AI SPEECH RECOGNITION IN HIGH-SCHOOL EFL SPEAKING ASSESSMENT: EFFECTIVENESS, MEANINGFULNESS, ENJOYMENT

Abstract

This mixed-methods study examined the use of an AI-powered automatic speech recognition (ASR) tool (SpeechAce) for high-school EFL speaking assessment, with a focus on perceived effectiveness, meaningfulness, and enjoyment. Thirty students completed pre- and post-surveys, and ten students participated in semi-structured interviews; classroom observations were documented using a structured checklist. The intervention lasted two weeks (four 40-minute meetings) in which students accessed SpeechAce on their smartphones via the web and followed an iterative speak–check–revise routine. Quantitative results (paired-samples t-tests) showed significant gains in perceived speaking skill and meaningfulness, whereas enjoyment remained stable. Qualitative findings indicated that automated, non-judgmental feedback supported lower anxiety, increased willingness to speak, and greater learner autonomy, while teacher facilitation (warm-ups, modelling, and brief mini-lessons) helped students interpret feedback and set revision goals. These findings suggest that ASR-supported assessment can function as a formative learning episode in school contexts when tasks, feedback interpretation, and revision opportunities are explicitly structured.