Machine translation as a form of feedback on L2 writing

Abstract

With advances in artificial intelligence (AI), many language teachers have started exploring the classroom implications of AI-powered technology, including machine translation (MT). To examine the usefulness of MT technology in writing instruction, we conducted a mixed-methods study comparing two types of written feedback: comprehensive direct Teacher Corrective Feedback (TCF), and MT feedback. Participants were 23 Japanese university students in an intact L2 writing classroom. Sample size adequacy was confirmed through a priori power analysis. Participants were instructed to describe a picture prompt in L2 English and then in L1 Japanese. Half the participants received first TCF then MT on their L2 English text, while the order was reversed for the other half. Participants in both conditions were then asked to study the feedback and describe the same picture prompt without the feedback. In the following phase, both groups completed the same tasks in reverse order. Participants also responded to a survey exploring their engagement with the feedback. Results reveal that: 1) TCF improved complexity; 2) MT improved accuracy and fluency; and 3) variation in outcomes may be explained by the different ways in which participants engaged with both TCF and MT. Implications for appropriate classroom use of MT are discussed.