Open access
•
Original Paper
•
Published: 22 October 2026
Hendrik van der Linden, Nurgan Dekosaputra, Kenji Takahashi
This randomized controlled trial (N=480) evaluates dynamic Socratic prompting delivered by fine-tuned educational LLMs. Students utilizing adaptive scaffolding demonstrated a 24.6% increase in debugging self-efficacy and superior problem decomposition compared to passive reference cohorts.
Open access
•
Original Paper
•
Published: 18 October 2026
Elena Rostova, Sarah Jenkins, Chen Wei
Integrating synchronized eye-tracking, galvanic skin response, and real-time audio transcript parsing to identify cognitive overload during complex mathematical derivations. The proposed multimodal transformer achieved 91.4% accuracy in predicting task frustration.
Open access
•
Original Paper
•
Published: 12 October 2026
Faisal Al-Mansoor, Marcus Thorne, Dewi Sartika
To comply with strict data privacy mandates (GDPR/APEC), we introduce a differential-private federated learning framework that trains institutional retention predictors without sharing raw learner logs. The distributed model matched centralized baseline AUC within 1.2%.
Open access
•
Original Paper
•
Published: 05 October 2026
Claire Beauchamp, Ratna Kumalasari, David K. Miller
A cross-national comparative study analyzing 12,000 student interactions with generative chatbots across secondary schools. We identify six prompt topologies that foster dialectical inquiry rather than answer-retrieval complacency.
Open access
•
Original Paper
•
Published: 28 September 2026
Tatsuya Mori, Angela Gomez, Nurgan Dekosaputra
Tracking spatial visualization gains among 650 middle school learners over an 18-week immersive curriculum. The reinforcement-learning difficulty regulator reduced learning attrition by 41% while maintaining optimal flow states.