[Survey and Practical Report] Promoting Listening and Speaking Development through an AI Conversation App: Challenges and Prospects from Classroom Practice

This classroom practice report integrates an AI conversation application into a first-year univer- sity listening course. During the first half of the semester, individual scaffolding was provided, while in the latter half, the focus shifted to general reminders. Students' engagement was tracked through 10 weekly usage logs and questionnaires administered at three points across the semester. Results showed that on-time task completion peaked in Week 6 (94.7%) but gradually declined thereafter, with non-completion reaching 23.7% by the end of the semester. Meanwhile, learners' evaluations of the app remained stable at a moderate level in terms of usefulness and enjoyment, and their con- fidence in listening to academic content -- though still low -- increased slightly over time (T1 = 1.66, T3 = 1.78). Open-ended responses revealed perceived benefits in phonological awareness, difficulties in securing study time, and mixed opinions regarding automatic speech recognition (ASR). These findings suggest that automatic notifications alone are insufficient to sustain engagement; however, interventions at key points -- such as reiterating evaluation criteria or making small adjustments to task volume and difficulty -- can support continuous learning and foster incremental gains in learners' confidence in academic listening.