Abstract
The rapid advancement of neural machine translation (NMT) and large language models (LLMs) has transformed how students engage with second language (L2) learning, particularly in English-medium instruction (EMI) contexts. While machine translation (MT) tools can provide valuable support for students to process, produce, and learn their second language, the uncritical or excessive use of MT may hinder the language development, breach academic integrity, and result in over-reliance and cognitive disengagements. Against this background, this study investigated how Chines L1 students at EMI Sino-Foreign cooperative universities understand, perceive, and use MT. Based on this, the study further sought to conceptualize and empirically validate machine translation literacy (MTL) as a domain-specific, multi-dimensional form of digital literacy, and to quantitatively examine its relationship with students’ perceptions and usage patterns of MT.Adopting a sequential mixed-methods design, the study combined a large-scale survey (N=236) with follow-up focus group interviews. Based on theoretical frameworks on machine translation literacy (Bowker, 2019, 2020a, 2020b, 2021a, 2021b, 2022, 2023, 2024a, 2024b; Bowker & Ciro, 2019), a Machine Translation Literacy Scale was developed to measure students’ literacy of machine translation. Further, drawing on the Technology Acceptance Model (TAM), an additional scale was designed to measure students’ perceptions and usage of MT. Through exploratory and confirmatory factor analysis, MTL was validated as a multidimensional construct, comprising Understanding of Data-Driven Mechanisms (DDM), Training and Output Awareness (TRA), Task Awareness (TASK), Privacy and Data Safety (PRI), and Editing Skills (EDT). Hierarchical Regression Modeling further explored the relationships between machine translation literacy, students’ perceptions of MT, and MT usage patterns. Qualitative interviews complemented the quantitative findings by providing fine-grained insights into how different dimensions of MTL manifest in practice and influence student usage behavior.
The study reached five key findings. First, MTL was revealed as a multidimensional construct that integrates technical, cognitive, and ethical understandings of today’s AI-driven machine translation tools. Second, while students showed relatively strong literacy in Training and Output Awareness (TRA) and Task Awareness (TASK), their understandings remained insufficient in MT’s Data-Driven Mechanisms (DDM) and Privacy and Data Safety (PRI). Students’ Pre- and Post-Editing Skills (EDT) were moderately developed but often underutilised, as they tended to prioritize efficiency and task completion over deeper engagement with MT outputs. The imbalanced development of MTL dimensions highlighted the need for targeted pedagogical interventions to strengthen technical understanding and foster critical evaluation of MT output. Third, the five MTL dimensions were found to be interrelated and mutually reinforcing. Fourth, while MTL predicted positive Perceived Usefulness (PU) and Perceived Ease of Use (PEOU), these perceptions did not straightforwardly translate into greater adoption of MT. Instead, actual usage of MT was mediated by complex social and contextual factors, suggesting that the TAM framework alone is insufficient to capture the full dynamics of MT adoption in education. Fifth, as discovered in the interviews with students, the widespread lack of clear institutional and classroom policies on MT use left them uncertain about its legitimacy, leading to anxiety, guilt, or hesitation, which underscored the urgent need for coherent pedagogical and institutional support regarding the use of machine translation and other AI language technologies in the context of language learning.
| Date of Award | 19 Jul 2026 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Derek Irwin (Supervisor) & Xiuzhi Liu (Supervisor) |
Cite this
- Standard