Research on Personalized Learning Strategies for Parallel Computing Courses in the Era of Artificial Intelligence: From the Perspective of Situated Learning Theory

Abstract

The rapid advancement of artificial intelligence poses new challenges to the cultivation of parallel computing talents. Traditional teaching approaches often suffer from the dual dilemmas of “separation between learning and application” and “cognitive overload.” Grounded in situated learning theory and empowered by artificial intelligence, this study proposes a personalized learning strategy model driven by a dual-wheel mechanism of “situatedness-driven + AI-empowerment.” A three-tier dynamic support system encompassing “diagnosis, recommendation, and regulation” is constructed, along with a multi-level task system anchored in authentic engineering contexts. The model is implemented and evaluated in a parallel computing course. The practical case centers on “parallel clustering analysis of large-scale high-dimensional data with holes,” guiding students through progressive mixed programming practices with OpenMP, CUDA, and MPI. Evaluation results show that the proposed model significantly improves students’ system-level tuning capabilities and engineering confidence: the experimental group achieved a speedup 2.4 times that of the control group. When facing real-world pain points such as load imbalance, 72% of the experimental groups proactively adopted dynamic data redistribution strategies, reducing communication waiting time from 38% to 14%. The study confirms that the effective integration of situated learning and AI empowerment can resolve the long-standing challenges in parallel computing education, achieving the personalized teaching goal of “context-driven instruction and learner-adaptive guidance.” Future work will expand interdisciplinary context case libraries, introduce multimodal cognitive load monitoring, and promote the development of an autonomous HPC education ecosystem.