Dynamic interplay between cognitive load and teaching presence among university English teachers in generative AI-augmented instruction: a longitudinal mixed-methods study

Abstract University teachers who bring generative AI into the classroom face a double demand: they must manage the mental cost of working with the tool while sustaining the pedagogical presence their students rely on. This study asks how the two are related. Survey data were collected from 186 English teachers at 24 Chinese universities across three waves of a single semester (Weeks 2, 8 and 15), and 28 of these teachers were interviewed once the final wave had closed. Structural equation modeling indicated that extraneous cognitive load was negatively associated with all three teaching presence dimensions, most strongly with facilitating discourse, whereas germane load was positively associated with presence, most clearly with direct instruction. AI proficiency moderated the extraneous-load pathway, flattening it for more proficient teachers, and AI attitude carried a small share of that association. Across the three waves extraneous load fell while germane load and teaching presence rose. The interviews corroborated these patterns for most participants and identified teachers for whom they did not hold, including some who experienced intrinsic demand as productive. Because the design is observational, spans one semester, and relies throughout on self-report, the results describe covariation rather than developmental or causal effect. The study nonetheless extends the Community of Inquiry framework by specifying cognitive antecedents of teaching presence, and offers situated guidance on tool procurement, tiered training and graduated support in Chinese higher education. Full instruments, the coding framework, the analysis code and the de-identified data set are provided at Related files.