AI-Assisted Web-Based Corrective Feedback To Improve Senior High School Students’ Listening Comprehension

Abstract

Listening receives limited class time at the senior high school level, and large classes rarely allow teachers to give personal written feedback on students’ open-ended answers. Web-based listening tools and AI scoring systems both exist, but the two are seldom integrated so that learners receive immediate corrective feedback on free-text listening responses. This study addressed that gap by developing and testing a web-based learning system assisted by artificial intelligence (AI), and it examined both the feasibility and the effectiveness of the product. The Research and Development method was applied through the Integrative Learning Design Framework across three phases, namely exploration, enactment, and evaluation. The product is a website named Ear Up! that delivers CEFR-graded audio lessons, collects essay answers, and uses an OpenAI GPT-4o model to return automated corrective feedback and next-lesson recommendations. The system was trialed with 28 tenth-grade students. Expert and user evaluation produced a mean feasibility score of 4.67 out of 5, in the very feasible band. Under a one-group pretest-posttest design, listening for details improved at a medium level (N-gain = 0.67) and listening for inference at a low level (N-gain = 0.25), while classical mastery rose from 29.41% to 82.35% and from 32.35% to 67.65%. The findings indicate that automated corrective feedback can extend listening practice beyond limited class time, yet higher-order inference still depends on teacher mediation. AI-assisted corrective feedback is therefore best positioned as a teacher-supervised supplement in future language learning.