题名：可解释自动批阅模型构建与应用
Title: Construction and Application of an Explainable Automatic Marking Model
作者：卢宇，章志，马安瑶，陈鹏鹤
Authors: Lu Yu, Zhang Zhi, Ma Anyao, Chen Penghé
摘要：自动批阅是数字化教学平台与智能化教育评价的重要实现形式和基本功能。基于深度学习的自动批阅模型逐步成熟但其内部结构复杂且决策过程不透明，导致用户难以信任其批阅结果并影响大规模部署。本研究提出了可解释自动批阅模型的基本框架，包含自动批阅基础模块、自动批阅解释模块与自动批阅交互模块。在此基础上，本研究构建了可解释自动批阅模型的实例并嵌入智能导学系统开展准实验研究。实验结果表明，嵌入可解释自动批阅模型的智能导学系统，有效提升了学习者对自动批阅功能和系统整体的信任度，也有助于提高技术接受度，交互模块的解释性信息也不会增加学习者的认知负荷。最后，本研究提出了可解释人工智能在教育领域开展自动批阅的研究建议和展望。
Abstract: Automatic marking is an important implementation form and basic function of digital teaching platforms and intelligent educational evaluation. The automatic marking models based on deep learning have gradually matured, yet their internal structures are complex and the decision-making processes are opaque, which makes it difficult for users to trust their marking results and affects large-scale deployment. This study has put forward the basic framework of an explainable automatic marking model, which includes the basic module for automatic marking, the explanation module for automatic marking and the interaction module for automatic marking. On this basis, this study has constructed an instance of the explainable automatic marking model and embedded it into an intelligent tutoring system to conduct a quasi-experimental study. The experimental results show that the intelligent tutoring system with the embedded explainable automatic marking model has effectively enhanced learners' trust in the automatic marking function and the whole system, and also helped to improve the technology acceptance. Moreover, the explanatory information of the interaction module will not increase learners' cognitive load. Finally, this study has put forward research suggestions and prospects for the application of explainable artificial intelligence in carrying out automatic marking in the field of education.
关键词：自动批阅;深度神经网络;可解释人工智能;人机交互;智能导学系统;
Keywords: Automatic marking; Deep neural network; Explainable artificial intelligence; Human-computer interaction; Intelligent tutoring system.
基金资助：北京市教育科学“十四五”规划2021年度重点课题“人工智能驱动的新一代智能导学系统构建研究”（CHAA21036）；
Fund Project: Key Project of the 14th Five-Year Plan of Beijing Educational Science in 2021 "Research on the Construction of a New Generation of Intelligent Tutoring System Driven by Artificial Intelligence" (CHAA21036).
DOI：10.13966/j.cnki.kfjyyj.2023.05.010