Year: 2026
Type: article
Abstract: Artificial intelligence (AI) tools for speaking practice are increasingly used in language classrooms, raising questions about how such feedback is experienced relative to teacher-led practice. This convergent mixed-methods study examined post-course learner perceptions of AI-mediated and teacher-led speaking feedback in a university English speaking course. Fifty-three students at an English-medium university completed an end-of-course survey after eight weeks of either AI-mediated feedback (FlowSpeak) or teacher-led feedback. Four constructs were examined: competence, relatedness, expectancy, and cost. Quantitative analyses (N = 47) showed no statistically significant between-group differences on competence, relatedness, expectancy, or cost, although descriptively the AI group reported somewhat higher competence and expectancy and lower cost. Given the modest sample size, posttest-only design, and insufficient statistical power for small-to-medium effects, these patterns are interpreted as inconclusive rather than as evidence of equivalence. Thematic coding of brief open-ended comments (N = 48) suggested that learners experienced AI as a private, low-judgement space for repeated practice, while teachers and peers remained the main sources of social connection. AI appeared to reduce social-evaluative pressure for some learners while introducing effort-, time-, and technology-related burdens. In this implementation, AI-mediated speaking feedback was generally perceived as complementing rather than replacing classroom relationships, but stronger claims require larger samples, baseline measures, and richer qualitative data.
Authors: Mohsen Askari, Azam Samadi Rahim
Language: English
Subfield: Language and Linguistics
Field: Arts and Humanities
Domain: Social Sciences
SDG: Quality education
Open Access status: diamond
OpenAlex ID: W7167443427