大学生人工智能素养智能评价体系研究——理论与实证研究

Research on an Intelligent Evaluation System for College Students’ Artificial Intelligence Literacy—Theoretical and Empirical Research

摘要: 在人工智能迅猛发展与广泛渗透的数字经济时代，人工智能素养已跃升为大学生必备的核心竞争力。然而，当前高校在人工智能素养评价方面普遍面临评价方式主观化、评价维度单一化以及过程评价黑盒化等现实困境，难以客观量化学生的算法实操能力与计算思维模式。鉴于此，本文提出并构建了一种融合层次分析法与抽象语法树技术的智能评价体系。该研究首先依据“认知–构建–创造–责任”四维框架重构了大学生人工智能素养评价指标；随后，开发了多专家协同的AHP权重决策系统，利用严格的一致性检验与群组几何平均算法确立了科学的指标权重；在此基础上，创新性地引入AST静态代码分析技术，实现了对学生源代码中关键AI知识点的自动提取与映射；最后，借助知识图谱技术实现了评价结果的可视化呈现。实证结果表明，该体系能够有效规避传统评价的主观偏差，精准诊断学生的编程逻辑、算法实现与AI思维短板，为高校人工智能教育的精准施教提供了坚实的科学依据。

Abstract: In the digital economy era, characterized by the rapid development and widespread penetration of artificial intelligence, AI literacy has emerged as a core competency for university students. However, higher education institutions currently face practical challenges in AI literacy assessment, including subjective evaluation methods, unidimensional assessment criteria, and a “black-box” approach to process evaluation, making it difficult to objectively quantify students’ practical algorithm skills and computational thinking models. In response to these issues, this paper proposes and constructs an intelligent evaluation system that integrates Analytic Hierarchy Process (AHP) and Abstract Syntax Tree (AST) technologies. The study first reconstructs the AI literacy evaluation indicators for university students based on a four-dimensional framework of “Cognition-Construction-Creation-Responsibility”; subsequently, it develops a multi-expert collaborative AHP weight decision system, which establishes scientific indicator weights through rigorous consistency checks and a group geometric mean algorithm; building on this, it innovatively introduces AST static code analysis technology to achieve the automatic extraction and mapping of key AI knowledge points from students’ source code; finally, it leverages knowledge graph technology to visualize the evaluation results. Empirical results demonstrate that the proposed system can effectively avoid the subjective biases of traditional assessments, accurately diagnose students’ shortcomings in programming logic, algorithm implementation, and AI thinking, and provide a solid scientific basis for precise teaching of AI literacy in higher education.