对话式智能导学系统研究现状及趋势  

Current Status and Trend of Conversation-Based Intelligent Tutoring System

作　　者：屈静 刘凯[2,3] 胡祥恩 杨钋[1] 蒋卓轩[4] QU Jing;LIU Kai;HU Xiangen;YANG Po;JIANG Zhuoxuan(Graduate School of Education,Peking University,Beijing 100871,China;School of Psychology,Central China Normal University,Wuhan 430079,China;College of Educational Science,Bohai University,Jinzhou 121007,China;ID AI Research,Shanghai 20001,China)

机构地区：[1]北京大学教育学院,北京100871 [2]华中师范大学心理学院,湖北武汉430079 [3]渤海大学教育科学学院,辽宁锦州121007 [4]京东AI研究院,上海200001

出　　处：《开放教育研究》2020年第4期112-120,共9页Open Education Research

基　　金：国家自然科学基金项目“面向青少年网络适应的个性化信息服务优化方法研究”(71974072);中国博士后资助项目“基于角色冲突的虚拟小组讨论构建原则及教学效果研究”(2016M590702)。

摘　　要：对话式智能导学系统通过模仿人类自然语言对话辅导,能够促进学习者的综合分析、定性推理等深度学习能力。本研究基于深度学习视角,用文献分析法对其概念内涵、理论基础、架构特点和学习效果等相关研究进行梳理,指出对话式智能导学系统对深度学习具有明显的促进作用,但也存在学习效率欠佳、深度学习支持不足及开发成本过高三个亟待解决的问题。为进一步推动对话式智能导学系统的发展,本研究建议重视跨学科合作、引入通用智能导学框架、考虑潜在的伦理问题,同时着重关注多模态交互方式、多维度情感计算和多代理团队学习三个新兴研究方向。As an interdisciplinary frontier research field, Conversation-based Intelligent Tutoring System(CBITS) has been highly anticipated because it is a possible path for the scale, automation, and humanization of deep learning. From the three domains(cognitive, intrapersonal, interpersonal) of deep learning perspectives, this study analyzed literature of the definition of CBITS, theoretical basis, architecture features, learning effects and then pointed out that CBITS can promote learners’ comprehensive analysis, qualitative reasoning as well as other deep learning abilities by imitating human’s one-on-one natural language tutoring. However, CBITS’s learning efficiency, insufficient deep learning comprehensive support, and the high cost of development still need to be improved. For better efficiency and effect, the future research of CBITS will mainly focus on multimodal interaction, multi-dimensional affective computing, and multi-agent team tutoring. At the same time, interdisciplinary collaboration, introducing the Generalized Intelligent Framework for Tutoring(GIFT) and the ethics of Artificial Intelligence should also be taken into future consideration.

关 键 词：深度学习 对话 对话式智能导学系统 通用智能导学框架 

分 类 号：G434[文化科学—教育学]