Adapt or adopt? Examining the efficacy of ChatGPT in providing translation feedback

Abstract

Translation feedback is crucial for improving student translation quality and cultivating translation competence, yet it remains a labour-intensive and time-consuming task for teachers. This study explores the potential of ChatGPT, a generative artificial intelligence (GenAI) model, as an alternative or supplement to traditional teacher feedback in translation education by comparing the content and strategy of these two feedback sources, as well as students’ adoption and perception of these sources. A mixed-methods crossover experimental design was employed to assess feedback from ChatGPT and teachers across four translation tasks. The findings indicated that ChatGPT tended to offer direct reference translations and implicit general feedback, focusing predominantly on lexical and morphosyntactic issues. In contrast, teacher feedback was more indirect and elaborated. Students showed a higher rate of acceptance with teacher feedback than with ChatGPT feedback. Interviews revealed that students held positive views towards teacher feedback but expressed mixed attitudes towards ChatGPT feedback regarding its quality, reliability, comprehensibility, and affective connection. The findings contribute to the discourse on the transformative role of GenAI in translation education and offer insights for integrating ChatGPT in translation teaching practice.