AI生成文本检测技术研发赋能学科建设路径研究

A Study on the Pathways for Empowering Discipline Construction through the R&D of AI-Generated Text Detection Technology

摘要: 针对生成式人工智能(AIGC)引发的学术不端风险与考核同质化危机，本研究立足人工智能赋能教育的政策导向，在相关理论与前期技术研究基础上，构建“技术检测、制度规范、教育赋能”三位一体的探索性协同框架。技术上，尝试构建“双轮驱动”大语言模型文本检测体系，并通过公开数据集实验初步验证其检测性能；制度上，从应用边界、评价体系、监管流程、长效保障四个维度开展全链条设计，探索设计“系统检测、专家复核、过程溯源”三级监管机制；教育上，将检测技术转化为认知反馈工具，设计三阶段培养路径与“教师–生成式AI–AI生成文本检测–学生”四元协同教学组织形式。研究为智能时代高校学术诚信治理与人才培养模式转型提供了一个探索性的框架，为后续教育场景实证研究与实践验证提供参考。

Abstract: To address the risks of academic misconduct and the homogenization of assessment practices associated with generative artificial intelligence (AIGC), this study is guided by the policy orientation of AI-enabled education, building on relevant theoretical perspectives and prior research on AI-generated text detection, and develops an exploratory collaborative framework integrating technological detection, institutional regulation, and educational empowerment. Technically, a dual-driven large language model-based text detection system is explored, with its detection performance preliminarily validated through experiments on public dataset. Institutionally, a whole-process framework is designed across four dimensions: application boundaries, assessment systems, regulatory procedures, and long-term safeguards, with an exploratory three-tier regulatory mechanism comprising system-based detection, expert review, and process traceability. Educationally, AI-generated text detection is repositioned as a cognitive feedback tool, and incorporated into a three-stage talent development pathway and a four-party collaborative teaching model involving instructors, generative AI, AI-generated text detection, and students. The study provides an exploratory framework for academic integrity governance and the transformation of talent cultivation models in higher education in the intelligent era, offering a reference for subsequent empirical research and practical validation in educational settings.