可解释自动批阅模型构建与应用

Research on the Construction and Application of Explainable Automatic Scoring Model

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。

Automatic scoring is an important realization form and basic function of intelligent educational evaluation.Deep learning-based automatic scoring models are gradually maturing but their internal structure is becoming increasingly complex and the decision-making process is not transparent,which makes it difficult for users to trust their score results and affects large-scale deployment.Therefore,this study proposes a basic framework for an interpretable automatic score model,which includes an automatic score base module,an automatic score interpretation module,and an automatic score interaction module.On this basis,this study constructs an example of the interpretable auto-score model and embeds it into an intelligent tutoring system,and further carries out quasi-experiments to verify its effectiveness.The experimental results show that the intelligent tutoring learning system embedded with the interpretable auto-score model effectively enhances learners'trust,and also contributes to their technology acceptance.Meanwhile,the explanatory information in the interaction module does not increase the cognitive load of learners.Finally,this study provides suggestions and outlooks for the research of explainable Al in educational areas.