The Framework Design of Intelligent Assessment Tasks Recommendation System for Personalized Learning

A novel multi-agent artificial intelligence (AI) architecture is introduced to enhance engineering education through agentic collaborative assistants. Conventional educational AI systems, particularly those utilizing large language models (LLMs), exhibit limitations in generating context-specific instructional content and addressing diverse cognitive levels. To overcome these constraints, a distributed, task-specialized system has been developed, comprising hierarchical macro-agent teams with defined roles: Team 1 (Chapters) analyzes the syllabus to establish a logical progression of learning; Team 2 (Content) curates and generates instructional material; Team 3 (Testing) produces assessments aligned with Bloom’s taxonomy; and Team 4 (Evaluation) ensures quality based on standard rubrics and outcome alignment. The distributed architecture enables efficient task parallelization, specialization, agent accountability, and modular optimization—capabilities unattainable in monolithic LLM-based systems. Quantitative analysis indicates significant gains: a 14–26% improvement in content quality (technical accuracy, conceptual clarity, cognitive coverage), a 35% reduction in computational requirements, a 68% decrease in API costs, and a 62% increase in processing speed. Performance is particularly strong in complex domains such as embedded systems, where depth of expertise and cognitive richness are essential. The modular structure supports incremental updates without requiring full system retraining, ensuring alignment with evolving engineering knowledge. Enhanced transparency through decision tracking and process visibility allows educators to validate and refine generated content. This approach demonstrates how multi-agent specialization and collaboration can effectively address limitations of monolithic systems, offering a scalable solution for delivering high-quality, domain-specific educational materials across engineering disciplines.