Timed second language writing performance: effects of perceived teacher vs perceived automated feedback

Automated writing evaluation (AWE) software has received much attention by researchers and practitioners as it has the potential of reducing the time necessary for providing second language (L2) students with written corrective feedback on their writing. While more practitioners are taking advantage of AWE affordances, research has indicated students engaged in untimed writing may still possess reservations about its usefulness as a feedback provider compared to that of L2 writing teachers. As trust affects student writer behavior, it is important to understand how the potential relationship between trust and writing outcomes are related to the source of writing feedback. This is especially important in contexts in which high stakes timed writing is the most frequent type of writing that L2 writers find themselves engaging in. Using a quasi-experimental design, we addressed these issues by examining the effects that perceived feedback source had on two groups of high-intermediate second language writers (n = 61 perceived teacher feedback group; n = 60 perceived AWE feedback group) enrolled in academic reading and writing courses in Northern Taiwan. Both groups received the same type of instruction and feedback for 18 weeks with the only caveat of perceiving the feedback source as coming from only the teacher or the AWE software. Results showed that the perceived AWE feedback group significantly outperformed the perceived teacher feedback group in their L2 writing by the end of the course. In addition, the perceived AWE feedback group put more trust in feedback on grammar rules and lexical choices than the perceived teacher feedback group. This indicates that high-intermediate L2 writers may trust AWE software more for specific types of feedback, such as those related to grammar and vocabulary. These and other results suggest decisions on providing AWE or teacher feedback should be made based on error type.