skip to main content
10.1145/3696952.3696985acmotherconferencesArticle/Chapter ViewFull TextPublication PagesiciipConference Proceedingsconference-collections
research-article
Open access

Profiling Chinese Student Interpreters’ Usage Intention of ChatGPT-assisted Translation by Q Methodology Based on Technology Acceptance Model

Published: 21 November 2024 Publication History

Abstract

This study investigates the Chinese student interpreters’ usage intention of ChatGPT-assisted translation with Q methodology. Guided by technology acceptance theory (TAM), the study aims to explore the usage intention profiles and consensus and differences between them. Q sort tasks were undertaken by 30 participants and their commentary data was elicited to complement Q sort analysis. From the factor analysis, four factors were obtained, namely deeper-purpose seekers, big picture reflectors, deep understanders and coordinators. The distinguishing statements include the subject, technology, information and society elements. The article concludes with a discussion of the pedagogical implications of learning interpreting of university students.

1 Introduction

The technology acceptance theory (TAM) believes that new technologies can be truly integrated into practice only when users are willing to accept and continue to use them (Davis et al., 1989). The individual acceptance of technology by using intention has been shown to be effective in predicting and explaining the actual use of new technologies (Venkatesh et al., 2003). The application of AI in the natural language processing field has resulted in the creation of intelligent chatbots and virtual assistants capable of both understanding and producing human language (Caldarini et al., 2022). The ChatGPT model has already seen increasing usage in applications such as translating languages (Chatterjee, 2023), which promotes the customization of teaching, reconstructs the relationship between teachers and students, and expands the field of translation studies (Zhou, 2023:134-146). In addition to the user-friendly development concept and intelligent information technology (IT), the reason for the population of ChatGPT is that it captures the user's information needs and provides good user experience.
At the same time, some people are pessimistic about ChatGPT, believing that its emergence will pose a serious challenge to traditional industries and social order (Debby, 2023; Chatterjee, 2023). In the education sector, for example, it has seriously disrupted the traditional order of teaching and learning. For instance, AI essay writing systems, which generate essays based on specific parameters or prompts, have disrupted conventional teaching and learning practices. This raises concerns about academic integrity, as students might misuse these systems to submit work that is not their own (Dehouche, 2021).
In response to the rise of ChatGPT-assisted translation, it is crucial for student interpreters to enhance their digital literacy and proficiency with technological tools, while also reinforcing their primary role as interpreters. Building on Dwivedi's (2023) exploration of the opportunities and challenges presented by ChatGPT, this study aims to examine the usage intention of ChatGPT-assisted translation of Chinese student interpreters. The researcher adopts a Q methodological design to holistically capture the learners’ profiles of usage intention.

2 Literature review

2.1 ChatGPT-assisted translation

ChatGPT represents a cutting-edge language translation system that leverages advanced natural language processing techniques to provide more accurate and faster translations compared to other systems. Built on the GPT-2 model, which has been extensively pre-trained on a vast corpus of text data, ChatGPT has undergone evaluations for various machine translation tasks, including translation prompts, multilingual translation, and translation robustness (Jiao et al., 2023). Experimental results indicate that ChatGPT, when using specifically designed translation prompts, can deliver performance that is on par with or superior to commercial translation systems for high-resource languages (Gao et al., 2023).
Compared with ChatGPT-assisted translation studies, its application in interpreting is at the exploratory stage. The ChatGPT technology with the characteristics of matching terminology and context, interactive language learning and two-way knowledge construction may change the data resources, technology iterations and scientific paradigms in interpreting. This has brought some additional technological benefits to the development of AI interpreting. A more obvious technological advantage in interpreting is demonstrated by the near transfer of the theme-related knowledge in the pre-interpreting preparation. Secondly, the ChatGPT application in interpreting can reduce the cognitive processing anxiety of interpreters to a certain extent. Finally, ChatGPT can significantly enhance the quality and efficiency of interpreting assessment management (Zheng et al.).

2.2 Usage Intention

Information systems researchers have proposed using intention models from social psychology as a theoretical foundation for studying the determinants of user behavior (Swanson, 1974). Among these models, Fishbein and Ajzen's (1975) Theory of Reasoned Action (TRA) stands out as particularly well-researched and effective in predicting and explaining behavior across various domains. TRA is highly general, designed to explain virtually any human behavior (Ajzen and Fishbein, 1980), making it suitable for examining the determinants of computer usage behavior specifically.
Davis (1986) introduced the Technology Acceptance Model (TAM), an adaptation of the Theory of Reasoned Action (TRA), specifically designed to explain computer usage behavior. TAM builds on TRA by establishing causal linkages between two key beliefs: perceived usefulness and perceived ease of use, and their influence on users’ attitudes, intentions, and actual computer adoption behavior. In China, researchers have begun applying TAM in educational contexts. Carey and Kacmar (2010) examined how culture and language affect technology acceptance and attitudes among Chinese students, revealing that interface features tailored to cultural preferences impact students’ perceived usefulness and usage intentions. Additionally, TAM has been applied to distance education via the Internet (Chen & Qiu, 2005) and to studying university teachers’ IT acceptance behavior in classroom teaching from an empirical perspective (Ren & Zhai, 2012).
With artificial intelligence technologies like ChatGPT becoming the future trend in IT development, it is essential to study the usage intentions and behaviors of ChatGPT users, particularly student interpreters. Drawing on the factors of technology-based self-directed foreign language learning proposed by Pan Xiaoquan (2021)—perceived usefulness, learning attitudes, ease of use, technological self-efficacy, and subjective norms—the following research questions were formulated:
What are the usage intention profiles of Chinese student interpreters'use of ChatGPT-assisted translation?
What are the consensus and differences between these intention profiles?

3 The study

This study examines the usage intentions of 30 Chinese student interpreters at various proficiency levels. The findings aim to provide educational stakeholders with insights to develop measures and strategies for the secure adoption, rejection, or controlled use of ChatGPT in interpreter education. Translation technology encompasses various technical methods applied to manual translation, machine translation, and computer-assisted translation.

3.1 Research context

When it comes to ChatGPT-assisted translation, people naturally associate it with machine translation. In China, The Teaching Guidelines for Undergraduate Translation Major released in 2020 lists “Translation Technology” as a compulsory course for the first time, which highlights its importance (Wang & Li, 2021). Generally speaking, Translation technology refers to the different types of technical means applied to human translation, machine translation and computer-assisted translation (Bowker, 2002). The translation function is only one of its by-products of ChatGPT.
Due to its translation function, ChatGPT can be preferred and used by student interpreters as a kind of translation technology. Although ChatGPT has the potential to assist instructors by generating course materials and providing suggestions, and to act as a virtual tutor for students by answering questions and facilitating collaboration, its use also presents challenges. These challenges include the generation of incorrect or fake information and the potential to bypass plagiarism detectors (Lo, 2023).
This study was conducted at the College of Foreign Languages and Literature of a prestigious comprehensive university located in a major southeastern city in China. The Department of Translation at this university offers bachelor's degrees in translation, as well as MTI and Ph.D. programs.

3.2 Participants

For the study, 30 students were recruited from Chinese-English interpreting courses or research groups. Most participants were native Mandarin speakers majoring in interpreting at the undergraduate, graduate, and Ph.D. levels.  In particular, 8 (27%) non-English undergraduate participants come from the optional interpreting course. According to the definition of Sawyer (2004: 72), the participants are at the entry and intermediate levels. Participant demographics are shown in Table 1.
Table 1:
ParticipantGenderAgeMajor*Received Training of ChatGPT (formal or informal)
Q1Female19Artificial IntelligenceBeginners' Guide
Q2Female44Translation StudiesBeginners' Guide; Application to disciplines; Application to translation
Q3Male22MTIBeginners' Guide
Q4Female18Medical JurisprudenceNone
Q5Male25Translation StudiesBeginners' Guide
Q6Female25Translation StudiesApplication to translation
Q7Male20Computer Science and TechnologyApplication to disciplines
Q8Female24MTIBeginners' Guide
Q9Male21Mathematics and Applied MathematicsBeginners' Guide;Application to disciplines
Q10Male20ChemistryNone
Q11Male23MTIBeginners' Guide
Q12Female19SociologyNone
Q13Male39Translation StudiesBeginners' Guide; Application to translation
Q14Male30Translation StudiesBeginners' Guide;Application to disciplines
Q15Female24MTIBeginners' Guide
Q16Female38Translation StudiesNone
Q17Male24MABeginners' Guide
Q18Male25MTINone
Q19Female23MTINone
Q20Female22TranslationBeginners' Guide
Q21Female35Translation StudiesBeginners' Guide; Application to disciplines; Application to translation
Q22Female18Economic ManagementApplication to disciplines
Q23Female23MTIBeginners' Guide
Q24Male22TranslationNone
Q25Female23TranslationBeginners' Guide
Q26Male22PhysicsBeginners' Guide; Application to disciplines
Q27Male18TranslationBeginners' Guide
Q28Male19TranslationNone
Q29Female18TranslationNone
Q30Female22TranslationBeginners' Guide
Table 1: Demographics
*Notes: English majors include Translation (undergraduate), MTI (Master of Translation and Interpreting), MA, and Translation Studies (Ph.D). The rest non-English majors come from the optional interpreting course.
Their classification information is shown in Table 2.
Table 2:
Classification of
Participants
Male
N.
Female
N.
Average
Age
Participant Number of Received training of
ChatGPT Use (multiple choices)
Beginners' GuideApplication to
disciplines
Application to
translation
Undergraduates from optional interpreting course4419.634 
Translation major (undergraduates)3420.64  
MTI, MA4423.56  
Ph.D3433.7434
Table 2: Classification Information
From Table 2, the researcher finds that the undergraduates of translation major receive the least training of ChatGPT use. And the reason for that lies in their preference to do translation by themselves rather than by machine translation, which is obtained from the later interview. Another significance of Table 2 is their high attention of training related to ChatGPT's guide and application, which indicates their acute perception of new technology.
Q methodology studies do not seek to generalize from a large population, similar to qualitative research (Watts & Stenner, 2012). A successful Q methodological study can involve a relatively small number of participants (typically 20-40) whose perspectives are crucial to the research topic (Watts, 2015). It means that in this PQ analysis, all the viewpoints of participants will be considered thoroughly and in keeping with this methodological holism.

3.3 Research design: Q methodology

This study employed Q methodology, developed by British physicist and psychologist William Stephenson in the 1930s (Stephenson, 1935). This method allows researchers to systematically uncover key subjective viewpoints among participants (Watts, 2015). It helps explore significant feelings and opinions held by individuals and highlights core attitudes or feelings shared within a group or community (Irie, 2014). Q methodology is similar to qualitative inquiries but includes statistical analysis, making it a mixed method on the continuum of quantitative and qualitative paradigms (Watts & Stenner, 2012; Watts, 2015). It has been successfully applied in political science, marketing research, policy studies, and education (e.g., Yang & Montgomery, 2013), and recently introduced to L2 motivation research (Irie & Ryan, 2015; MacIntyre et al., 2017). A typical Q methodological study involves five steps: (1) collecting related statements (concourse); (2) developing the sample statements (Q set), usually around 40 but ranging from less than 20 to over 250 (Millar, 2022); (3) gathering participants’ opinions (Q sort); (4) factor analysis of these Q sorts; and (5) interpreting the emergent factors in relation to the statements. The factors represent distinctive viewpoints shared within the participant group, grouping individuals according to their shared opinions on a certain issue. The participant sample size in a Q study usually ranges from 20 to 40, though effective studies can be conducted with fewer participants (Watts, 2015).

3.4 Data collection and data analysis

To identify the complexity of student interpreters’ usage intention of ChatGPT-assisted translation, the researcher subdivides the topic into ChatGPT use and acceptance of technology. A diverse range of sources, including academic literature (e.g., Zhang et al., 2023; Zheng et al., 2023; Dwivedi et al., 2023; Gordijn et al., 2023, etc.), popular science magazines, mainstream media platforms, and conversations with educators and peers, were utilized in an unstructured manner to collect various items encompassing different discourses on usage intention. Subsequently, two domain experts and two seasoned Q researchers were consulted to assess the statements and capture the nuanced perspectives on usage intention. The statements formed the preliminary concourse and were imported into online Q sorting platform1. Then, four of the participants were invited to do the pilot test, after which the final concourse was formulated from 40 to 36 statements by culling repeated, ambiguous or indirectly related ones according to their feedback. That's Step 1 Concourse and Step 2 Q set.
Step 3 Q sort. The researcher made appointments with the participants one by one on online Tencent Meeting (a reliable cloud video meeting product), whose screen sharing and recording functions facilitated the real-time watching, communicating and using playback for the researcher. Participants received detailed information about the research project and signed an electronic consent form. They then filled out a brief questionnaire to provide additional demographic information. Following this, participants engaged in a card-ranking activity involving 36 items on usage intention of ChatGPT-assisted translation. Each item was assigned a hierarchical position in a forced-choice, quasi-normal, and symmetrical distribution based on how well the statement was perceived to reflect the participant's understanding of the topic. After the ranking they would complete comments on 7 most agree, disagree or neutral statements. The face-validity dimension required items to be sorted from ‘most disagree’ to ‘most agree,’ with the most negative value (−5) on the left and the most positive value (+5) on the right. This dynamic process aimed to create a single, holistic configuration of all items, reflecting each participant's continuous comparison between them. The resulting product is viewed as a dynamic representation of their perspective on the subject (Lunberg, 2019). The sorting task for each participant took about 20-30 minutes to complete.
Figure 1:
Figure 1: Forced-choice, quasi-normal and symmetrical distribution continuum for 36-card sorts
Figure 2:
Figure 2: One of 30 sorting
For Step 4, the researcher used PQMethod2, a web-based platform that allows to manage Q methodology studies online, including inverted factor analysis. In brief, the analysis of the “Chinese Student Interpreters’ Usage Intention of ChatGPT-assisted Translation” Q sort began with an exploratory factor analysis adhering to standard statistical conventions to determine the identification and retention of factors (Rieber, 2020). A correlation matrix between and among Q sorts was generated using centroid extraction. After the centroid factor analysis was performed, 7 factors were seen as an output. And then a manual rotation of the factors was performed, thus Factor 1, Factor 2, Factor 3 and Factor 5 out of the former 7 factors were extracted. Therefore, the extracted 4 factors (eigenvalue > 1.00, according to Watts & Stenner, 2012) were performed as a varimax rotation, and 5 Varimax factors with high loading would be the output (due to the order of Factor 5 and the total number to be chosen). A significant loading value of 0.43 (p < .01) was calculated by the equation 2.58 × (1÷\(\sqrt {{\rm{No}}.{\rm{\ of\ items}}}\)) (here No. is 36), indicating that any Q sort with a single rotated factor loading above 0.43 could be taken as a member of this factor group, while confounding factors should be removed. After the successful completion of last routine run, a four-factor structure was identified (labeled as Factor 1, Factor 2, Factor 3 and Factor 4), explaining 63% of the total variance. Factor 1 included seven Q sorts, accounting for 21% of the variance, while Factor 2 comprised three Q sorts, explaining 9% of the variance, and Factor 3 of nine Q sorts with 17% of the variance and Factor 4 of seven Q sorts with 16% of the variance, shown in Table 3.
Table 3:
Profiles of Usage IntentionDistinctive featuresQ sortingUse frequency of ChatGPT-assisted translation
(0 for not mentioned)
%
expl. Var.
Factor 1:
deeper-purpose seekers
• preference to do translation by myself and then let ChatGPT help to revise it
• strong interest in the research on ChatGPT-assisted translation
• less sense of career crisis affected by ChatGPT
Q2, Q4, Q10, Q11, Q17, Q20, Q23Q2, Q10, Q11, Q23

1-2 times per week of using ChatGPT-assisted translation
21
Factor 2:
big picture reflectors
• belief in irreplaceability of ChatGPT to human interpreters
• confidence in the technology continuity of ChatGPT
• full trust in the person-to-person communication not intervened by chat model of ChatGPT
• AI products unable to cultivate students’ critical thinking and problem-solving abilities
Q14, Q18, Q2109
Factor 3:
deep understanders
• perceived usefulness and willingness to use
•strong belief in the replaceability of ChatGPT to human interpreters
• dependence on ChatGPT in translating
Q6, Q9, Q19Q6, 1-2 times per week of using ChatGPT-assisted translation
Q9, over 10 times
Q19, 3-5 times
17
Factor 4:
Coordinators
• concern about the interpreter's anxiety not reduced by use of ChatGPT
• deep concern about the impact of technological and social ethics
Q1, Q5, Q13, Q16, Q24, Q25, Q28Q5, Q13

1-2 times per week of using ChatGPT-assisted translation
16
Table 3: Q sorting groups
Notes: Due to the confounding factors, Q3, Q7, Q8, Q12, Q15, Q22, Q26, Q27, Q29, Q30 were excluded from any group.

4 Findings

The researcher analyzed the four factors by closely examining the individual rankings and the overall configuration of the Q statements within the relevant factor array, as well as the statements ranging from “least like” to “most like” (Zheng, 2023:6). The following section provides a qualitative interpretation of four usage intentions for ChatGPT-assisted translation among student interpreters. And the distinguishing statements of each intention profiles are summarized in Table 3.

4.1 Usage intention profiles of Chinese student interpreters

4.1.1 Factor 1: deeper-purpose seekers

The seven sorts that loaded on Factor 1 can be best characterized as deeper-purpose seekers. Table 4 summarizes the distinguishing statements, characterizing the dominant usage intention profile of this group.
Table 4:
Distinguishing statements in Factor 1Q-Sort ValueZ-Score
Agreed21. In order to improve my translation skills, I prefer to do translation by myself first and then let ChatGPT help to revise it.31.28
 22. I am interested in the application research of ChatGPT-assisted interpreting.31.15
Disagreed1. ChatGPT may affect my future career and I feel a career crisis.-2-0.97*
Table 4: Sorting results of Factor 1
Note: Asterisk (*) Indicates Significance at P < .01).
The seven Q sorts loading on Factor 1 share specific purpose — the application research of ChatGPT-assisted interpreting (statement no. 22) and know the route to seeking it— practice translation competence (21). They have less career crisis of being translator (1). The reason for deeper-purpose seeking may lie in their feel of career crisis. The probability of automation for translation and interpreting was estimated at 38%, categorizing these professions as “medium-risk” occupations (Vieira, 2020). Commentary data from Q20, a participant with a defining loading of .6408 on Factor 1, supports this notion: “I think the process of translation should be translator-oriented with artificial intelligence such as ChatGPT as assistance. And ChatGPT cannot replace the subjective choice and judgement of the translator. The academic research should follow the technology trend and explore the possibility of its assistance to translation. (Comment on Statement 21 from Q20)”.
Related to their use frequency of ChatGPT-assisted translation–4 out of 7 Q sorts at 1-2 times per week (shown in Table 3), this comment illustrates that student interpreters do not fear being ourperformed by ChatGPT, but rather restrained by the technology's limitations and market acceptance (Vieira, 2020). They have clear purpose. Thus, the researcher characterize this group as deeper-purpose seeker.

4.1.2 Factor 2: big picture reflectors

The three sorts that loaded onto Factor 2 have the largest distinguishing statements shown in Table 5. Factor 2 shares similar strong confidence with Factor 1 in the irreplaceability of the human interpreters by ChatGPT. But they contain opposite Z-scores with Factor 1 on Statement 21, and 22, which shows that this group would not do translation by themselves first, nor is interested in the application research of ChatGPT-assisted interpreting.
Table 5:
Distinguishing statements in Factor 2Q-Sort ValueZ-Score
Agreed2. No matter how ChatGPT technology advances, it will not replace human interpreters.41.88*
 11. AI products such as ChatGPT provide students with quick and concise answers but do not help students develop critical thinking and problem-solving skills.31.07
 4. ChatGPT is pre-trained by large data and there are data security risks.20.76
 3. ChatGPT has been popular one night and may be as short-lived as any other technology.10.37*
 1. ChatGPT may affect my future career and I feel a career crisis.00.26*
Disagreed15. Due to the risk of ChatGPT, I am not willing to continue to use it, nor would I recommend it to others.-1-0.42
 7. Some low-end, preliminary translators may indeed be replaced by ChatGPT.-1-0.45*
 22. I am interested in the application research of ChatGPT-assisted interpreting.-2-0.61
 16. ChatGPT is far away from me and it doesn't matter whether I use it or not.-2-0.68
 31. ChatGPT was developed in the West, and the information output is likely to embody westernized ideas.-3-1.24
 21. In order to improve my translation skills, I prefer to do translation by myself first and then let ChatGPT help to revise it.-4-1.57
 20. I am concerned that the human-ChatGPT dialogue will reduce person-to-person communication.-5-2.11*
Table 5: Sorting results of Factor 2
Note: Asterisk (*) Indicates Significance at P < .01).
The participants grouped under Factor 2 concern more on the drawbacks of ChatGPT, such as vulnerability to develop students’ critical thinking and problem-solving skills (11), data security risks (2). Though realizing the drawbacks, this group will use it (15, 16), and be confident in its authenticity of the information output (31), and its noninterference to person-to-person communication (20).
In particular, three participants in this group have not experienced the ChatGPT-assisted translation (shown in Table 3) and probably their comments come from indirect perspective. Comment data from Q 14 (with a defining loading of .5311 on Factor 2) and Q18 (with a defining loading of .5270 on Factor 2) exemplifies this intention profile: “The algorithm on which ChatGPT is based is different from the human cognitive system. Unlike a computer program, the flexible human cognition processes are not computable. The creativity of human language makes language evolve while machine learning is always lagging. There are over 7,000 human languages, and ChatGPT cannot handle languages spoken by the minority. (Comment on Statement 7 from Q14)”.
The intention profile here can be described as big picture reflectors, which involves harmonizing the capabilities of technology with one's personal philosophy of ChatGPT usage. Participants grouped in Factor 2 still have confidence in the irreplaceability of the human interpreters by ChatGPT though they will not do translation by themselves first nor research of ChatGPT.

4.1.3 Factor 3: deeper understanders

The three sorts that loaded on Factor 3 can be characterized as deeper understanders with statements shown in Table 6.
They definitely express the acceptance intention of ChatGPT-assisted translation (14) and are concerned about the replaceability of human interpreters by ChatGPT (2, 7) . On the other hand they notice the threat to human society and education resulted by over-reliance or blind worship (6).
Table 6:
Distinguishing statements in Factor 3Q-Sort ValueZ-Score
Agreed14. I think ChatGPT functions well and I would like to continue using it.41.41
 7. Some low-end, preliminary translators may indeed be replaced by ChatGPT.20.70*
Disagreed4. ChatGPT is pre-trained by large data and there are data security risks.-1-0.26*
 6. The over-reliance on AI, or even blind worship or superstition, are the greatest threat to human society and education.-1-0.61*
 21. In order to improve my translation skills, I prefer to do translation by myself first and then let ChatGPT help to revise it.-2-0.69
 2. No matter how ChatGPT technology advances, it will not replace human interpreters.-3-1.54*
Table 6: Sorting results of Factor 3
Note: Asterisk (*) Indicates Significance at P < .01).
Participant Q9 (with a defining loading of .7484 on Factor 3) and Participant Q6 (with a defining loading of .5507 on Factor 3) shared their comments. “In the age of AI, it is necessary to understand how AI works and how to empower translation work with GPT to avoid being eliminated from society. (Comment on Statement 6 from Q6); In the fields of translation, quantitative research, big data and finance, the development of AI such as ChatGPT will reduce the need for technical positions. (Comment on Statement 2 from Q9)”.
In fact, the three sorts in this group all have experience of ChatGPT-assisted translation, and their relatively high use frequency per week is Q6 at 1-2 times, Q9 over 10 times and Q19 at 3-5 times as shown in Table 3. The ongoing interest in interpreting technologies suggests that these tools are likely to become a permanent fixture in the work of interpreters (cf. Kalina & Ziegler, 2015). The intention profile here can be described as deeper understanders, which means they trust the performance and security of ChatGPT and will apply it into translation. They even think it will replace the human interpreters.

4.1.4 Factor 4: Coordinators

The remaining seven sorts that loaded on Factor 4 can be characterized as Coordinators which means those people who can solve the conflict successfully. Table 7 summarizes the distinguishing statements.
The participants who loaded on Factor 4 notice the influence of ChatGPT to technological and social ethics (33) and the existing interpreters' cognitive processing anxiety (12). Their reaction was to enhance the translation competence by more practice (21).
Table 7:
Distinguishing statements in Factor 4Q-Sort ValueZ-Score
Agreed33. The emergence of ChatGPT will affect the existing system of technological and social ethics.30.90
 21. In order to improve my translation skills, I prefer to do translation by myself first and then let ChatGPT help to revise it.30.64
Disagreed12. ChatGPT technology can reduce interpreters' cognitive processing anxiety.-2-1.15*
Table 7: Sorting results of Factor 4
Note: Asterisk (*) Indicates Significance at P < .01).
The narratives from Q1 (with a loading of .4803 of Factor 4) and Q16 (with a loading of .6701) shed some light: “Cases show that ChatGPT can now perform tasks traditionally undertaken by humans, which will lead to unemployment to some extent. Next is the “black box” Technology problem. Although ChatGPT as a language model is unlikely to lead to potential malicious incidents, it is necessary to formulate related laws or regulations. (Comment on Statement 33 from Q1); ChatGPT may be able to translate formulated sentences, but it is difficult to translate language with specific cultural imagery or poets, which are also the difficulties that interpreter come across. So ChatGPT will not reduce the interpreter's cognitive anxiety. (Comment on Statement 12 from Q16)”.
The commentary focuses on the technological ethics and interpreter's anxiety. Related to their use frequency of ChatGPT-assisted translation — 2 out of 7 use it 1-2 times per week (shown in Table 3). It can be seen that the participants in this group prefer to do translation by themselves to relieve their anxiety and concern with the technological ethics. There is an undeniable need for further research on how technology enables, mediates, and constrains the practice of interpreting (Mellinger, 2019). That is Factor 4 Coordinators.

4.2 Consensus

The consensus statements mean those that do not distinguish between ANY pair of factors.
As to the channel of knowing ChatGPT, all agreed on the information pushed from social media (35). They have understood its technological advantage (5) and found the differences between ChatGPT and other search engines (30). In a word, they feel happy, interested and are willing to try ChatGPT-assisted translation (8, 9, and 18), which embodies their IT literacy (19). Students’ internal positive acceptance plays a crucial role in promoting their usage intention of ChatGPT-assisted translation (Man, 2023).

5 Discussion

In an effort to advance research on Chinese student interpreters’ usage intention of ChatGPT-assisted translation, the present study profiled four groups: deeper-purpose seekers, big picture reflectors, deep understanders and conciliators. These four profiles largely fit into Rieber (2020: 2540) categorization of deeper understanders, big picture reflectors, deeper-purpose seekers and teaching/ technology Schismists, while the last one was revised to coordinators. Factor 1 is characterized with less sense of career crisis and research enthusiasm of the application of ChatGPT in interpreting compared with much sense of career crisis and no research enthusiasm of it. Factor 3 believes in the replaceability of human interpreters by ChatGPT and trusts its data security, which is totally opposite to that of Factor 2. The following section aims to explain these observations with reference to the ChatGPT empowerment to interpreting learners.

5.1 Integration of interpreting and ChatGPT

ChatGPT is widely utilized for various NLP tasks, including text generation, language translation, and answering numerous questions (Dwivedi, 2023). As an AI tool, it represents the integration of interpreting with technology, similar to the previously mentioned “training in the use of information technology for subject preparation before, during, and after assignments.” (Gile, 1995).
As to the usage of ChatGPT in interpreting, the researcher gets the data from the demographic questionnaires answered before Q setting. 70% of the participants would like to use ChatGPT in per-interpreting preparation, 20% during interpreting and 20% for post-interpreting assessment. Therefore, it can be concluded that most student interpreters use ChatGPT as a search engine at present. Furthermore, all participants share virtually unanimous attitude towards the usage effect shown in Table 8. Although translation quality is frequently imperfect, it generally allows users to access content that would otherwise be unavailable to them. (Carl, 2018).
Table 8:
StatementsFactor
1234
28. The quality of ChatGPT translation is not so different from that of other translation engines.-0.81
27
-1.29
32
-0.24
21
-0.57
25
30. ChatGPT does not differ much from other search engines such as Baidu and Google when doing background knowledge search for interpreters.-0.85
28
-1.35
33
-0.61
28
-0.70
26
Table 8: Factor scores with corresponding ranks
As for the integration of interpreting and ChatGPT, it needs a long way to go due to the essence of interpreting— from speech recognition, code transfer to expression, which requires equipment and connection.

5.2 Caution of the technology risk

Concerning the weakness of ChatGPT, Farrokhnia et al. (2023) listed “insufficient depth of understanding,” “challenges in assessing response quality,” “potential for biases and discrimination,” “deficiency in high-order thinking skills,” “risks to educational integrity,” “inadequate contextual comprehension,” “threats to academic honesty,” “reinforcement of educational discrimination,” “facilitation of plagiarism in education and research,” and “decline in high-order cognitive skills.”
ChatGPT is an AI technology that is based on large amounts of data. The flow of the data inevitably involves data security, especially in the complex international environment. Data security matters to national security. It is necessary to strengthen the management of data and raise awareness of data security of the masses before ChatGPT is open sourced on a large scale. In addition, the core technology of ChatGPT is developed by OpenAI, a foreign company, which will result in the technical security. How to effectively avoid the problem of technology containment in the future is also one of the issues to be considered.
A potential risk is the impact of ChatGPT on the efficiency of knowledge-based tasks. As a generative tool, ChatGPT can produce new data rather than merely analyzing existing information. Consequently, the absence of notable enhancements in the productivity of knowledge work may lead to frustration among leaders and policymakers.

6 Conclusion

This research examined the intention of Chinese student interpreters to use ChatGPT-assisted translation at the College of Foreign Languages and Literature in a prominent key comprehensive university situated in a southeastern metropolitan city in China. The 30 participants surveyed were students of Chinese-English interpreting courses or members of research groups, with the majority having Mandarin Chinese as their first language and interpreting as their major from undergraduates, graduates and Ph.D. programs. In particular, 8 (27%) non-English undergraduate participants come from the optional interpreting course.
With Q methodology, the participants are sorted into four factors/ groups according to their different usage intention: Factor 1, deeper-purpose seekers, are with strong interest in the research on ChatGPT-assisted translation and preference to do translation by themselves first and less sense of career crisis; Factor 2, big picture reflectors are featured by belief in irreplaceability of ChatGPT and confidence in the technology continuity of ChatGPT; Factor 3, deep understanders are characterized by strong belief in replaceability of ChatGPT to human interpreters, disagreement on the threat of ChatGPT to the society and education and suspicion of data security; and Factor 4, coordinators, have deep concern about the impact of technological and social ethics, and about the interpreter's anxiety not reduced by use of ChatGPT. Indeed, participants appeared to form their intentions toward using ChatGPT-assisted translation primarily driven by their expectations that it would enhance their performance in their respective academic achievement, which accords with the user acceptance theory.
In brief, they feel happy, interested and are willing to try ChatGPT language translation, which embodies their IT literacy. However, progressing in the profession typically requires greater mastery of technology, not less. Translators can only hope to control their work when they are critically aware of the tools at their disposal (Gil & Pym, 2006). Given the challenges of ChatGPT-assisted translation, student translators and interpreters should be trained to work with it.
Pedagogical implications from this study can be applied to the teaching and learning of interpreting within the context of ChatGPT. As shown in this study, the majority of the students received little or just the beginners' guide by themselves. Teachers may instruct ChatGPT's application to respective disciplines or application to translation. Curriculum designers may consider opening new related courses to broaden students’ horizon. At research institutes, concerned efforts are expected to develop the practicality of ChatGPT-assisted interpreting.
It is acknowledged that the Q methodology used in this study has not been widely applied in intention research or translation studies in general. Although efforts were made to expand the range of statements regarding student interpreters’ intentions to use ChatGPT-assisted translation, further refinement could yield more insightful results. Nonetheless, this study demonstrates the potential of using Q methodology as a mixed method to comprehensively capture learners’ usage intentions within a specific context. Future research on intention could benefit from this approach.

Acknowledgments

I would like to extend my gratitude to Yizhu LI for providing a step-by-step guide on the Q methodology used in this article. I also thank Yongyan ZHENG and Xiuchuan LU for their discussions and comments on my study, as well as the anonymous reviewers for their invaluable feedback.

Footnotes

1
Q sorting platform (Q-Tip: Q-Method Testing and Inquiry Platform), was jointly developed by University of Wisconsin-Madison, Indiana University and University of Guelph, https://qtip.geography.wisc.edu/#/.

References

[1]
Ajzen I, (1975). Fishbein M. Belief, attitude, intention and behavior: an introduction to theory and research. Addsion-Wesley, Reading MA.
[2]
Ajzen, I. (1980). Understanding attitudes and predicting social behavior. Englewood cliffs.
[3]
Bowker, L. (2002). Computer-aided translation technology: A practical introduction. University of Ottawa Press.
[4]
Brown, S. R. (1980). Political subjectivity. New Haven, CT: Yale University Press.
[5]
Caldarini, G., Jaf, S., & McGarry, K. (2022). A literature survey of recent advances in chatbots. Information, 13(1), 41.
[6]
Carey, J. M., & Kacmar, C. J. (2010). Cultural and language affects on technology acceptance and attitude: Chinese perspectives. International Journal of Information Technology, 16(1), 1-19.
[7]
Carl, M. & Braun, S. (2018). Translation, interpreting and new technologies. In K. Malmkjaer (ed), The Routledge Handbook of Translation Studies and Linguistics. London: Routledge, 374-390.
[8]
Chatterjee, J., & Dethlefs, N. (2023). This new conversational AI model can be your friend, philosopher, and guide, and even your worst enemy. Patterns, 4(1), 100676.
[9]
Chen, Y. & Qiu M. M. Application of TAM in distance education based on Internet. The Chinese Journal of ICT in Education. 2005 (2), 72-73.
[10]
Cotton, D. R., Cotton, P. A., & Shipway, J. R. (2023). Chatting and cheating: Ensuring academic integrity in the era of ChatGPT, Innovations in Education and Teaching International.
[11]
Davis, F. D. (1986). A technology acceptance model for empirically testing new end-user information systems: Theory and results (Doctoral dissertation, Massachusetts Institute of Technology).
[12]
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: A comparison of two theoretical models. Management science, 35(8), 982-1003.
[13]
Dehouche, N. (2021). Plagiarism in the age of massive Generative Pre-trained Transformers (GPT-3). Ethics in Science and Environmental Politics, 21, 17-23.
[14]
Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., & Wright, R. (2023). “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management, 71, 102642.
[15]
Gao, Y., Wang, R., & Hou, F. (2023). How to Design Translation Prompts for ChatGPT: An Empirical Study. arXiv e-prints, arXiv-2304.
[16]
Gil, José Ramón Biau, and Anthony Pym. "Technology and translation (a pedagogical overview)." Translation Technology and its Teaching, Intercultural Studies Group, Universitat Rovira i Virgili, Tarragona (2006).
[17]
Gile, Daniel (1995). Basic Concepts and Models for Interpreter and Translator Training. Amsterdam and Philadelphia: John Benjamins.
[18]
Hoffman, R. R. (1997). The cognitive psychology of expertise and the domain of interpreting. Interpreting, 2(1-2), 189-230.
[19]
Ire, K. (2014). Q methodology for post-social-turn research in SLA. Studies in Second Language Learning and Teaching, 4(1), 13-32.
[20]
Irie, K., & Ryan, S. (2015). Study abroad and the dynamics of change in learner L2 self-concept. Motivational dynamics in language learning, 343-366.
[21]
Jiao, W., Wang, W., Huang, J. T., Wang, X., & Tu, Z. P. (2023). Is ChatGPT a good translator? Yes with GPT-4 as the engine. arXiv preprint, arXiv:2301.08745. https://arxiv.org/abs/2301.08745
[22]
Kalina, S.; Ziegler, K. (2015). Technology, in: F. Pöchhacker (ed). Routledge Encyclopedia of Interpreting Studies. New York: Routledge, pp. 410–412.
[23]
Kang, Z.F. (2024) New orientations of national interpreting education in the ChatGPT era. Translation Research and Teaching,(01):1-9.
[24]
Ling, X.X., Wang D. M. & Yuan J. (2023). Reflection on Technology Ethics and Academic Ethics in the Context of ChatGPT Craze. Journal of Xinjiang University (Philosophy and Social Sciences). 44(04), 123-136.
[25]
Liu, Y., Han, T., Ma, S., Zhang, J., Yang, Y., Tian, J., & Ge, B. (2023). Summary of ChatGPT/ GPT-4 research and perspective towards the future of large language models. arXiv preprint arXiv:2304.01852.
[26]
Lo, C. K. (2023). What is the impact of ChatGPT on education? A rapid review of the literature. Education Sciences, 13(4), 410.
[27]
Lu, X.C., Zheng, Y.Y. (2019) On the dual foreign language learning motivation of English majors in Chinese universities, Fudan Forum on Foreign Languages and literature, (01):28-36.
[28]
Lundberg, A. (2019). Teachers’ beliefs about multilingualism: Findings from Q method research. Current Issues in Language Planning, 20(3), 266-283.
[29]
Lundberg, A., de Leeuw, R., & Aliani, R. (2020). Using Q methodology: Sorting out subjectivity in educational research. Educational Research Review, 31, 100361.
[30]
MacIntyre, P. D., MacKay, E., Ross, J., Abel, E., Ortega, L., & Han, Z. (2017). The emerging need for methods appropriate to study dynamic systems. Complexity theory and language development: In celebration of Diane Larsen-Freeman, 97-122.
[31]
Man, D. L. & Mo J. (2023) An empirical study of MTI students’ intention to use translation technology. Foreign Language and Translation (01), 86-91.
[32]
Mellinger, C. D. (2019). Computer-assisted interpreting technologies and interpreter cognition: A product and process-oriented perspective. Tradumàtica, (17), 0033-44.
[33]
Millar, J. D., Mason, H., & Kidd, L. (2022). What is Q methodology?. Evidence-Based Nursing, 25(3), 77-78.
[34]
Mu, L. (2022). Translation teaching research in China. Shanghai Foreign Language Education Press.
[35]
Pan, X.Q. (2021) On the college English teaching: the perspective of hybrid space. Guangming Daily Publishing House.
[36]
Ren, X. H. and Zhai, N. (2012). An empirical research of information technology acceptance behaviors of teachers in college classroom teaching. Journal of Distance Education, 2012, 30(02): 84-90.
[37]
Rieber, L. P. (2020). Q methodology in learning, design, and technology: An introduction. Educational Technology Research and Development, 68(5), 2529-2549.
[38]
Sawyer, D. B. (2004). Fundamental aspects of interpreter education. Fundamental Aspects of Interpreter Education, 1-328.
[39]
Schäler, R., Way, A., & Carl, M. (2003). EBMT in a Controlled Environment. Recent advances in example-based machine translation, 83-114.
[40]
Swanson, E. B. (1974). Management information systems: appreciation and involvement. Management science, 21(2), 178-188.
[41]
Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS quarterly, 425-478.
[42]
Vieira, L. N. (2020). Automation anxiety and translators. Translation Studies, 13(1), 1-21.
[43]
Wang, H.S. & Li Y. (2021). Translation technology teaching in the New Era: problems and countermeasures, (03):13-21.
[44]
Watts, S. & Stenner, P. (2012). Doing Q methodological research: Theory, method & interpretation. Doing Q Methodological Research, 1-248.
[45]
Watts, S. (2015). Develop a Q methodological study. Education for Primary Care, 26(6), 435-437.
[46]
Yang, Y., & Montgomery, D. (2013). Gaps or bridges in multicultural teacher education: A Q study of attitudes toward student diversity. Teaching and Teacher Education, 30, 27-37.
[47]
Zhan C. (2010). Thirty years of interpreting teaching in China: development and current situation. Journal of Guangdong University of Foreign Studies, 21(6), 89-92.
[48]
Zhang H., Liu C, Wang D. B. & Zhao Z. X. (2023) Research on the influencing factors of ChatGPT users’ intention. Information studies: Theory & Application, (04),15-22.
[49]
Zheng, Y. H., Zhang M. C., Wang X. N., Zhang B. J. Cheng L. L. & Rao G. Q. (2023). Multi-domain Changes and Challenges Caused by ChatGPT. Journal of Tianjin Normal University (Social Sciences), (3): 49-63.
[50]
Zheng, Y., Lu, X., & Ren, W. (2019). Profiling Chinese university students’ motivation to learn multiple languages. Journal of Multilingual and Multicultural Development, 40(7), 590-604.
[51]
Zheng, Y., Lu, X., & Ren, W. (2020). Tracking the evolution of Chinese learners’ multilingual motivation through a longitudinal Q methodology. The Modern Language Journal, 104(4), 781-803.
[52]
Zheng. (2023) Q methodology within a transdisciplinary shift in applied linguistics: innovation and future directions. Journal of Foreign Languages, 46(01):2-10.

Index Terms

  1. Profiling Chinese Student Interpreters’ Usage Intention of ChatGPT-assisted Translation by Q Methodology Based on Technology Acceptance Model
            Index terms have been assigned to the content through auto-classification.

            Recommendations

            Comments

            Information & Contributors

            Information

            Published In

            cover image ACM Other conferences
            ICIIP '24: Proceedings of the 2024 9th International Conference on Intelligent Information Processing
            November 2024
            419 pages
            ISBN:9798400718076
            DOI:10.1145/3696952

            Publisher

            Association for Computing Machinery

            New York, NY, United States

            Publication History

            Published: 21 November 2024

            Check for updates

            Author Tags

            1. ChatGPT-assisted translation
            2. Chinese student interpreters
            3. Q methodology
            4. usage intention profiles

            Qualifiers

            • Research-article

            Conference

            ICIIP 2024

            Acceptance Rates

            Overall Acceptance Rate 87 of 367 submissions, 24%

            Contributors

            Other Metrics

            Bibliometrics & Citations

            Bibliometrics

            Article Metrics

            • 0
              Total Citations
            • 115
              Total Downloads
            • Downloads (Last 12 months)115
            • Downloads (Last 6 weeks)115
            Reflects downloads up to 21 Dec 2024

            Other Metrics

            Citations

            View Options

            View options

            PDF

            View or Download as a PDF file.

            PDF

            eReader

            View online with eReader.

            eReader

            HTML Format

            View this article in HTML Format.

            HTML Format

            Login options

            Media

            Figures

            Other

            Tables

            Share

            Share

            Share this Publication link

            Share on social media