Effects of Generative AI Chatbot on English Learners’ Speaking Anxiety: A Case Study
摘要
This study examines how generative AI chatbots reduce English-speaking anxiety among Chinese university students through a four-week quasi-experiment. 62 participants from a University in Wuhan were randomly assigned to an experimental group using Doubao, an AI chatbot developed by ByteDance with context-aware dialogue, instant feedback, and dynamic topic, and a control group practicing without AI. Both groups completed two 10-m weekly sessions using IELTS Speaking Test Part 1 topics. Data were collected through pre- and post-test questionnaires based on the Foreign Language Classroom Anxiety Scale (FLCAS), translated and adapted into Chinese, alongside semi-structured interviews. Results revealed significant anxiety reduction in the EG compared to the CG, with a statistically marked difference. Participants emphasized that Doubao reduced.their fear of errors and peer judgment for its encouraging practice environment. But they also highlighted the necessity for more coherent topic shifts and motivational features. The findings confirm generative AI chabots’ efficacy in creating low-anxiety practice spaces, supplementing traditional language pedagogy and offering practical insights for EFL educators.