Investigating the Effects of ChatGPT-Assisted English Speaking Practice Based on Student Engagement Levels

This study examined the impact of college student’s engagement with ChatGPT-based speaking assignments on their speaking proficiency and perceptions of AI-assisted learning. A comparative analysis was conducted between highand low-engagement groups, excluding mid-engagement participants. Students completed ChatGPT-based speaking assignments twice a week over nine weeks, and both quantitative and qualitative data were collected. Paired samples t-tests indicated significant improvement in Responding to Questions and Expressing an Opinion for both the high- and low-engagement groups. In contrast, independent samples t-tests revealed that the high-engagement group significantly outperformed the lowengagement group in overall speaking performance on the post-tests across both speaking tasks. These findings suggest that consistent participation meaningfully contributes to language development. Additionally, student’s perceptions of ChatGPT differed by engagement level. While low-engagement students acknowledged its convenience and emotional support, highengagement students reported enhanced confidence, reduced anxiety, and greater satisfaction with the learning process. Qualitative responses further highlighted differences in perceived strengths and limitations of AI, as well as preferences for human versus AI instruction. The results underscore the pedagogical potential of structured, AI-integrated speaking tasks and the crucial role of sustained engagement in maximizing both cognitive and affective learning outcomes.