This longitudinal mixed-methods exploratory study examines changes in English as a Foreign Language (EFL) learners’ speaking performance, speaking-related anxiety, and learning motivation within an AI-integrated Task-Based Language Teaching (TBLT) framework. Fourteen low-proficiency adult learners (CEFR A1–A2) in a continuing education context participated in a two-semester intervention incorporating AI tools for speaking practice, feedback, and task support. Given the small sample size and non-normal data distribution, non-parametric analyses were conducted. A Friedman test revealed a statistically significant difference in speaking performance across four time points, χ²(3, N = 14) = 19.57, p < .001, with a moderate-to-large effect size (Kendall’s W = .47). Post-hoc Wilcoxon signed-rank tests indicated significant within-semester improvements in both Semester 1 (Z = −2.51, p = .012, r = .67) and Semester 2 (Z = −2.64, p = .008, r = .71). In Semester 1, speaking-related anxiety significantly decreased (Z = −3.18, p = .001, r = .85), while confidence increased significantly (Z = −2.94, p = .003, r = .79), both with large effect sizes. In contrast, motivation showed small-to-moderate positive changes in Semester 2, although these did not reach statistical significance. Qualitative reflections further suggested that AI tools supported rehearsal, reduced anxiety, and enhanced learner confidence. These findings highlight meaningful longitudinal patterns and suggest the potential of AI-integrated TBLT to support both performance and affective development in low-proficiency EFL learners.

Authors:
Wei-chi Pan, Wenzao Ursuline University of Languages, Taiwan

About the Presenter(s)
Vicky Wei-chi Pan is a lecturer at Wenzao Ursuline University of Languages. She focuses on EFL writing instruction, AI-assisted feedback, and is researching assessment literacy in vocational-track university students.

See this presentation on the full schedule – Friday Schedule

Bookmark