聚焦问题求解能力的研究生《人工智能》研讨式案例驱动的教学构建

Constructing a Seminar-Based and Case-Driven Teaching Model for Postgraduate “Artificial Intelligence” Focused on Problem-Solving Skills

摘要: 针对目前高校研究生《人工智能》课程在跨学科授课时暴露出的理论推导枯燥、知识孤岛化、重模型调用轻底层剖析等问题，本文基于近年授课跟踪实践，提出了一种剥离传统讲授模式的研讨式案例驱动的教学重构方案。依托逆向问题抛出、细粒度实验架构和自带课题答辩等环节，打通课堂测验、研究选题和学生真实学位论文之间的壁垒。教学改革实践数据反馈表明，该教学体系成功引导了超过三分之一的选课研究生将课堂答辩和期末作业孵化为跨学科科研成果的雏形，扭转了学生科研选题无有效解决方案和仅会调参的浅层学习现状。

Abstract: In response to the issues prevalent in the interdisciplinary teaching of the postgraduate “Artificial Intelligence” course, including tedious theoretical derivations, fragmented knowledge silos, and an over-reliance on model invocation over underlying analysis, this paper proposes a reconstructed seminar-based and case-driven teaching model that breaks away from traditional lecture formats, based on teaching tracking practices in recent years, by leveraging modules such as reverse problem posing, fine-grained experimental architectures, and project defenses based on students’ actual degree thesis. Empirical data feedback from the teaching reform demonstrated that this pedagogical system has successfully guided over one-third of the enrolled postgraduates to incubate their class defenses and final assignments into prototypes of interdisciplinary scientific research outcomes. This effectively reverses the superficial learning status quo wherein students lack effective solutions for their research topics and only know how to tune parameters.