智能教学反馈的目标、机制与架构

The Goals,Mechanisms and Architecture of Intelligent Teaching Feedback

反馈是有效教学中不可或缺的要素,智能技术在教育领域的兴起为形成性、个性化的教学反馈带来新的可能。对相关研究进行系统性梳理能够为智能反馈系统的设计、开发与应用提供指导。从目标、作用机制和系统架构三个维度对智能教学反馈的研究现状进行分析后发现:基于智能技术的教学反馈旨在为学习者的认知、情感与态度、元认知带来积极影响,促使学习者改进当前状态和学习成效。在反馈作用机制层面,基于正确/错误响应的自动反馈通常采用基于规则的匹配模式,通过对比答案正误来触发反馈;个性化教学反馈的关键在于构建学习者模型,使教学反馈适应不同学生的特征,并结合本体论或采用多轮人机对话的方式实现教学反馈的个性化;人机协同反馈通过实现教师、学生与计算机之间的互动与互补,来提升教学反馈的有效性。在系统架构层面,反馈系统由输入层、分析层、推理生成层三个层次组成,通过一系列算法模型和反馈机制生成反馈决策,其与学习者模型、专家模型、领域模型的相互作用决定着教学反馈的精确性与适应程度。

Feedback is an indispensable element of effective teaching and learning,and the rise of intelligent technologies in education has brought new possibilities for formative and personalized feedback.A systematic review of related research can provide guidance for the design,development and application of intelligent feedback systems.An analysis of the current state of research on intelligent technology-enabled feedback in terms of objectives,mechanisms of action,and system architecture has revealed that intelligent technology-based feedback aims to bring positive effects on learners’cognition,emotions and attitudes,and metacognition in order to motivate learners to improve their current states and learning outcomes.At the level of feedback mechanism,an automatic feedback based on correct and incorrect responses usually adopts a rule-based matching model to trigger feedback by comparing correct and incorrect answers;the key to personalized feedback is to build a learner model to adapt feedback to different students’characteristics and to realize personalized feedback by combining ontology or using multiple rounds of human-computer dialogue;a human-computer collaborative feedback can be achieved by realizing interaction and complementarity between teachers,students and computers.At the system architecture level,the feedback system consists of three layers:input layer,analysis layer and inference generation layer,generating feedback decisions through a series of algorithmic models and feedback mechanisms,whose interactions with learner models,expert models and domain models determine the degree of accuracy and adaptability of teaching feedback.