semiotics of place: Autonomy and space. In G. Murray (Ed.), Social dimensions of autonomy in language learning (pp. 81–99). Palgrave Macmillan.

Mynard, J., Ambinintsoa, D. V., Bennett, P. A., Castro, E., Curry, N., Davies, H., Imamura, Y., Kato, S., Shelton-Strong, S. J., Stevenson, R., Ubukata, H., Watkins, S., Wongsarnpigoon, I., & Yarwood, A. (2022). Reframing self-access: Reviewing the literature and updating a mission statement for a new era. Studies in Self-Access Learning Journal, 13(1), 31–59. https://doi.org/10.37237/130103 

Benson, P. (2011). Teaching and researching autonomy in language learning (2nd ed.). Routledge.

Winstone, N., & Carless, D. (2020). Designing effective feedback processes in higher education: A learning-focused approach. Routledge.

Yan, D. (2023). Impact of ChatGPT on learners in an L2 writing practicum: An exploratory investigation. Education and Information Technologies, 28, 13943–13967. https://doi.org/10.1007/s10639-023-11742-4