Postgraduate EFL learners often struggle to attain academic speaking proficiency because traditional instruction offers limited opportunities for individualized practice, delayed feedback, and inadequate support for anxiety regulation. This quasi-experimental study evaluated whether adaptive and collaborative AI can address these constraints by comparing three AI-driven interventions—personalized learning pathways (PLP), conversational AI tutors, and AI-mediated collaborative platforms—with traditional computer-assisted language learning (CALL) in 414 postgraduate EFL students. A pretest–posttest design with analysis of covariance, rather than raw-score comparison, statistically controlled baseline speaking performance. Outcomes targeted fluency (speech rate, pause frequency), accuracy (error density), and syntactic complexity (clauses per T-unit), as well as plus metacognitive awareness and speaking anxiety. PLP’s adaptive algorithms produced the largest and most educationally meaningful gains: fluency improved by 3.21 points; accuracy improved through an approximately 37.5% reduction in errors; and complexity increased by 1.8 clauses per T-unit. These improvements were statistically reliable, exceeded those of the other AI conditions and CALL, and corresponded to very large effects sizes. PLP learners also reported enhanced metacognitive self-monitoring (approximately 88%) and an approximately 44% reduction in speaking anxiety. Findings indicate that well-designed AI can disrupt fossilized errors, accelerate nonlinear skill integration, and create responsive, learner-centered speaking environments for advanced EFL learners.
Xu et al. (Sun,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: