Key points are not available for this paper at this time.
ABSTRACT Artificial intelligence (AI)‐based pronunciation feedback is increasingly used in English as a foreign language learning, yet its pedagogical value depends not only on the delivery of automated feedback but also on how learners interpret, evaluate, and use such feedback in their pronunciation learning. Drawing on learner‐mediated perspectives on feedback processing, this study examined a baseline‐controlled longitudinal pathway linking perceived diagnosticity of AI pronunciation feedback, trust in AI pronunciation feedback, pronunciation self‐efficacy, and overall pronunciation performance. A two‐wave full‐panel longitudinal design was employed in a public college English course at a teacher education university in southwestern China. The participants were 790 first‐year students who regularly used FIF Oral English for after‐class pronunciation practice. Questionnaire data and teacher‐rated pronunciation performance were collected at both time points, with an interval of approximately ten weeks. Data were analyzed through reliability analysis, confirmatory factor analysis, longitudinal measurement invariance testing, and structural equation modeling with bootstrapped estimates of indirect associations. The results showed that, after controlling for baseline levels, perceived diagnosticity at Time 1 significantly predicted trust in AI pronunciation feedback at Time 2. Trust at Time 2 was positively associated with pronunciation self‐efficacy at Time 2, which in turn was positively associated with overall pronunciation performance at Time 2. The chained indirect association from perceived diagnosticity to overall pronunciation performance through trust and pronunciation self‐efficacy was also significant. These findings suggest a learner‐mediated pathway in which AI pronunciation feedback is linked to pronunciation performance when learners perceive it as diagnostically useful, regard it as trustworthy, and connect it with their beliefs about pronunciation improvement.
Huang et al. (Tue,) studied this question.