From speculation to construction: Practice and theoretical innovation in AI-empowered inquiry-based project-based learning in thermology
从思辨到建构：人工智能赋能热学课程探究式项目化学习的实践与理论创新

摘要

热学作为物理学专业核心必修课程，承载着从宏观热力学到微观统计物理的完整知识体系。然而，传统教学模式难以让学生亲身经历理论建构中充满困惑、辩驳与重构的思维历程，学生对“熵”“最概然分布”“可逆性”等概念的深层逻辑理解不足。本研究提出以人工智能为认知伙伴的“理论探究式项目化学习”路径，建构了“对话—建构—超越”（Dialogue-Construction-Transcendence, DCT）理论探究环模式。Dialogue阶段以苏格拉底式诘问制造认知冲突，Construction阶段以逻辑检验与可视化辅助理论论证，Transcendence阶段以角色反转实现前提批判与知识内化。该模式植根于分布式认知、概念转变与探究性对话三大理论支柱，并配套四维论证品质评价量规与强制性无AI建构制度，促成学生对热学核心理论的深度思辨与前提批判。论文系统阐述了DCT模式的理论基础、结构逻辑与教学设计原则，并给出了可操作的教学实施方案、提示词模板与评价体系，为理论密集型学科中利用AI进行高阶思维培养提供了一个初步的概念框架。本文提出的DCT模式尚属理论建构阶段，后续将通过试点研究加以验证。

Abstract

As a core compulsory course for physics majors, Thermology carries a complete knowledge system from macroscopic thermodynamics to microscopic statistical physics. However, traditional teaching models make it difficult for students to personally experience the thinking process full of confusion, debate, and reconstruction in theoretical construction, resulting in insufficient deep logical understanding of concepts such as “entropy, ” “most probable distribution, ” and “reversibility.” This study proposes a “theoretical inquiry-based project-based learning” pathway with artificial intelligence as a cognitive partner, and constructs the “Dialogue-Construction-Transcendence” (DCT) theoretical inquiry cycle model. In the Dialogue stage, Socratic questioning is employed to create cognitive conflict; in the Construction stage, logical verification and visualization assist theoretical argumentation; in the Transcendence stage, role reversal facilitates premise critique and knowledge internalization. Grounded in the three theoretical pillars of distributed cognition, conceptual change, and exploratory talk, the model is equipped with a four-dimensional argumentation quality rubric and a mandatory AI-free construction system, promoting students' deep reflection and premise critique of core thermology theories. This paper systematically elaborates the theoretical foundation, structural logic, and instructional design principles of the DCT model, and provides an operable teaching implementation plan, prompt templates, and evaluation system, offering a preliminary conceptual framework for cultivating higher-order thinking with AI in theory-intensive disciplines. The DCT model proposed in this paper is still at the theoretical construction stage and will be verified through pilot studies in subsequent research.