GenAI as a learning partner

Supporting self-regulated learning over time without replacing effort

Abstract

Generative AI (GenAI) offers significant potential to scaffold self-regulated learning (SRL) by acting as an adaptive agent or “co-regulator”. However, effectively balancing technological assistance with student effort requires AI systems that recognise SRL not as a static trait, but as a temporal and personalised process. This study investigates these dynamics over a full semester, utilising surveys at the beginning, during and end of semester to track the SRL, motivation, and emotion of 75 first-year university students. We first examine pre- and post-semester shifts, finding individual consistency alongside systemic declines in metacognitive knowledge and wellbeing. We then analyse week-to-week fluctuations, identifying curriculum demands—such as major assessment deadlines—as primary drivers of shifts in student internal states. Finally, we provide a proof-of-concept demonstration by leveraging this longitudinal and contextual information within a Large Language Model to generate tailored support directions. Our findings demonstrate that, when provided with personal, temporal, and contextual information, GenAI can identify appropriate directions for SRL support that respond to a learner’s evolving cognitive and metacognitive needs. This work underscores that, for GenAI to function as an effective learning partner that preserves rather than diminishes student effort, it must be designed with strong contextual awareness, adapting its scaffolding to support students without replacing their cognitive effort.