This study aims to examine the effectiveness and underlying mechanism of a Human–AI co-teaching model implemented within a 5E instructional framework in a Basic English Grammar course. It also aims to examine if Human–AI co-teaching enhances learner engagement and learning results relative to traditional teaching, and to clarify the instructional process behind this effect through a systematic mechanism.
A quasi-experimental design was employed, involving an experimental group receiving Human–AI co-teaching and a control group receiving traditional instruction. Data were collected through post-course questionnaires and learning assessments. Descriptive statistics, independent-samples t-tests and serial mediation analysis (Model 6) were conducted to examine both the effectiveness and the underlying mechanism of the proposed model.
The results indicate that students in the Human–AI co-teaching group demonstrated significantly higher levels of learner engagement and learning outcomes than those in the traditional instruction group. Furthermore, the findings reveal that students’ perceptions of Human–AI co-teaching positively influence learning outcomes through a sequential mediation mechanism involving teacher–AI integration quality and learner engagement. This suggests that the effectiveness of AI-supported instruction is not solely outcome-based but depends on the quality of instructional integration and student engagement.
This study contributes to the literature by providing a mechanism-level explanation of Human–AI co-teaching, rather than focusing solely on its direct effects. It introduces teacher–AI integration quality as a key construct and demonstrates how AI-supported instruction operates within a structured 5E learning framework. By employing a quasi-experimental design, the study offers robust empirical evidence and advances understanding of how AI can be effectively integrated into language education.
