Unlocking potential in underachieving learners: self-regulation development through AI-supported e-mentoring among socioeconomically disadvantaged students

Underachieving students from socioeconomically disadvantaged backgrounds often experience difficulty sustaining self-regulated learning (SRL), which contributes to persistent gaps in participation and academic achievement. This study evaluated a three-tiered AI-supported e-mentoring intervention designed to strengthen SRL through goal clarification, micro-strategy prompting, reflective coaching, and growth-oriented maintenance support. In a cluster randomized controlled trial with pre-intervention, post-intervention, and short follow-up assessments, students in the AI-supported mentoring condition showed greater gains than controls in SRL, with improvements observed in both scale-based SRL indicators and platform-based behavioral markers (e.g., planning and task completion). These gains were accompanied by better academic performance, learning engagement, and persistence. Mechanism-oriented analyses further suggested that SRL operated as a proximal pathway linking the mentoring intervention to broader learning-related outcomes. Equity-focused moderation analyses indicated a compensatory pattern, with larger benefits among students with lower baseline self-regulation and greater socioeconomic constraints. Follow-up estimates suggested that effects were largely maintained after the intensive phase, although a modest attenuation was observed. Robustness checks and missing-data diagnostics supported the stability of the findings across alternative specifications. Overall, the findings suggest that AI-supported e-mentoring, implemented with teacher oversight and safety protocols, can support SRL development and promote more sustained engagement and achievement among underachieving learners.