融合人工智能与成果导向教育理念的“基因工程”智慧课程构建与教学实践

Construction and teaching practice of an AI- and OBE-integrated smart course Genetic Engineering

摘要:

针对北京理工大学研究生“基因工程”课程在CRISPR-Cas9等核心知识点上概念易混、理论难以迁移到实验方案设计、教学资源分散导致学习链条割裂，以及复杂思维过程与科研能力难以进行过程性评价等突出问题，本研究探索了人工智能(artificial intelligence, AI)与成果导向教育(outcome-based education, OBE)深度融合的课程建设思路。课程团队基于教学诊断结果，构建了“知识图谱-能力矩阵-智能学伴”三位一体的课程架构，将421个核心知识点与研究生毕业要求能力指标点建立系统映射，形成支持个性化学习路径与能力达成度追踪的“知识-能力”关联矩阵；面向研究生科研训练需求，设计科研任务型AI学伴，围绕基因治疗载体设计等综合任务，利用检索增强生成(retrieval-augmented generation, RAG)与课程专属大模型分析学生决策路径和任务产出，用于复杂思维与科研能力的过程性评价；同时构建数据驱动教学分析驾驶舱，实时汇聚学习行为、知识掌握和能力表现数据，辅助教师识别概念瓶颈与能力短板，并优化教学干预。2025年秋季学期的教学实践初步表明，采用该平台的实验班的期末成绩高于2个对照班(P<0.05)，学生日均活跃度达87%，人均日提问5.3次，采用该平台的多数学生认为自主学习效率与前沿技术理解有所提升。研究结果初步表明，相较于侧重资源整合与内容推荐的常规智慧课程，本课程通过以真实科研任务为牵引、以能力达成为核心的AI赋能路径，在一定程度上回应了研究生层面知识迁移困难与能力评价不足等教学问题，为生命科学研究生课程的智能化改革提供了可借鉴的实践路径。

Abstract:

Several critical challenges exist in the graduate course Genetic Engineering at Beijing Institute of Technology, including conceptual confusion in core topics such as CRISPR-Cas9, difficulties in transferring theoretical knowledge to experimental design, fragmentation of teaching resources across multiple platforms, and the lack of process-oriented assessment for complex reasoning and research competence. In response, we propose a new paradigm that deeply integrates artificial intelligence (AI) and outcome-based education (OBE) in course design and implementation. Specifically, we construct a triadic framework of knowledge graph-ability matrix-intelligent learning companion, in which 421 core knowledge points are systematically mapped to the graduate program’s ability indicators, forming a dynamic knowledge-ability matrix that supports personalized learning paths and traceable competence development. A research task-oriented AI companion is further designed for the graduate context. By leveraging retrieval-augmented generation (RAG) and a course-specific large language model (LLM) around comprehensive tasks such as the vector design for gene therapy, the system analyzes students’ decision-making trajectories and task outputs to enable process-based evaluation of complex thinking and research skills. In addition, a data-driven teaching analytics cockpit aggregates learning behaviors, knowledge mastery, and ability performance in real time, helping instructors accurately identify conceptual bottlenecks and competence gaps and refine instructional interventions. Teaching practice conducted in the Fall 2025 semester suggests, within the context of this study, that students in the experimental class using the platform achieved higher final examination scores than those in two parallel control classes (P<0.05). The platform was also associated with high learning engagement, with a daily activity rate of 87% and an average of 5.3 questions per student per day. Most students reported observed improvements within this study in self-directed learning efficiency and understanding of advanced technologies. Overall, the findings indicate that compared with conventional smart courses primarily focusing on resource integration and content recommendation, this AI- and OBE-integrated course-centered on authentic research tasks and competence attainment-partially addresses challenges related to knowledge transfer and competence evaluation in graduate education, providing a referenceable practical pathway for smart-course reform in life science graduate programs.