Reinforcement learning personalized instruction for psychological engagement self-regulated learning motivation and...

Adaptive educational technologies are increasingly used to personalize learning pathways. However, limited evidence has examined whether reinforcement learning (RL)-personalized instruction is associated with psychologically meaningful learning outcomes in Professional English education. This study compared an RL-personalized instructional pathway with a fixed-path control pathway among 200 learners across an 8-week, 16-session Professional English program. Student-level data included baseline characteristics, English proficiency, psychological engagement, self-regulated learning, motivation, user experience, and performance outcomes. In contrast, session-level data captured learner states, instructional actions, policy type, reward values, difficulty, task category, accuracy, response time, task completion, time on task, and momentary engagement process scores. The groups were comparable at baseline in demographic characteristics and English proficiency. The RL-personalized group showed higher process-level learning activity, including higher reward values, greater exposure to difficulty, more completed tasks, longer time on task, shorter response times, higher task accuracy, and higher momentary engagement process scores. Post-intervention outcomes also favored the RL-personalized group for psychological engagement, self-regulated learning, motivation, user experience, gain score, and final achievement. Adjusted models showed that these associations remained after controlling for baseline English proficiency and participant-level covariates. Because baseline levels of psychological engagement, self-regulated learning, and motivation were not measured, psychological outcomes were interpreted as adjusted post-intervention between-group differences rather than as within-person improvement. These findings suggest that RL-personalized Professional English pathways may support psychologically responsive adaptive instruction when learner-state estimation, difficulty calibration, feedback, professional task relevance, and support for self-regulation are integrated into the learning design.