Exploring students' self-regulated learning behavioural patterns and perceptions in an English speaking task within a generative AI-supported immersive VR environment

Abstract

Generative AI (GenAI) has introduced new opportunities for self-regulated learning (SRL) by enabling adaptive, conversational support in immersive educational environments. Yet little is known about how SRL unfolds in real time during oral communication tasks with GenAI support in immersive virtual reality (iVR). Most studies rely on post-task questionnaires, leaving unclear how different types of agent support influence regulatory processes and whether GenAI agents provide more effective support than scripted ones. This randomised experimental study compared college students' SRL behaviours under two conditions: a control group interacting with a scripted agent (34 students) and an experimental group interacting with a GenAI agent (37 students). Their interactions were coded using a validated SRL framework and analysed through behavioural sequence analysis. Results showed that learners with GenAI support demonstrated more frequent reflection behaviours (21.2 % vs. 8.9 %) and greater diversity in their regulatory strategies, indicating more adaptive engagement across SRL phases. Within the GenAI condition, two learner profiles were identified: strategic learners, who sustained longer and more diverse regulatory cycles, and reactive learners, whose behaviours were shorter and more affect-driven, demonstrating that individual learners responded differently to GenAI support. Regression and ANCOVA analyses confirmed that these behavioural profiles significantly predicted learners' perceived SRL gains, after controlling for baseline differences. These findings show that GenAI agents provide adaptive support that strengthens learners' planning, monitoring, and reflection processes in immersive speaking tasks. They also underline the educational significance of linking behavioural profiles with adaptive agent design to advance personalised support in iVR language learning environments.