Japanese University EFL Student Insights on the Emerging ChatGPT Phenomenon

,


INTRODUCTION
The focus of this study is to give an emic perspective from Japanese university students studying English as a foreign language (EFL) concerning the use of generative artificial intelligence (GenAI) for academic assignments.This study provides a qualitative analysis of information gathered from three participants who used ChatGPT for writing and discussion assignments in a communication course of a liberal arts department at a university in Tokyo.In this paper, we explore the use of generative artificial intelligence (GenAI) in academic settings, focusing specifically on ChatGPT, a prominent model in the realm of GenAI.The research question explored in this study relates to various aspects connected to the application of GenAI in an academic setting during the first year of mainstream use.What are some early perceptions of Japanese university EFL students regarding the use of AI programs, including ChatGPT?What implications can educators derive from these insights to help inform future planning and pedagogy considering the role of this emerging technology in education?
The classroom-based research integrated the students' real-world interest in the emerging technology of GenAI.Themes, derived from coding analysis of a written survey (see Appendix A), semi-structured interviews, class discussions, and follow-up correspondence exploring the students' experience using ChatGPT for academic assignments, include efficiency, reliability, ethics, EFL utilization, and unique insights.

LITERATURE
The Unified Theory of Acceptance and Use of Technology (UTAUT) developed by Venkatesh, Morris, Davis, and Davis (2003) refers to user acceptance of technology based on four elements: performance expectancy (PE), effort expectancy (EE), social influence (SI), and facilitating conditions (FC).An example of PE would be a consumer buying the latest model of iPhone with expectations of better functionality than the previous versions.EE explains how the ease of use for a particular technology influences usability and therefore its popularity.SI can be illustrated by how a friend's positive opinion of a video game console could influence you to buy one.Ideal FC are determined by how the infrastructure or network conditions, such as reliable internet or processing speed, can facilitate the adoption of specific software in a particular environment (Vankatesh et al., 2003).The use of ChatGPT in a university setting meets these standards.A study of over 2240 university students by Abdaljaleel et al. (2024) supports that the perception and popularity of GenAI tools are heavily influenced by those factors, which form the basis of UTAUT.This indicates that the use of such tools in the academic environment is worthy of consideration in terms of impact and application.
In November of 2022, OpenAI released ChatGPT.It grew quickly in popularity with over 100 million active users after just two months of the release.This broke the record for the most rapid expansion of any application in history (Buriak et al., 2023;Wu et al., 2023).In March 2023, the further improved ChatGPT-4 version was released with more functionalities including image input (Wu et al., 2023).In September 2023, OpenAI (2023) expanded those capabilities by adding a voice function for the ChatGPT application.In January of 2024, chat.openai.com received 1.6 billion visits, with over a quarter of the users being between the ages of 18 and 25, with Japan being in the top 5 countries using ChatGPT at 3.28% of the populace (SimilarWeb, n.d.).This implies that university-aged Japanese students are high among the growing number of ChatGPT users.model AI tool for university assignments, found that students in the experimental group incorporating the real-time aid of an AI chatbot doubled their achievement score from 40.6 to 81.1.Students in the control group who did not use the AI, instead interacting with the instructor in real-time, only increased their score from 43.4 to 65.2 (Essel et al., 2022).
The efficiency of GenAI technology has been discussed in terms of positive experiences, evolving expectations, and negative impressions.(Abdaljaleel, 2024;Chan & Hu, 2023;Fyfe, 2023;Kanabar, 2023;Lepik, 2023;Murray & Williams, 2023;Tossell et al., 2024;Warner, 2023).Kanabar (2023) provides a comprehensive view of student opinions on using ChatGPT in a project-management course.Most of the students reported that the tool was beneficial."Personally, I think that chat GPT is helpful, providing ideas you never thought of before.""It is a very helpful tool not only in doing homework but with professional projects too.And it will make research easier.""ChatGPT is a very helpful tool and can help you with research your topic in-depth and can save time" (pp.394-395).There are many positive statements, yet some include caveats to the benefits of efficiency that are found across relevant literature.Tossell et al. (2024) discuss many of the uses, such as generating ideas, information gathering, clarifying concepts for writing, and providing actionable feedback.However, there were mixed results.Responses indicate that students found ChatGPT best suited as a tool to assist in a "collaborative manner" (p.1075).One student from the study commented, "I thought it was more of an assessment of our editing skills than our opinions on the topic" (p.1075).Fyfe (2023) concurs by stating, "When writing with AI becomes more like the assembly and editing of texts, we change the role of author to something more like an editor, curator, or mediator" (p.1402).One participant in their study added, "I felt more like Dr. Frankenstein, stitching together half sentences and incoherent AI words into something more cohesive" (p.1402).The study described by Fyfe (2023) was designed so that the students had to collaborate with ChatGPT instead of relying on full AI generation.This presented challenges.At first, some students felt that it would be simple." [I]ntegrating artificially generated text into my writing was more of a curse than the blessing I thought it might be when this project was first explained" (p1399).In the end, 87% stated that it would have been easier to write the paper on their own (Fyfe, 2023).

Reliability
Besides the challenge of editing AI-generated writing, another issue is discerning which information is accurate.One student stated, "[S]ometimes the passage said the exact opposite of what was true, but the way it was worded seemed so professional and authentic I was almost convinced" (Fyfe, 2023(Fyfe, , p.1399)).
ChatGPT provides answers in a natural, confident, and verbose manner which can create a false notion of reliability (Shoufan, 2023).The tool sometimes disseminates false information, including references termed "hallucinations" (Kanabar, 2023;Lepik, 2023).Working with an AI tool can feel like collaborating with an efficient "writing buddy", but this phenomenon can have its disillusionment as soon as students find that their partner has provided "hallucinated references and content that needs to be verified" (Lepik, 2023, p.35).This can lead the users from blindly trusting to feeling the need to confirm the generated information (Lepik, 2023).
Tools like ChatGPT derive insights from massive pools of data in a process called deep learning.Some of these sources are "potentially biased or discriminatory" (Tossell et al., 2024(Tossell et al., , p.1079)).The language learning model has no insight into the users' motivation.Nor does it make ethical choices of the text it creates.It is using "algorithmic guesses and more" to respond to prompts (Kanabar, 2023, p.397).ChatGPT cannot self-regulate in terms of misleading information or the ability to gauge credibility, therefore requiring human supervision (Chan & Hu, 2023).GenAI tools have been accused of contributing to the "generation of misinformation, the dissemination of fake news, and even the promotion of plagiarism" (Maciel, 2023, p.111).
However, during their study, Tossell et al. (2024) note the change in student perception from a "cheating tool" to one necessitating "human oversight and calibrated trust" (p.1069).Shoufan (2023) concurs, illustrating that students who use the tool generally gain an understanding of these flaws in ChatGPT and feel hopeful about technical enhancements solving these issues in time.Shoufan (2023) continues by stating that the ideal scenario in which students use ChatGPT would be if they have "prior knowledge" or "adequate background in the relevant field of study" so they can effectively generate useful prompts and "critically evaluate responses provided by the system" (p.38813).
Most research reviewed for this paper agrees that students should not completely rely on text created by GenAI tools.Furthermore, if AI is to be embraced by academic institutions for student use, effort must be put into protecting academic integrity (Chan, 2023;Chan & Hu, 2023;Kanbar, 2023;Lepik, 2023;Shoufan, 2023;Tossell et al., 2023;Van Wyk, 2024).This provides an opportunity for students and educators to consider the ethics of AI in contemporary academia.

Ethics
Since 2023, many scholarly articles (Bishop, 2023;Chan, 2023;Chan & Hu, 2023;Cotton et al., 2023;Fitria, 2023;Fyfe, 2023;Kanabar, 2023;Mohammadkarimi, 2023;Price & Sakellarios, 2023;Smolansky et al., 2023;Van Wyk, 2024) have been published concerning ethics and assessing the use of AI-generated writing in student assignments.According to Kanabar (2023), in the spring semester of 2023, it was becoming evident that strategies to prohibit GenAI would not be realistic.Lepik (2023) suggests that banning the technology would inadvertently entice students to try it, as "forbidden fruit is sweet" (p.35).They also note that it is time-consuming for teachers to police cases of AI-based scholastic misconduct (Lepik, 2023).Furthermore, the use of AI detectors to point out AI-generated writing in student assignments is currently struggling to keep up with GenAI ranging in detection ability from fairly accurate to fair guess (Krishna et al., 2023;Price & Sakellarios, 2023;Walters, 2023;Weber-Wulff et al., 2023).The challenge of pinpointing cheating this way is problematic for In a study by Van Wyk (2024) of academics from a faculty of education, 90% of the participants had favorable views concerning the usefulness of AI software for teaching and learning.Yet, for another survey question, 90% expressed that consequences should be clarified to prohibit cheating."Some participants proposed harsh punishment, such as expulsion from all academic activities, forfeiting of grades or deregistration from courses or qualifications to avoid cheating or academic dishonesty" (Van Wyk, 2024, p.6).Price (2002) proposes that an effective way to form a unified understanding of plagiarism among students is to engage in dialogue with them about ethics.This is especially practical in the context of EFL.Chan (2023) states that there are different perceptions of what constitutes "plagiarism" between cultures.This is not as simple as multiculturalism however, as illustrated by a study of ten female Australian undergraduate business school students that compounds this by revealing there was not "a single universal ethical perspective" on the use of GenAI amongst any of the participants of similar demographics (Murray & Williams, 2023).Therefore, a classroom discussion on plagiarism could be meaningful both practically and pedagogically before using GenAI for academic assignments.
Plagiarism can be a subjective concept (Fyfe, 2023).Especially when viewed from a multicultural lens (Chan, 2023).The idea becomes further distorted with the complexities of utilizing AI tools.What constitutes original thought if one is using software that provides assistance based on deep learning (Fyfe, 2023)?What of automated writing evaluation (AWE) software such as Grammarly?Fritria (2021) notes that a number of these grammar-checking tools currently run on AI, though utilized differently than ChatGPT and other GenAI.Therefore, lines may be blurred when considering a student following suggestions from Grammarly and other AWE."If the teacher is grading the student based on their writing ability, perhaps grammar-checking software is giving the teacher a skewed sample of the student's true writing" (Price & Sakellarios, 2023).Kanabar (2023) notes that the results of their study found that "Students consider using ChatGPT unethical to use…Not exercising the traditional approach to conducting research or homework was troubling to many students" (p.397).As the use of AI grows, ethical principles may transform as technology nudges humanity to change.Boyle (2016) comments from a traditional humanistic viewpoint on ethics in writing, resettling within new parameters of emerging technology by stating, "[E]thics in a posthuman practice are not ideals imposed upon conditions for actions we ought do but are instead ongoing exercises whose aim is to compose new capacities for conducting ourselves within expanded media ecologies" (p.549).
Cheating and plagiarism are viewed in a relative sense through a multicultural lens (Chan, 2023).In the context of EFL classes, due to cultural diversity, students and teachers may hold divergent perspectives regarding the definition of academic dishonesty concerning the utilization of GenAI software.In Mohammadkarimi's (2023) study on perceptions of academic dishonesty, all 67 EFL teacher participants agreed that AI can harm "students' commitment to academic honesty, perceiving it as enabling dishonesty and hindering skill development" (p.105).However, another perspective on ethics and GenAI in an EFL context could coincide with Price's (2002) suggestion that the relative concept of plagiarism can be an effective starting point from which to have an impactful discourse.

EFL Utilization
Stephen Atlas ( 2023) has a chapter devoted to "ChatGPT for Communication" which describes several productive applications for GenAI in the context of .For example, ChatGPT can "simulate conversations"(p.64).It can be an effective "virtual language partner" providing back-and-forth dialogue in a "low-stakes setting" where students aren't hindered by shame or anxiety (p.64).Though Chan (2023) mentions that for language students, creating adequate prompts for ChatGPT in the target language has its own linguistic difficulties, Atlas (2023) alternatively proposes that the creation of prompts and subsequent responses can be opportunities for conversation practice with possibilities to explore different scenarios.
ChatGPT can be used as a translator.Atlas (2023) expands on the benefits of this in a languagelearning context by pointing out ChatGPT's ability to use and understand natural language including "idiomatic expressions and recognizing cultural references" (p.67).Lepik (2023) adds that while not always perfect, ChatGPT can translate even scholarly texts between languages, and "it outperformed Google Translate" (p.34).Since Professor Atlas' book was published, OpenAI (2023) announced upgrades to ChatGPT including the capability to communicate with voice.Therefore, ChatGPT can currently be used as a real-time spoken word translator.
With opportunity and creativity, ChatGPT can be used for a variety of situations to aid EFL learners.Barrett & Pack (2023) refer to a study by Fan (2023) that examines EFL students' use of AWE programs such as Grammarly.These tools are useful for advising writers with actionable feedback.However, EFL learners do not always understand the feedback provided by the AWE.Therefore, ChatGPT can be used to facilitate a deeper understanding of AWE suggestions through its translation abilities (Barrett & Pack, 2023).

Unique Insights
The utilization of ChatGPT and other GenAI tools adds a complex array of benefits, pitfalls, and transformative scenarios in many aspects of life.In academia, educators and students are discovering novel opportunities to use the technology to enhance learning.Unique insights emerge that address a variety of facets concerning GenAI's influence on education.Notably, Kanabar (2023) identifies what they term "Student Syndrome" (p.388).This is characterized by procrastination evident in students.Kanabar (2023) proposes that by utilizing ChatGPT, students can produce research and assignments of superior quality even when facing time constraints.2023) reference a survey of one thousand university students conducted by Welding (2023), revealing that 43% of respondents utilized GenAI for academic tasks, with over half of those relying on the tools for major assignments and exams.However, despite the widespread use of AI, there remains a lack of institutional guidance on ethical considerations surrounding its use.This highlights the need for educational institutions to address ethical concerns and provide students with the necessary support and guidance.
The enthusiasm of students towards the future of GenAI is evident, as the following quote, by a student involved in research done by Tossell et al. (2024) illustrates, "Coolest assignment I've done to date.I think tools like ChatGPT will change our future and assignments like these are paramount to understanding the direction we want to take them" (p.1076).

METHODOLOGY Research Design Overview
The research design of this study incorporates a qualitative thematic analysis focusing on student use of ChatGPT in assignments.The research is based on data gathered from a fiveweek project incorporating a survey (see Appendix A), semi-structured interviews, informal discussions, and follow-up correspondence.Data was collected from three participants (n=3) in a class-based environment.The design of the following analysis incorporates both open coding and axial coding in a layered process of successive passes revealing overarching themes in the data.

Study Participants
The participants of this study were students enrolled in an elective EFL communication course offered by a liberal arts department of a Tokyo-based university.Spanning from first-year to fourth-year students, the participants voluntarily enrolled for the class which speaks of their overall motivation to enhance their English skills.
The students' English language proficiency levels ranged from intermediate to advanced.Participant 1 had an Eiken level 2, Participant 2 reported scoring a 6.5 on the IELTS test, and Participant 3 received a score of 980 on the TOEIC.To consolidate the understanding of the skill levels, these test scores translate to CEFR proficiency levels of B1, B2, and C1 respectively, indicating progressively advanced abilities in speaking, listening, reading, and writing, allowing for effective communication, comprehension, and speaking on various topics (MEXT, 2018;Tannenbaum & Wylie, 2008).
The group consisted of two females and one male, but I do not consider any information derived from this study to be gender specific, so there will be no mention of such.Participant 1 was a first-year student, and the other two participants were fourth-year students.This distinction may subtly shape their perspectives due to the proximity of post-graduate life, though it was not specifically mentioned.

Introduction of the Research Plan
On the first day of the class, I introduced the idea and asked the students if they would be interested in doing a study on student use of ChatGPT for academic assignments.They were very positive about the idea.In an informal preliminary discussion, they said they had limited exposure to ChatGPT.When discussing AI, one mentioned using Google Translate, another said they used Apple's Siri, and the third said they regularly used Grammarly.Although one had experimented with ChatGPT none of them had fully used it for academic assignments.In the 6 th week of class, I formally introduced the research plan and explained the design of the study to the students.I emphasized that even though the activities involved with the study were concurrent with the established plan for the class in the second half of the semester, they were free to opt out of the research at any time with no adverse consequences to their course evaluation.All were enthusiastic about the plan and remained so throughout the study.

Data Collection
To investigate the participants' impressions of using ChatGPT for academic assignments, a careful design of lessons with a flipped-class structure was implemented to facilitate data collection.The study unfolded over several weeks, during which the participants engaged in academic writing and class discussions (in English) on diverse topics related to global and national current affairs.The topics included population (growth and decline), economics (inflation vs. deflation), the pros and cons of genetically modified foods, and potential revisions to the Japanese Constitution.
Each week a new topic was announced.The students were given open-ended essay questions that could be broadly interpreted.For example: "Do you think that there is a population problem?If Yes -How can we fix the problem?If No -Explain why you think that there is not a problem."It is not defined whether this question is referring to the exponential growth of the global population or the present situation in Japan, which is that of population decline.These questions were to be researched and answered as essays in academic writing.In the following week's class, the participants held oral discussions explaining their answers and clarifying their positions.
The research participants were given specific directives.The instructions were for the participants to utilize ChatGPT for at least one of their assignments, and to write at least one other without any help from GenAI.They were free to mix and match as desired for the other two assignments.The students were instructed not to disclose which assignments they wrote with GenAI.Therefore, during the oral discussions, they were required to uphold and elaborate The structured approach of this study aimed to facilitate reflection of the participants on their experiences.The plan cultivated a secure environment from which the students could explore some of the possibilities of AI technology for academic work.The design allowed them to compare the different processes in an emotional and analytical sense.It provided a context from which the participants could engage with the technology in a real-world scenario yet buffeted from any possible negative consequences under the umbrella of this study.That and the intimate group size were deemed ideal to illicit honest and unfiltered perspectives for the data that was collected.
The culmination of the data collection process occurred after the fourth assignment was completed.The participants responded to survey questions (see Appendix A), and in the fifth week, we held semi-structured interviews.The surveys, interviews, and follow-up correspondence offered valuable insights into their perceptions and impressions of the experience.

Analysis
For this study, an inductive coding approach was employed in the thematic analysis process.Initially, I used open coding to explore the data without predetermined categories, aiming to identify and understand emerging themes.The first set of passes yielded 17 code categories including benefits, criticisms, cultural ideas, dangers, enjoyment, ethics, and justification.Subsequently, I conducted axial coding to structure those codes, combining them whenever possible and organizing the data in a manner that more closely suggested final themes.
At this point, all the participants received a copy of the codes for review and comment.No opposition was raised by any of the participants to the direction that the coding had taken.One participant wrote a responding letter: There is nothing I would like to complain about.Thank you for adding what I wrote in my follow-up email.If any, it may be typos, especially in the interview transcript, so I don't really care about it.It was so much fun to read the coding things.It was a rare opportunity to know how others perceive my opinions!(Participant 3, personal communication, September 22, 2023) The next step involved synthesizing the codes into final themes for use in further analysis, interpretation, and consideration.These themes were efficiency, reliability, ethics, EFL utilization, and unique insights.They were selected due to their conciseness and universality across the coded data.
To verify and ensure the reliability of the coding process, ChatGPT was used in the final step of code retrieval.I input all the coded data into the AI, including the survey responses, semi-structured interview transcripts, and follow-up correspondence letters.This was done sequentially, separating the data from each participant to maintain methodological integrity and ensure the nuances of individual perspectives were maintained.I provided the five themes to ChatGPT and asked it to verify them by providing specific examples in the form of quotes.The program analyzed all the data and retrieved relevant quotes to illustrate participants' perspectives based on the finalized themes from the analytic process described above.

FINDINGS
The results of the study are presented below.The students were free to choose which topics to enlist the use of ChatGPT.Their choices and methods varied.For example, Participant 1 (P1) used AI only for the first assignment on population.They submitted the results verbatim even though they did not agree with the opinions expressed by the AI.Participant 2 (P2) used it three times exploring different techniques each time.For the question on economics, they mostly used AI-generated work after rearranging sections.For the assignment on the Japanese Constitution, they described a collaboration of "fifty-fifty" with the opinions written by P2 and the explanations written by the AI.For the topic of genetically modified foods P2 simply garnered ideas from ChatGPT but wrote the essay on their own.Participant 3 (P3) used ChatGPT for the assignment on economics.They asked numerous questions of ChatGPT in Japanese, and through dialogue with the GenAI narrowed the focus and compiled essays, which P3 translated into English on their own.Each of the students had an individualized experience working with ChatGPT for the assignments.After completing the four weeks of assignments, they shared their experiences by answering long-form questions in a survey, which were further expanded upon in a semistructured interview.Later, there were some follow-up correspondence emails used for clarification and expansion of ideas.Upon analysis of their responses and discussions during the study, five shared themes emerged.Efficiency, reliability, ethics, and EFL utilization were mentioned by all, as well as some unique insights by each, warranting particular attention.

Efficiency
All the participants acknowledged the speed of ChatGPT.However, their considerations on time and the scenarios discussed varied.P1 expressed that ChatGPT can be used by students to save time and effort when writing an assignment.Yet, they stated, "I think it's more difficult than the normal method for writing reports."P1's opinion was based on their experience using a completely AI-generated answer for the assignment on population problems.They mentioned that their opinion of the issue differed from ChatGPT's and that when prompted, the tool did not provide sufficient support for the generated information.This experience led P1 to express reservations about the use of the tool because to write a supported and quality academic assignment one would need to take the time to verify all the information generated by AI.P2 expressed, "It's really useful, and I could shorten the time to spend on tasks."They emphasized the efficiency of AI in generating ideas, finding appropriate words or phrases, and understanding terminology related to the topics.One notable instance that P2 mentioned was the request of ChatGPT to write a 400-word essay.The ability to dictate the limitations of a word count was one of several examples mentioned that gave them the opinion that the technology could be easily adapted for a variety of tasks.
P3 concurred with P1 in the idea that the speed of ChatGPT is weighted by the inefficiency of working with the tool for someone who is more hindered by exactitude.This can be an issue, especially in the context of these assignments.The students were not simply to write an essay and submit that for final evaluation alone.They also needed to come into the classroom for group discussion and present their ideas, supporting and even debating the points in some cases.P3 stated, "[W]hen you argue based on 'borrowed' opinion, it takes extra time to comprehend what the idea you are writing is."In connection to the concept of using ChatGPT for the expression of concepts that may be deemed controversial, P3 pointed out that the tool often appears to be taking strides at maintaining a neutral tone.In a case where the writer wants to fully support one side of a debate, P3 stated that for some it could be a waste of time to discuss opinions with AI.
P3 does see that GenAI tools such as ChatGPT have academic use that can be deemed efficient, from a certain point of view.P3 contends that not all homework in the Japanese education system is inherently productive, rather what might be described as busy work.However, the students must do these assignments to maintain their scores to get into a good high school or university.So, in these cases, P3 suggests that AI could be used to save time on less meaningful assignments.

Reliability
The next theme exhibits some overlap with efficiency, as unreliable work generated by ChatGPT can lead to inefficiencies, necessitating additional time for double-checking or fixing information in assignments.P1 experienced such a situation, disagreeing with certain information generated by AI and challenging it by asking, "[W]hy do you think so?"However, P1 reported that the tool could not show satisfactory support for the generated information.As P1 was working in the context of academic writing, they felt it wasn't useful for writing scholarly essays as there were no sources provided.P1 went on to express that ChatGPT cannot show quality sources because AI's "opinions are randomly put together from the Internet."P2 used ChatGPT for more assignments than the others, but "did not depend on ChatGPT 100%."According to them, students should not completely rely on ChatGPT, but instead for support.ChatGPT can be utilized for broad strokes such as summarizing ideas and thus aiding one's understanding.However, that can have its limitations.P2 mentioned the difficulty during the in-class discussion of the economics topic, for which they had heavily relied on GenAI.Once the conversation deviated from the prepared statements, P2 found it challenging to give comprehensive answers.

94
P3 encountered comparable challenges transforming AI written content into a discussion noting that it is difficult to discern the flow of logic when the ideas are not originally their own.P3 goes on to state that ChatGPT's answers are quite confident which can be compelling.This can be a trap for "immature writers and students."They followed up by stating, "Advanced writers who can doubt AI's answers do not need AI's help for creating their original ideas."Overall, P3 expressed that depending on content generated by AI involves indirectly drawing from the tool's existing pool of information.This not only places oneself in the position of using dubious information but also raises the possibility of plagiarism by proxy.

Ethics
Plagiarism was a concern shared by the participants to varying degrees, each offering different perspectives.For example, P1 did not mention ethics concerning the assignments that were done for the class discussions.Instead, they raised concerns regarding images generated by AI.P1 expressed an interest in AI-generated imagery but mentioned stories that they had heard of artists challenging AI-generated artwork in court because of copyright infringement.P2's expressed the opinion that using ChatGPT for getting ideas is ethical.However, using sentences verbatim as written by the tools is "problematic."In terms of ethical citation of AIgenerated writing, P2 stated that they do not think it is necessary to list a reference for ChatGPT, as it is not a living creative entity and therefore not generating opinions but automated responses.
P3's opinions on the ethics were more straightforward.Instead of considering ChatGPT's generated ideas as those of a complex automaton, they note that AI learns from "existing things and information."Therefore, to use GenAI's ideas as one's own is "indirectly plagiarism."P3 declares that "it should be forbidden to copy and paste what AI generates as your own idea."However, in connection to what they previously stated about efficiency and pragmatic use by students, when faced with less meaningful assignments, P3 considers that students will still use GenAI for assignments even if they are aware of the moral ambiguity and ethical implications.

EFL Utilization
Given that this study drew from a series of assignments within an EFL communication class, all participants referenced their experience with ChatGPT in the context of language learning.P1 specifically pointed out the challenge of understanding the vocabulary used by ChatGPT.The student had trouble with some of the terms, especially since the opinions generated to answer the essay question were not shared by the participant.P1 reported struggling with this when it came to the in-class discussion."Of course, I searched these meanings before discussion, but I couldn't use these words properly."P2 shared the idea that AI-generated assignments could put EFL students in a challenging position when faced with elaborating on concepts during real-time discussions.They also note the potential negative effects on language learning if a teacher is unable to detect the use of GenAI software in students' work.In such cases, the teacher may be unable to recognize areas for improvement, hindering language development.However, P2 also discussed positive aspects of AI in language learning.They noted that Japanese learners tend to be shy, suggesting that interacting with a computer could benefit them.P2 also notes the shortage of teachers for other foreign languages in Japan, proposing that technology could help fill the gaps by providing access to diverse language resources and instruction.
P3 focused on the translation ability of ChatGPT.They had previously worked with Google Translate regularly.As a result of the first-time use of ChatGPT for this project, P3 stated that its translation ability "probably surpasses Google Translate," especially in terms of natural language."I was surprised at ChatGPT's translation skills."

Unique Insights
The environment in which this small study was done facilitated open-ended observations and reflections on the use of the new technology for academic purposes.Apart from the four themes listed above were other insights and perspectives that add more context to the emic perspective that has been a goal of this study.For example, P1 used ChatGPT the least of all participants.After the initial experience, they did not use AI for any of the other essay assignments.When asked for a final statement on the study, they diverted from the use of text-generating AI to image-generating tools."I like to see illustrations which are drawn by using AI tool."Yet, they expressed some trepidation due to concerns about potential copyright issues related to imagegenerating AI technology.P2 explained that in the flipped-class structure of the communication course used as the setting for this study, the students prepared for discussions creating academic essays beforehand.These essays formed the basis for oral discussion during the group sessions.This setup provided an environment conducive to sharing ideas and gaining insights from diverse perspectives.P2 pointed out the contrast with many other classes, where students typically write and submit reports without engaging in meaningful exchanges to broaden their outlooks.To address this gap, P2 suggested that students could use ChatGPT to explore viewpoints beyond their own.P3 discussed the phenomenon of GenAI's rising popularity through the lens of social psychology.Their comment, which draws a connection between the eagerness for knowledge and the desire for social inclusion, offers a unique perspective on why individuals turn to GenAI for information.The viewpoint is best conveyed in their own words: AI and its prompt responses satisfies people's desire for information.Recently, we have been accustomed to reach new information.Internet and SNS greatly contribute to swift spread of information, and people are becoming too impatient to be ignorant.Or we are quite scared of being left out.I wonder desire for knowledge is somewhat similar to desire for inclusion.AI can offer adequate amount of new information regardless of its credibility in a snap.It comforts people that they are not left out.

Overall Perceptions
The range of viewpoints expressed by this small group of Japanese university EFL students shows an emic perspective reflecting both caution and optimism regarding the use of GenAI for academic assignments.The participants in this study echoed those reported in the literature.As mentioned in Kanabar's (2023) study, students appreciate the efficiency of ChatGPT for generating novel ideas, bringing to light an array of viewpoints, and overall streamlining the writing process.However, the students involved in the study shared concerns similar to those discussed in Fyfe (2023) regarding the challenge of managing AI-generated content in a way that pulls authorial control into a realm of editing over creating.
Furthermore, the participants expressed apprehension about the validity of information and its sources as presented in AI-generated writing, such as was addressed in several studies mentioned in the literature (Chan & Hu, 2023;Kanabar, 2023;Lepik, 2023;Maciel, 2023;Shoufan, 2023;Tossel et al., 2024).Even within this group of three participants several nuances in perspective were evident, underscoring the necessity for educators to adopt a flexible approach, acknowledging both the potential benefits and challenges of AI integration into the academic and language learning environments.

Implications for Pedagogy
Pedagogical adjustments aimed at ensuring the reliability and ethical sourcing of information used in AI-generated content for academic assignments require careful consideration, especially within the realm of language learning.In this context, where diverse linguistic and cultural backgrounds intersect, establishing a shared understanding of academic integrity is crucial, particularly in EFL settings where addressing plagiarism demands a culturally sensitive and measured approach, as mentioned in Chan (2023).
To effectively engage students on this important topic, tailored assignments, within the EFL courses, that focus on ethics and responsible use of AI, can be invaluable.Building on Price's (2002) suggestions of involving students in dialogue concerning plagiarism and discussions of Tossell et al. (2024) and Lepik (2023) regarding students' evolving approach to writing with AI, these assignments could provide a foundation for scaffolding digital literacy skills and ethical considerations into assignments.In this manner, educators can empower students to discern reliable information and critically evaluate AI-generated content, while also reinforcing ethics, academic integrity, and cultural sensitivity.

Further Research
The initial concept for this research was that of a longitudinal study originating with this data from the first year of mainstream ChatGPT access by students for academic work and extending to data collected along the same lines in the future.As mentioned in Tossell et al.'s (2024) exploration of the incorporation of ChatGPT in academic assignments, student perceptions change from that of a "cheating tool" to a resource albeit necessitating human oversight.What Longitudinal studies exploring the trajectory of student attitudes and engagement with AI, as it becomes more incorporated into their lives, could inform educators on directions to evolve practices and pedagogical approaches.

Limitations
While this study offers valuable insight into the experiences and perceptions of students regarding the use of ChatGPT in academic assignments, it is important to acknowledge the several limitations of this study.As the lecturer for the course and the primary researcher, I had a teacher-student relationship with the participants which could have influenced their responses.The number of participants (n=3) provided a comfortable environment to share thoughts but is limited and comes from only one class at one university in the capital city of the country.Also, the participants were all EFL students with a somewhat similar proficiency level.This sample restricts the ability to comprehensively generalize the findings.
The study spanned five weeks.A larger study including a broader sample of participants and a longer period for data collection could enhance the scope of information utilized for the study.Furthermore, there are a growing number of similar GenAI tools, yet the participants in this study all chose to work with ChatGPT.Therefore, the findings may not be applicable to all AI systems or platforms.

CONCLUSION
In conclusion, this study gives insight into a small group of Japanese university EFL students' impressions of ChatGPT when used for academic assignments in the first university semester following its release.By drawing on both the literature review and findings sections, this study highlights the importance of adopting a measured and culturally sensitive approach to AI integration in language learning pedagogy.As GenAI becomes more commonplace, educators must remain aware of current trends that could aid in developing engaging and productive methods to manage student use of the tools.The integration of GenAI into education requires the balance of potential benefits with ethical considerations and pedagogical imperatives, ensuring that students are equipped with the skills and knowledge to navigate the digital landscape responsibly and ethically.
://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK 87 educators since student gains in comprehension and skills are unclear and difficult to evaluate (Smolansky et al., 2023).
Print ISSN: 2054-6297(Print) Online ISSN: 2054-6300 (Online) Website: https://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK 89 Murray and Williams ( ://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK 90 Researcher-Participant Relationship I was the lecturer for the class and therefore held a teacher-student relationship with all participants.I had previously taught both senior students in an academic writing course and thus had an established rapport with them.With only three participants in the class, we enjoyed a more relaxed and open environment, allowing for free-flowing conversations and a sense of familiarity.
://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK 91 on the positions expressed in their written essay submissions, even if they were fully generated by ChatGPT.
://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK 93 ://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK (Participant 3, Study Survey, June 30, 2023) Print ISSN: 2054-6297(Print) Online ISSN: 2054-6300 (Online) Website: https://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK ://www.eajournals.org/Publication of the European Centre for Research Training and Development-UK 97 will be the student perceptions of GenAI two years from now, ten years from now?