AI in an L2 translation class

Chapter 10 AI in an L2 translation class Sonia Vandepitte Universiteit Gent, Belgium Teaching translation from the so-called native language (L1) into a second language (L2) requires approaches different from teaching translation into the L1. With the advent of some generative AI tools made freely accessible everywhere, a powerful resource has become available that can be put to good use in the L2 translation class. This contribution will report on such an exploratory class that involved a translation assignment in which ChatGPT was used as an editing tool. It will explain the translation setting and detail the design of how AI was integrated into translation training from a European language into English. It will describe the students’ learning objectives and the steps taken to facilitate their paths towards those goals. By illustrating the students’ main demonstrated competences, the class design will be reflected upon and some ideas for alternative future classes will be suggested. It will be shown that integrating AI into the L2 translation classroom enriches L2 translation learning, and a constructive critical attitude will be called for. 1 Introduction Autumn 2022 brought the tool of generative AI (GenAI) to the general public. In a translation training class, this meant that students could now produce translations and edit texts on the basis of the multilingual capabilities of a user-friendly, conversational device which relied on a huge knowledge database. Because it took into account contextual features, its translations in well-resourced language pairs, and especially with the English language, turned out to preserve tone and meaning better than any automated tool before. In addition, its translations of idioms and culture-specific concepts in those languages also proved to contain fewer failures than the previous models of neural machine translation (MT). If Sonia Vandepitte. 2026. AI in an L2 translation class. In JC Penet, Joss Moorkens & Masaru Yamada (eds.), Teaching translation in the age of generative AI: New paradigm, new learning?, 191–210. Berlin: Language Science Press. DOI: 10.5281/zenodo.17641082 Sonia Vandepitte any organisation could use such a tool, the importance of making students familiar with it in their translation training therefore also became essential. The next section will review the most relevant characteristics of a training class teaching translation into English as the students’ second language, translation in a GenAI context, and some pioneering projects that already integrated GenAI into a translation class from the outset. 2 Literature review 2.1 Teaching translation into the L2 Translation into a second language, i.e. a language which is not the translator’s main language, often abbreviated as L2, is often considered to require a higher cognitive load (Krings 1986, Göpferich 2017), although various scholars have demonstrated the high quality of the outcome of such translation processes (e.g. Grosman et al. 2000). Translation training programmes, too, often contain modules with practice in this translation direction. The reason is simple: in many countries there is a need for translation into languages, usually a so-called dominant language or a lingua franca (see e.g. House 2018), for which no translators are available. Hence, training programmes from the EMT L2 Translation working group (Vandepitte 2023) have collected a set of best practices for teaching L2 translation. First, it was agreed that most students were somewhat less eager to translate into an L2, had less confidence in their abilities, and were more hesitant about their progress than in their translation into their main language. First, those considerations required appropriate approaches to the general course design. It was suggested that the inclusion of real translation projects (if possible, with extra remuneration for students) worked favourably. If they also include cooperation with fellow class students who can act as bilingual revisers for their first drafts or other students (abroad) who can collaborate as monolingual reviewers because their main language is the targeted L2, students’ confidence in their ability to produce a well-written translation is boosted. If students’ initial competence varies considerably, or if classes are large, this discrepancy may be identified at the beginning of the course and groups organised in such a way that better-skilled students support others (enhancing students’ social skills) As far as class design is concerned, the following best practices were mentioned as suitable to produce better insight into the translation task ahead. In the first place, much attention should be devoted to text choice. Especially, the relevance, length and topic of the texts have an impact on students’ willingness to 192 10 AI in an L2 translation class start a difficult task. Texts commonly required by the translation market or texts shared by professional organisations are preferable. A topic that is too specific will render the text too demanding and an opportunity for students to observe their own progress will be lost. The teacher can provide specialised glossaries, introduce feed-up sessions, in which both grammatical, lexical and cultural stumbling blocks are pointed out, let the students make a critical analysis of the communicative situation (the translation brief), or organise a class discussion on the domain to which the text topic belongs. Students can receive peer and teacher feedback when first and final drafts are published anonymously on the learning platform, allowing students the opportunity to learn from each other’s work. 2.2 Translation in a generative AI context With the development of GenAI (Eloundou et al. 2023, Felten et al. 2023, Akhtar 2024), translation practices are offered a further step in the automation process. While the use of MT (both statistical and neural) has been amply discussed (e.g. Moorkens et al. 2025), research of the use of a chatbot like ChatGPT (OpenAI n.d., Roumeliotis & Tselikas 2023) in the translation process is still in its infancy. It focuses mainly on the quality performance of the generative tool in comparison to that of MT. Robinson et al. (2023), for instance, found that the amount of resources in a particular language plays an essential role. Their study involving about 200 languages showed that quality measures in terms of two different metrics of translations with languages for which there is a high volume of resources for GPT models were as high as or equally high as those for traditional MT models. With most languages having a low volume of resources, however, GPT model performance consistently lagged behind. Geng et al. (2024), too, emphasised that non-English large language models (LLMs) can only be developed if available data sets are large enough. Mujadia et al. (2023), on the other hand, report positively on their experiments with translations involving English and 22 Indian languages using LLMs based on Meta’s LLaMa. Other scholarly work on translation by ChatGPT has been carried out in different knowledge domains. Ülkü (2023) focussed on the hospitality industry services that will likely benefit from the use of GPT-4. He predicted that real-time translation by means of the generative model will facilitate effective communication between hotel staff and guests, but did not provide any criticism or caution, nor any data to support his optimistic belief in the translation quality of the tool. Teng (2024) researched the field of poetry translation and comparatively explored faithfulness, expressiveness, and elegance in both a human and GPT-4 193 Sonia Vandepitte translation of a Chinese poem. Cao & Zhong (2023) carried out a statistical investigation to assess the translation quality of a Chinese recipe into English by GPT4. Drawing on both automatic and human evaluation systems, they found that together with other LLMs, GPT-4 outperformed the fine-tuned MT models and the fine-tuned multilingual encoder-decoder models. GPT-4 created a translation that incorporated cultural adaptations and alternative names for ingredients. Notably, it was more closely aligned with the source text than the human reference text, even though it produced double the number of tokens of the translations produced by the other models. Within the legal area, Giampieri (2024) submitted an English arbitration clause to the GPT chatbot of 2024 and prompted it to translate it into Italian. She reported some lexical inaccuracies and inconsistencies that needed to be post-edited. In summary, the quality as judged by these preliminary studies, which were not yet available when the decision was taken to integrate GenAI into the class, clearly depends on a number of factors. First, the number of resources available to the generative tool plays a big role, and languages for which there are few resources yield poor AI-tool output. However, for a language pair like ChineseEnglish, ChatGPT outperformed other MT-models and, in particular, also showed successful translation choices in cultural contexts. Secondly, the text type also has an impact on the translation quality of the generative tool, with legal texts, for instance, requiring special attention to how the source text meaning is rendered. 2.3 Integrating generative AI into the translation class Including GenAI such as ChatGPT in education has recently been mentioned in various reports as a promising aid to improve students’ writing (Cao & Zhong 2023, Labadze et al. 2023, Ouyang et al. 2023), and, in particular, their writing in English as a Foreign or Second Language (EFL/ESL) (Fitria 2023). One advantage of AI-powered chatbots is clear: ChatGPT can provide instant and personal feedback at a time when the learner’s mind is set to learn, so that its relevance and impact may be higher than when they receive similar feedback at a later juncture. This will fundamentally change L2 learning: the AI tools will be able to detect gaps in individual students’ performances and even give feedback and alternative exercises. The worldwide availability of ChatGPT towards the end of 2022 also raised translation teachers’ awareness that the new technological developments after the introduction of CAT tools and the various forms of MT would bring about yet another set of changes to the translation process. 194 10 AI in an L2 translation class This awareness led to one of the first attitudinal examinations: Sahari et al.’s study (2023) investigated the opinions of both translation students and teachers about the usefulness of ChatGPT when compared with Google Translate. They concluded that students and teachers did not agree on which tool was preferable. However, the AI tool was recognised as a facilitator of typing and spelling, but the tool’s results for fine-tuning and double-checking could not be trusted. Wang et al. (2024) also investigated students’ attitudes to the use of ChatGPT in translation. They did so within the framework of the Unified Theory of Acceptance and Use of Technology, whose model points at factors influencing participants’ attitudes. They found that despite a positive attitude towards ChatGPT, students still commonly used MT systems. Gao (2024) describes a pilot programme, in which the texts of political speeches were submitted to ChatGPT 3.5. The translations produced were discussed by the students and compared to translations produced by professional UN translators. They found that the GPT model was successful in producing first draft translations with appropriate lexical and grammatical information. However, it was not yet capable of sufficiently dealing with complex linguistic, cultural, ideological or other contextual nuances. Cao & Zhong (2023) produce statistical evidence for the usefulness of ChatGPT in the context of a translation class. They carried out an analysis of translations submitted by Chinese ESL/EFL learners and compared both the overall quality (measured using BLEU scores) and some linguistic features (Coh-Metrix) of three types of student translations of a short press release: some translations only underwent self-feedback, others elicited teacher feedback, and a third set received feedback by ChatGPT. The results of this analysis indicated higher BLEU scores and syntactic Coh-Metrix scores for the translations produced after teacher feedback and the lowest score was given to the GPT feedback generated translations. Translations with ChatGPT feedback only scored better in the lexical domain. They suggest that ChatGPT “relies on replicating patterns from its training data rather than performing true syntactic analysis, thereby failing to offer substantive feedback” (Cao & Zhong 2023:13). While Gao’s program integrated GenAI into the translation class process as a tool that is prompted to perform the translation itself, Cao and Zhong designed a process in which the tool is requested to act like an educator and to provide feedback to the students’ translations. The following sections will describe and discuss an exploratory course which integrated ChatGPT into the students’ translation process in yet another way. This course from a European EMT translation programme in Spring of 2023 had students translate into an L2, a task in which students could benefit if they 195 Sonia Vandepitte could access the opinions of a target language speaker. The course therefore had students use ChatGPT as a monolingual reviser of their own translation work. 3 Course design for ChatGPT in an L2 translation class As mentioned above, the course reported here had students use a GenAI tool, mainly for editing. The course was part of a Master’s translation programme and required students to be able to translate from their L1 into English as an L2. The use of the AI-tool at the revision/review stage of their translation process would therefore give them more confidence about the choices they had made in their L2 text production. The class included 28 students, and although English is not the only L2 in their programme, their level of English proficiency can be described as C1-C2. At the time of submitting course descriptions, GenAI had not yet become publicly accessible and was therefore not included in the course description. Hence, it was a primary ethical and legal requirement of the course design to assess whether the set of learning objectives already identified in the course description contractually allowed for the introduction of AI into this final-year master’s course. It turned out that the most general learning objective, acting in an unpredictable and complex context, and the learning objective including the appropriate methods applied during the student’s individual cognitive translation process, applying the appropriate digital (reference) tools, made it possible to include an exercise with GenAI. At the start of the course, before some of the research reviewed above was published, few students had heard about GenAI, which they would become more familiar with in their compulsory language technology module of the programme. In order to introduce them very quickly to the GenAI tools, the students were presented a straightforward list with direct AI website links from which they could choose one, including ChatGPT-3.5, Perplexity and Jasper. There was also a reference to future developments such as Meta’s Llama and Google Gemini. The student’s use of a generative tool was integrated into their translation process as follows. First, students were asked to translate a passage from the source text and place it on the course collaboration platform. Second, they were asked to revise and review a team partner’s translation independently. Third, the original translator would submit the edited translation to an AI-tool, asking it to improve the text or giving it another prompt of their choice, and pasting the outcome underneath the edited translation. Finally, they would compare the AI tool version with the revised translation and indicate any differences with 196 10 AI in an L2 translation class an annotation; they would put a + sign in the balloon if they agreed with the proposed change or write an alternative solution, if they did not agree with the AI tool version. Similarly to previous non-AI translation processes, students were reminded to adopt a frame of mind that is similar to that of a reader of the target text (in this case, a foreign student, see below) and to ask themselves what expectations the typical target reader might have in terms of register and terminology. They were also reminded to interpret the source text as well as possible, and to be aware of spelling, layout, grammar and terminology. As a result, students’ work would look as follows. Figure 1 shows the first draft submission of a student and the two editing comments made by their peers, one of which has been complemented with a teacher comment. Figure 1: A student’s first draft translation with peer and teacher feedback. Figure 2 illustrates the GPT-3.5 output of an editing prompt given by the original translation student and her own annotated comments with pluses or alternative solutions. Further noteworthy details of the translation setting in this course are the following. First, a real-world client was involved: the university’s legal office needed a translation of its University Codex from the students’ main language into English. This source text meant that legal language was involved and their 197 Sonia Vandepitte Figure 2: The GPT-edited version of the peer-revised draft (from Figure 1) with the translating student’s comments. target audience consisted of foreign staff and students. Since this translation setting included a paying client, students would also be remunerated (with book tokens) for their extra translation work. In addition, the client was available for any questions that might arise relating to the translation and students would, hence, be able to practice client communication. For this purpose, a working document was prepared, called ‘Questions to the client’, in which students could write about their translation problems. This document was discussed in class and questions that were retained for referral to the client were collected and sent by the teacher. Each communication from the client was then also provided to the whole class in the same document. The students could further consult both a parallel and comparable corpus and glossaries, which had been compiled in previous classes for translations for the same client and the same target text. Finally, as in other L2 translation classes, students were able to work on a student collaboration platform, in which each student’s contribution was visible relatively anonymously to everyone. All materials provided by the students who participated in this project have received student consent for research purposes and may be used to illustrate what the students were capable of in this pilot project. 4 Findings First, students did not do much experimentation with a variety of GenAI tools but only applied ChatGPT and only used it in their revision process. It also turned out that some students confused Machine-Translated texts with AI-generated texts and referred to, for instance, Deep-L, which was already familiar to them. However, students did not prompt ChatGPT to produce a translation. 198 10 AI in an L2 translation class More importantly, students were able to identify differences between the translation version that had undergone monolingual and bilingual revision by their peers and the text as edited by the GenAI tool used. In particular, they were not only able to identify better formulations in the L2, but also unwelcome omissions, additions, and even (sometimes minor) differences of meaning between their own versions and the ones provided by ChatGPT. The following set of examples shows that students largely agreed with ChatGPT improvements. They involved the more idiomatic use of the Anglo-Saxon genitive (less frequently used in their own native language) (1), the avoidance of a split to-infinitive construction (2) and the use of the active rather than the passive voice (3): (1) a. Peer-revised draft an opinion by the Executive Board is requested b. GPT-edited version the Executive Board’s opinion is requested c. Student comment + (2) a. Peer-revised draft the chair can decide to exclusively send the document to members of the Executive Board b. GPT-edited version the chair can decide to send the document exclusively to members of the Executive Board c. Student comment + (3) a. Peer-revised draft the overview of decisions is published by the secretariat of the Board of Governors on the intranet b. GPT-edited version the secretariat of the Board of Governors must publish an overview of decisions on the intranet c. Student comment + The next set of examples (4–6) illustrates students’ recognition of undesirable addition and omission as provided by ChatGPT: 199 Sonia Vandepitte (12) The translation aims to accurately convey the meaning of the original Dutch text. However, nuances in language and cultural context might not be fully captured. For more specific or legal interpretations, it is always recommended to consult with a legal professional familiar with Belgian law and university governance. […] “Bestuurders” has been translated as “governors” to reflect the role of these individuals in the university’s governance. “Raad van Bestuur” and “Bestuurscollege” have been translated as “Board of Governors” and “Executive Board” respectively, based on common English terms used in university governance. The phrase “maatschappelijk geëngageerde” has been translated as “socially engaged” to convey the university’s commitment to societal issues. (Google 2024) Nor did the course design include any evaluation stage of the students’ individual use of ChatGPT, which is an important stage of the student learning process. Our complex and ever-changing world will need translators to be able to translate texts themselves, performing all the traditionally required cognitive writing processes, as well as to post-edit a GenAI-tool. 6 Conclusion To conclude, integrating AI into the L2 translation classroom with a real-world task not only responds to the call of leaders at all levels of government to introduce AI into teaching in order to maintain global competitiveness. It also facilitates the road to reaching basic learning objectives, and, therefore, enriches L2 translation learning in a way that enhances reflective thought. The latter may in itself increase intrinsic motivation, a focal point in much L2 translation training. Considering the importance of the role of GenAI in education (Labadze et al. 2023), this exploratory project recommends the critical application of a GenAItool to be added as an additional learning objective in future L2 translation courses. It further shows teachers the direction in which the course design could be improved. In order to retrieve all the relevant information, students can be given a template for their assignments with space for their first draft of the translation, the monolingual and bilingual revision of that draft, a screenshot of students’ prompts showing the generative tool used, an AI chatbot output with students’ comments, and a final conclusion on the comparison between the revised 206 10 AI in an L2 translation class version and the AI chatbot output (including the number and type of changes accepted/ignored and whether they would use the tool again). Devoting course time to explore different translation-generative tools and to discuss their output is also recommended. While this qualitative inquiry has only focused on just one aspect of using AI in translation training, other teaching implications have been left undiscussed: ways in which a chatbot can help the teacher set up an examination text on a topic from texts seen in class or design a schedule for peer collaboration in large groups. However, as Gao writes: “GPT is not, and will not be, an adequate substitute for the language expertise, cultural awareness, ideological sensitivity, and creative abilities to ‘re-write’ on the part of professional human translators” (2024:1). Therefore, the training of a constructive while at the same time critical attitude towards GenAI is essential, perhaps more strongly than ever before. This should keep the standards of academic integrity and increase awareness of local and international ethical and legal issues. Today, courts have decided that AI systems do not have any rights, nor do they have a legal personality (which is reserved for natural and juridical persons). Nor will AI systems ever show true intelligence (Chomsky 2023). The main problem with them derives from the fact that they are solely based on a volume of data and algorithms that apply to that volume and were provided by often unidentifiable humans. In spite of GPT’s recently implemented measures for bias reduction and content moderation, that volume may still contain bias, toxic language or other harmful content. Better training and more reliable data sources will not completely stop an AI-tool from yielding hallucinations or responses that do not mirror true facts, nor will protection measures be foolproof against malicious actors. In other words, quality control will remain an ongoing effort that users need to be aware of. In order to build the necessary constructive critical attitude, various roads can be taken by translation trainers. They can provide for a creative skills module in the translation curriculum (Guerberof Arenas & Asimakoulas 2023), they can make their students aware of sustainability issues in the evaluation of the technology they use (Moorkens et al. 2024), and/or they can attribute human-centred augmented intelligence (O’Brien 2024) a central place in their own course. References Akhtar, Zarif Bin. 2024. Unveiling the evolution of generative AI (GAI): A comprehensive and investigative analysis toward LLM models (2021–2024) and 207 Sonia Vandepitte beyond. Journal of Electrical Systems and Information Technology. 22. DOI: 10. 1186/s43067-024-00145-1. Cao, Siyi & Linping Zhong. 2023. Exploring the effectiveness of ChatGPT-based feedback compared with teacher feedback and self-feedback: evidence from Chinese to English translation. https://arxiv.org/abs/2309.01645. Chomsky, Noam. 2023. The false promise of ChatGPT. The New York Times. https: //www.nytimes.com/2023/03/08/opinion/noam-chomsky-chatgpt-ai.html. Eloundou, Tyna, Sam Manning, Pamela Mishkin & Daniel Rock. 2023. GPTs are GPTs: an early look at the labor market impact potential of large language models. https://arxiv.org/abs/2303.10130. Felten, Ed, Manav Raj & Robert Seamans. 2023. How will language modelers like ChatGPT affect occupations and industries? https://arxiv.org/abs/2303.01157. Fitria, Tira Nur. 2023. Artificial intelligence (AI) technology in OpenAI ChatGPT application: A review of ChatGPT in writing English essays. ELT Forum: Journal of English Language Teaching 12(1). 44–58. DOI: 10.15294/elt.v12i1.64069. https://journal.unnes.ac.id/sju/elt/article/view/64069. Gao, Fei. 2024. When minds meet machines: A pilot programme of using ChatGPT to teach political speech translation in a Chinese-to-English translation classroom. International Journal of Chinese and English Translation & Interpreting 5. 11 pp. DOI: 10.56395/ijceti.v3i1.94. Geng, Xiang, Ming Zhu, Jiahuan Li, Zhejian Lai, Wei Zou, Shuaijie She, Jiaxin Guo, Xiaofeng Zhao, Yinglu Li, Yuang Li, Chang Su, Yanqing Zhao, Xinglin Lyu, Min Zhang, Jiajun Chen, Hao Yang & Shujian Huang. 2024. Why not transform chat large language models to non-English? https://arxiv.org/abs/ 2405.13923. Giampieri, Patrizia. 2024. AI and the BoLC: Streamlining legal translation. Comparative Legilinguistics 58. 68–90. https://pressto.amu.edu.pl/index.php/cl/ article/view/42397. Google. 2024. Gemini (aug 15 version). Large language model. Göpferich, Susanne. 2017. Cognitive functions of translation in L2 writing. In John W. Schwieter & Aline Ferreira (eds.), The handbook of translation and cognition, 402–422. Quebec, Canada: John Wiley & Sons. Grosman, Meta, Mira Kadric, Irena Kovačič & Mary Snell-Hornby (eds.). 2000. Translation into non-mother tongues in professional practice and training. Tübingen, Germany: Stauffenburg Verlag. Guerberof Arenas, Ana & Dimitri Asimakoulas. 2023. Creative skills development: Training translators to write in the era of AI. Hermes 63. 227–243. DOI: 10.7146/hjlcb.vi63.143078. 208 10 AI in an L2 translation class House, Juliane. 2018. The impact of English as a global lingua franca on intercultural communication. In Andy Curtis & Roland Sussex (eds.), Intercultural communication in Asia: Education, language and values, 97–114. Berlin, Germany: Springer. DOI: 10.1007/978-3-319-69995-0_6. Krings, Hans-Peter. 1986. Translation problems and translation strategies of advanced German learners of French (L2). In Juliane House & Shoshana BlumKulka (eds.), Interlingual and intercultural communication: Discourse and cognition in translation and second language acquisition studies, 263–276. Tübingen, Germany: Gunter Narr. Labadze, Lasha, Maya Grigolia & Lela Machaidze. 2023. Role of AI chatbots in education: Systematic literature review. International Journal of Educational Technology in Higher Education 20. 17 pp, 56. DOI: 10.1186/s41239-023-00426-1. Moorkens, Joss, Sheila Castilho, Federico Gaspari, Antonio Toral & Maja Popović. 2024. Proposal for a triple bottom line for translation automation and sustainability: An editorial position paper. Journal of Specialised Translation (41). 2– 25. DOI: 10.26034/cm.jostrans.2024.4706. Moorkens, Joss, Andy Way & Séamus Lankford. 2025. Automating translation. London, UK: Routledge. Mujadia, Vandan, Ashok Urlana, Yash Bhaskar, Penumalla Aditya Pavani, Kukkapalli Shravya, Parameswari Krishnamurthy & Dipti Misra Sharma. 2023. Assessing translation capabilities of large language models involving English and Indian languages. https://arxiv.org/abs/2311.09216. O’Brien, Sharon. 2024. Human-centered augmented translation: Against antagonistic dualisms. Perspectives 32(3). 391–406. DOI: 10.1080/0907676X.2023. 2247423. OpenAI. 2024. ChatGPT (Aug 15 version). Large language model. https://chat. openai.com/chat. OpenAI. N.d. Text generation models. https://platform.openai.com/docs/guides/ text-generation. Ouyang, Fan, Tuan Anh Dinh & Weiqi Xu. 2023. A systematic review of AIdriven educational assessment in STEM education. Journal for STEM Education Research 6. 408–426. DOI: 10.1007/s41979-023-00112-x. Robert, Isabelle S., Jim J. J. Ureel, Aline Remael & Ayla Rigouts Terryn. 2018. Conceptualizing translation revision competence: A pilot study on the ‘fairness and tolerance’ attitudinal component. Perspectives 26(1). 2–23. DOI: 10.1080/ 0907676X.2017.1330894. Robinson, Nathaniel R., Perez Ogayo, David R. Mortensen & Graham Neubig. 2023. ChatGPT MT: Competitive for high- (but not low-) resource languages. 209 Sonia Vandepitte In Proceedings of the Eighth Conference on Machine Translation (WMT), 392– 418. https://aclanthology.org/2023.wmt-1.40. Roumeliotis, Konstantinos I. & Nikolaos D. Tselikas. 2023. ChatGPT and OpenAI models: A preliminary review. Future Internet 15. 192. DOI: 10.3390/fi15060192. Sahari, Yousef, Abdu M. Talib Al-Kadi & Jamal Kaid Mohammed Ali. 2023. A cross sectional study of ChatGPT in translation: Magnitude of use, attitudes, and uncertainties. Journal of Psycholinguistic Research 52. 2937–2954. DOI: 10. 1007/s10936-023-10031-y. Teng, Shuhan. 2024. Exploring the effect of ChatGPT on the translation of poetry in literary works: A case study of the David Hawkes translation of A Dream of Red Mansions. Dean & Francis 1. 7. DOI: 10.61173/zgt1q485. Ülkü, Abdullah. 2023. ChatGPT-4 for hospitality: Implications. Journal of Tourism and Gastronomy Studies 11(3). 1727–1743. DOI: 10.21325/jotags.2023.1263. Vandepitte, Sonia. 2023. AI in a ‘translation into L2’ class. Presentation at the Autumn EMT Network Meeting. Salamanca, Spain. Vidakovic, Ivana. 2023. ChatGPT-prompts voor vertaling om de kracht van taal te ontsluiten. https://textcortex.com/nl/post/chatgpt-prompts-for-translation. Wang, Lulu, Simin Xu & Kanglong Liu. 2024. Understanding students’ acceptance of ChatGPT as a translation tool: A UTAUT model analysis. https://arxiv.org/ abs/2406.06254. 210