This study examines the influence of text genre, i.e. literary and tourism texts, and prompting techniques, specifically Chain-of-Thought (CoT) reasoning and Domain-Specific Prompts (DSP), on trainee translators' reading processes when evaluating GenAI-generated translations. Drawing on eye-tracking and post-task cued retrospection data from 30 trainees, the study found that (a) GenAI-generated translations of literary texts required longer and more frequent fixations than those of tourism texts during evaluation, indicating intricate genre-specific reading behaviours; (b) the use of DSP, when combined with CoT reasoning, reduced cognitive effort in translation quality evaluation than the CoT-only case, as evidenced by shorter fixation durations on the rating scale; (c) trainees demonstrated a general serial genre-independent reading pattern that begins with the source text, followed by the shifts between the source and target texts, and concludes with the rating scale. These findings offer insights into the cognitive mechanisms underlying reading for post-editing purposes across diverse genres in a GenAI era, further highlighting the necessity of integrating metacognitive awareness and tailored prompting strategies into translator training.
The rapid advancement of Generative Artificial Intelligence (AI) has shifted interpreting pedagogy towards a human-AI synergy, yet it has introduced unprecedented cognitive challenges, such as cognitive anchoring and correction overhead. Traditional instructional feedback often fails to address interpreting as a complex ill-structured problem. In response to this gap, we propose the Dynamic Synergy Model of Interpreting Coaching (DSMIC) as a cognitive stabiliser. The framework operationalises coaching through two synergistic sub-models: Interpreting Problem-Solving Coaching (IPSC) for tactical cognitive intervention and Student-Centered Interpreting Learning Process Coaching (SCILPC) for affective regulation. Through a six-week microgenetic case study of two novice interpreters, the research suggests that this heuristic approach effectively decouples learners from technological dependency. Findings, triangulated via the Interpreting Strategy and Resilience Grid (ISRG) and qualitative protocols, reveal a significant shift from passive algorithm followers to proactive cognitive commanders, evidenced by enhanced key information retention, strategic flexibility, and rapid cognitive recovery. Ultimately, the study reconceptualises interpreting competence as metacognitive orchestration, arguing that the interpreters' thinking premium - the capacity for adaptive expertise under uncertainty - is the core value that must be actively reconstructed through expert coaching in the AI era.
Translation and interpreting (T&I) have long been taught and assessed in higher education contexts worldwide. With the growing demand for grading student-generated T&I, researchers are focusing on automatic T&I assessment. In this article, we aim to initially investigate the utility of ChatGPT as an automatic evaluator for interpreting, by comparing ChatGPT-based scores with those of human raters. We recruited 45 raters from diverse backgrounds to assess 56 English-Chinese student interpretations. Twenty-four ChatGPT-based e-raters were also configured to assess the same interpretations, varying the reference rendition availability, scoring granularity, and model randomness. We conducted statisticalanalysis to examine the reliability, severity, validity, and accuracy of ChatGPT-based e-raters. Our findings reveal that: a) e-raters showed greater reliability and severity than human raters; b) e-raters generally matched human raters in concurrent validity; and c) e-raters outperformed some of their human counterparts in scoring accuracy for English-to-Chinese interpretations. Among the 24 e-raters, those scoring at the passage level exhibited lower reliability, severity and validity but, interestingly, achieved more accurate scores for English-to-Chinese interpretations. These results suggest that ChatGPT-based e-raters may be used as a secondary rater for cross-validation alongside human raters but should not be deployed as sole scorers.
Following in the footsteps of academic proponents of collaboration between spoken and signed language interpreters (SpLIs and SLIs), the authors of this study have bridged their respective areas of expertise to bring these two communities of practice together. This study presents data from a two-phase empirical investigation into the attitudes - understood as beliefs and perceptions - of SpLIs and SLIs towards the principles set out in various codes of ethics. The research design comprised an online questionnaire and three focus groups. The analysis was conducted from diachronic and synchronic perspectives, taking into account language modalities of the participants, as well as their professional experience. The findings confirmed substantial similarities and differences between SpLIs and (deaf and hearing) SLIs in defining and applying professional ethics. As a result, the authors advocate for close collaboration between the two groups of interpreters, which would enhance exchange of knowledge and experience and thus prepare them for the developments of the ever-evolving profession.
