An AI-Learner Shared Control Model Design for Adaptive Practicing

Online higher education offers great learning flexibility but demands learners’ high self-regulated learning (SRL) skills, especially in self-paced and asynchronous online learning. The lack of SRL skills in many learners often leads to poor academic outcomes, underscoring the need for SRL support. Our study introduces CAP (Confidence-based Adaptive Practicing), a model of adaptive practicing designed to enhance SRL in STEM disciplines. CAP incorporates knowledge tracing and question sequencing as two core functions. Unlike traditional adaptive learning systems that rely solely on machine control, CAP integrates learner confidence feedback and learner control in its rule-based intuitive algorithms. To avert the subjectivities of human judgement on learner confidence, CAP employs Thompson Sampling machine learning to refine the algorithms for adaptive accuracy and efficiency. This innovative AI-learner shared control approach has garnered positive feedback from field experts, highlighting its potential effectiveness in facilitating SRL.