从项目式学习到人机协同：生成式人工智能赋能高职工科竞赛育人的演进机制与可迁移路径

From Project-Based Learning to Human-AI Collaboration: Evolutionary Mechanisms and Transferable Pathways of Generative AI-Empowered Engineering Competition Education in Vocational Colleges

摘要: 文章以研究团队2021~2025年围绕全国大学生智能汽车竞赛与全国大学生电子设计竞赛开展的四项连续研究为对象，借用行动者网络理论并将其转译为“竞赛网络–多方协作权责–必经审核关口”的职教话语，回顾并剖析竞赛育人模式从“跨学科项目式学习”经“AI增强型探究式学习”到“三元人机协同育人”的演进机制。研究发现：三个阶段的实质差异不在于是否使用了AI工具，而在于竞赛网络的参与方构成变化与必经审核关口的位移——教师由“知识与方案的核心供给者”转型为“工程可行性的最终审定者”；大语言模型承接基础知识供给与方案生成职能后，其“建议失效”事件反而成为学生建立批判性工程素养的关键契机。本文把该素养操作化为三个维度、六项子维度与十六条可观测行为指标并配套观测量规，并以一个关键技术决策点为例完整追溯各参与方之间的协商链条；据此提炼“三元权责分工–四阶递进流程”的实施框架，并给出四条实训迁移路径。上述结论的成立依赖“实物可证伪”的强反馈环境与高职竞赛情境，向其他学科、教育层次与常规课堂的迁移尚待验证。

Abstract: Taking as its object four consecutive studies conducted by the authors’ team between 2021 and 2025 on the National University Students’ Intelligent Car Racing Competition and the National Undergraduate Electronics Design Contest, this paper borrows Actor-Network Theory and translates it into a vocational-education vocabulary of competition network, multi-party rights and responsibilities, and obligatory review gate. Three findings are reported. First, what distinguishes the three stages is not whether AI tools are used, but how the participants in the competition network are recomposed and where the obligatory review gate is relocated: the instructor shifts from the central supplier of knowledge and solutions to the final arbiter of engineering feasibility. Second, once large language models absorb the supply of foundational knowledge and the generation of candidate solutions, the episodes in which their advice fails in practice become the decisive moments for building students’ critical engineering literacy; the paper operationalises this literacy into three dimensions, six sub-dimensions and sixteen observable behavioural indicators with an accompanying observation rubric, and traces in full the chain of negotiation among students, instructors, the model, components, costs and contest rules at one key technical decision point. Third, an implementation framework comprising a triadic division of rights and responsibilities and a four-phase procedure is distilled, together with four pathways for migrating competition outcomes into routine practical training. These claims rest on a strong-feedback setting in which physical artifacts can falsify AI advice and on a vocational-college competition context, so transfer to other disciplines, educational levels and routine classroom teaching remains to be verified.