ARTIFICIAL INTELLIGENCE-BASED CHATBOTS AND THE FLUENCY–ACCURACY TRADE-OFF IN UNIVERSITY EFL LEARNERS’ SPOKEN ENGLISH PRODUCTION

Abstract

This thesis examines the fluency–accuracy trade-off in university EFL learners’ spoken production through Skehan’s model. AI chatbots are considered as a tool for low-stakes task repetition. Existing research shows that repeated chatbot practice improves fluency and confidence but does not automatically improve accuracy; the outcome depends on whether the chatbot provides explicit corrective feedback. The thesis recommends combining fluency-oriented repetition with feedback-rich, accuracy-focused chatbot tasks and calls for measuring fluency and accuracy separately in future research.