ScholarWorks@숙명여자대학교

초록

Today, children are confronted with unprecedented mental health concerns and lack timely, individualized psychoeducational aid, especially in under-resourced communities with scant professional help. The explosive advancement in 6G Internet of Everything (IoE) technology, with its ultra-low latency, massive connection density, and terabits-per-second throughput, enables real-time adaptive educational interventions in emergent learning spaces. Current methods for educational content generation with large language models fail to address issues of trustworthiness and developmental appropriateness for heterogeneous IoT devices. Therefore, this paper presents the fine-grained interpretable matrix for the psychoeducation (FGIM-P) model, which exploits the ultra-low latency backbone of 6G IoE networks to provide calibrated content through educational devices. FGIM-P structures large language models, so they first retrieve and then generate educational content by enhancing novelty, saliency, and relevance interpretability for a specific architectural purpose. The model implements a pair-wise extraction mechanism to central concept control output with interpretable masking matrices structure and verifiable content surrounding different developmental layers to achieve trustworthiness. ChildEdu/DailyLearn and student depression datasets demonstrate FGIM-P surpassing five state-of-the-art baselines in responsiveness and trustworthiness (cultural inclusivity) metrics.

키워드