大语言模型预训练逻辑启发下的预防医学知识体系构建

Construction of the Preventive Medicine Knowledge System Inspired by the Pre-Training Logic of Large Language Models

摘要: 生成式人工智能显著降低了医学知识检索、归纳和调用的成本，也使预防医学教育必须重新回答一个基础问题：当大量事实性信息可以通过外部工具快速获得时，学生仍需内化哪些知识，又应形成怎样的知识结构，才能支撑后续的专业判断与公共卫生实践？本文以大语言模型预训练逻辑为问题引子，以概念图、图式理论和情境学习理论等认知心理学与教育学理论为论证基础，提出预防医学“知识预训练”体系。该体系不主张增加课程数量，而是以公共卫生问题为组织中心，在现有学科课程之上建立纵向问题链，并将知识区分为需要稳定内化的核心知识与需要循证检索、动态更新的外部知识。文章进一步构建“生物医学基础–公共卫生方法–专业领域知识–人群与社会情境”四层知识结构，明确各层知识对应的职业胜任力培养要求，并给出课程设计团队、教师发展与核心/外部知识清单界定的实施策略，以营养与食品卫生学相关问题说明其教学应用。该框架旨在推动预防医学知识教学由课程覆盖和事实记忆转向知识结构、证据调用和问题表征，为后续专业任务训练奠定可持续的知识基础。

Abstract: Generative artificial intelligence has markedly reduced the cost of retrieving, summarizing, and invoking medical knowledge, forcing preventive medicine education to revisit a fundamental question: when large amounts of factual information can be rapidly obtained through external tools, what knowledge should students still internalize, and what kind of knowledge structure should they develop to support subsequent professional judgment and public health practice? Taking the pre-training logic of large language models (LLMs) as a starting point, and grounding the argument in cognitive and educational theories such as concept mapping, schema theory, and situated learning, this paper proposes a “knowledge pre-training” system for preventive medicine. Rather than increasing the number of courses, the system organizes learning around public health problems, establishes vertical problem chains across existing disciplinary courses, and distinguishes core knowledge that must be stably internalized from external knowledge that should be retrieved based on evidence and dynamically updated. The paper further constructs a four-layer knowledge structure—biomedical foundations, public health methods, professional domain knowledge, and population and social context—specifies the professional competencies to be developed at each layer, and provides implementation strategies covering interdisciplinary curriculum design teams, faculty development, and a methodological workflow for defining the core/external knowledge lists. Its teaching application is illustrated with problems related to nutrition and food hygiene. The framework aims to shift preventive medicine knowledge teaching from course coverage and factual memorization toward knowledge structure, evidence invocation, and problem representation, laying a sustainable knowledge foundation for subsequent professional task training.