计算机科学
人机交互
知识管理
人工智能
组分(热力学)
领域(数学)
钥匙(锁)
多媒体
工程类
匹配(统计)
作者
Pham Ba Hong Son,Bùi Văn Hồng
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-07-01
卷期号:: 1-5
标识
DOI:10.1109/istem-ed69857.2026.11734356
摘要
This paper proposes a transformational shift in STEM education through the “CogniSTEM” system framework, transitioning learners from passive consumers of educational technology to active “pedagogical engineers.” The study has two primary objectives: (1) to design and implement a localized, edge-computed adaptive learning system that integrates outcome-based education (OBE) with the CDIO (conceive, design, implement, operate) methodology; and (2) to empirically evaluate the system's effectiveness in optimizing cognitive load and instructional efficiency using Paas's cognitive efficiency metrics. The system architecture is driven by predictive processing and local retrieval-augmented generation (RAG), utilizing edge-computed small language models (e.g., Ollama) deployed locally to ensure ultra-low latency, complete data privacy, and zero-distraction environments. Grounded in cognitive science, the pedagogical engine anticipates cognitive bottlenecks before they occur. Instead of providing direct answers, the system deploys Socratic scaffolding based on Narciss's informative tutoring feedback to minimize prediction errors. A preliminary deployment with seven high school students across 25 learning sessions yielded a mean instructional performance of 89%, with an average cognitive load of 3.68/9 on the Paas scale. Notably, 32% of sessions were classified as having high instructional efficiency$(E>0.5)$, indicating that learners achieved high performance with minimal cognitive cost. These results suggest that the CogniSTEM framework effectively reduces extraneous cognitive load while fostering active cognitive engagement, thereby illuminating the intricate mechanics of learner cognition.
科研通智能强力驱动
Strongly Powered by AbleSci AI