初高中数学错题智能体技能的设计与实践

Design and Practice of Intelligent Agent Skills for Wrong-Answer Analysis in Junior and Senior High School Mathematics

摘要: 本研究面向初高中数学错题处理场景，提出一套基于Agent Skill架构的错题智能体设计框架。该框架围绕“采集–解析–诊断–报告”闭环工作流，将题目解析、错因诊断、学情分析与对话辅导封装为四个可调用的技能模块，并构建覆盖概念性、程序步骤性、计算性、符号表示性四类错误的错因模板库，将诊断输出与课程标准对齐，实现从“知道错了”到“知道怎么改”的可执行反馈。研究同时提出省时、高效、针对性三维评估指标体系，并就AI辅助学习的伦理边界展开讨论。

Abstract: This study addresses the problem of handling wrong answers in junior and senior high school mathematics, proposing a design framework for a wrong-answer intelligent agent based on the Agent Skill architecture. Centered on a closed-loop workflow of “collection-parsing-diagnosis-reporting”, the framework encapsulates four callable skill modules—question parsing, error diagnosis, learning situation analysis, and conversational tutoring—and constructs an error template library covering four error types: conceptual, procedural, computational, and symbolic. Diagnostic outputs are aligned with curriculum standards to deliver executable feedback that moves students from “knowing what went wrong” to “knowing how to improve”. The study further proposes a three-dimensional evaluation framework encompassing time efficiency, learning gains, and individual differentiation, and discusses the ethical boundaries of AI-assisted learning.