Enhancing Speaking Competence, Reducing Language Anxiety, and Increasing Motivation: Exploring the Effects of AI Tools on First-Year Plant Science Learners at Bonga University.

Abstract This study employed a quasi-experimental parallel mixed-methods design to investigate the effects of AI-assisted language learning tools on Bonga University Plant Science Department first year learners' speaking anxiety, speaking motivation, and speaking competence. A sample of 80 intermediate-level learners (M = 18.5, SD = 1.6) from Bonga University, Ethiopia, was divided equally into an experimental group (EG) using AI-assisted speaking tools (ChatGPT-4 voice mode, ELSA Speak Premium, Duolingo Max, and a custom progress dashboard) and a control group (CG) receiving traditional speaking instruction. Over a 10-week intervention, quantitative data were collected through pre- and post-tests measuring speaking anxiety (8 items, α = 0.84), speaking motivation (10 items, α = 0.89), and speaking competence (15-point test assessing pronunciation, fluency, vocabulary, grammar, and coherence). Qualitative data were gathered through semi-structured interviews with EG participants (n = 40). Results revealed that the EG demonstrated significantly greater reductions in speaking anxiety (post-test: EG, M = 23.45, SD = 7.92; CG, M = 29.87, SD = 6.54; ANCOVA F = 23.45, p < 0.001, η²=0.23, Cohen's d = 0.8), significantly higher increases in speaking motivation (EG, M = 38.92, SD = 7.43; CG, M = 29.56, SD = 6.87; F = 34.89, p < 0.001, η²=0.31, 0.81), and substantially greater improvements in speaking competence (EG, M = 24.67, SD = 4.15; CG, M = 18.34, SD = 4.62; F = 41.23, p < 0.001, η²=0.35, d=.84) compared to the CG. Qualitative content analysis revealed that AI-assisted tools reduced speaking anxiety through non-judgmental practice environments (85% of participants), mastery through unlimited repetition (78%), and visible progress tracking (70%); enhanced motivation through gamification (88%), adaptive challenge (82%), and immediate encouraging feedback (75%); and improved speaking competence through immediate specific corrective feedback (90%), extended practice opportunities (85%), and vocabulary/expression expansion (78%). Moderator analyses indicated that highly anxious learners showed larger anxiety reductions (d = 1.98 vs. 1.28), while technologically experienced learners showed larger competence gains (d=.82 vs. 0.65). These findings provide evidence that AI-assisted language learning effectively enhances Ethiopian university learners' speaking outcomes by simultaneously addressing affective and cognitive dimensions of oral language acquisition.