Managing behavior in AI-mediated English education: The interplay of teachers, technology, and communication willingness

This study investigates the relations between teacher AI behavior management competence and the willingness of university students to communicate in English in AI mediated English as a foreign language classroom in China. Drawing on the Cognitive Theory of Multimedia Learning and cognitive load theory, we theorize that good teacher orchestration of AI signaling and pacing reduces students' extraneous cognitive load and hence helps with communication readiness. Survey data was gathered from a sample of 454 students in three cities (Xi'an; Wuhan; Nanjing) and analyzed using partial least squares structural equation modelling in SmartPLS. Teacher AI behavior management competence was a significant predictor of the willingness to communicate in English directly and indirectly through reduced extraneous cognitive load. The mediated pathway was stronger when students experienced higher levels of personalization in AI nudges and higher levels of infrastructure reliability, as is clear from evidence of cueing/segmenting and system quality affecting cognitive efficiency and downstream engagement. Importantly, given that all of the observed classrooms had used the same AI enabled platform to some degree, the findings support within platform inferences on how well teachers orchestrate AI mediated behaviors management, rather than generalizations about the efficacy of AI versus non-AI instruction. The study extends multimedia learning theory from the level of design of the message level to the level of orchestration of the classroom level, and identifies conditions under which cognitively efficient AI orchestration is congruent with student communicative readiness.