Paper Type
Short
Paper Number
PACIS2026-1865
Description
Digital learning enables proactive intervention—providing guidance before learners explicitly request help. However, the effect of proactive support on learning performance remains unclear. On one hand, it removes the need to initiate help, potentially avoiding competence threats associated with requesting assistance. On the other hand, unsolicited help may signal inadequate ability and heighten perceived competence threats. Drawing on self-affirmation theory, we propose that learner involvement moderates this effect. Modern learning systems capture rich behavioral traces, enabling unobtrusive measurement of engagement-related behavior. Using a mixed-methods approach, we first develop a Graph Neural Network model to predict involvement from behavioral data during problem-solving, then implement this model in a digital learning system and conduct an experiment to examine how different forms of support are adapted to users’ behavioral states to influence learning performance.
Recommended Citation
Zhang, Jiayuan, "Designing Proactive Interventions for Dynamic Learner States in Digital Learning Systems" (2026). PACIS 2026 Proceedings. 11.
https://aisel.aisnet.org/pacis2026/ai_fow/ai_fow/11
Designing Proactive Interventions for Dynamic Learner States in Digital Learning Systems
Digital learning enables proactive intervention—providing guidance before learners explicitly request help. However, the effect of proactive support on learning performance remains unclear. On one hand, it removes the need to initiate help, potentially avoiding competence threats associated with requesting assistance. On the other hand, unsolicited help may signal inadequate ability and heighten perceived competence threats. Drawing on self-affirmation theory, we propose that learner involvement moderates this effect. Modern learning systems capture rich behavioral traces, enabling unobtrusive measurement of engagement-related behavior. Using a mixed-methods approach, we first develop a Graph Neural Network model to predict involvement from behavioral data during problem-solving, then implement this model in a digital learning system and conduct an experiment to examine how different forms of support are adapted to users’ behavioral states to influence learning performance.
Comments
02-FutureofWork