教育评价新趋向:智能化测评研究综述

New Trend of Educational Assessment:A Research Overview of Intelligent Assessment

教育评价改革在我国受到了前所未有的重视,然而受限于传统测评手段,个性化评价和过程性评价难以得到有效实现,教育评价改革需要新型的测评技术作为支撑。计算机技术与人工智能的发展为测评领域注入了新的活力,推动了智能化测评的产生,为我国教育评价改革提供了新的解决方案。目前,智能化测评已经在学生能力和知识水平评估、人格与心理健康评估以及教学过程评估等方面取得进展:(1)学生能力和知识水平评估突破纸笔测验的局限,转向过程性评价、综合能力评价和动态性评价;(2)人格与心理健康评估摆脱了对自陈量表法的依赖,依据多模态数据有望实现无痕式和伴随式评估;(3)智能分析技术助力教学反馈,使教学过程评估更加直接、便捷和精准。智能化测评对教育评价改革起到了重要促进作用,但在多模态数据应用、模型的针对性、精细化、准确性与可解释性以及信效度检验等多个方面仍然存在问题。后续需探索多模态数据的协同分析,在权衡模型的准确性和可解释性基础上,提高模型的针对性和精细化,打通信息科学与测评领域的学科壁垒,确保模型可实用、可泛化和可扩展。

The reform of educational assessment has received unprecedented attention in China.However,limited by traditional assessment tools,both personalized and process assessments are difficult to achieve effectively.Therefore,new assessment technologies are required by the reform.The development of computer and artificial intelligence technologies has injected new vitality into the field of assessment,and it promotes the emergence of intelligent assessment and provides new solutions for the reform of educational assessment.At present,intelligent assessment has made progress in the assessment of students’knowledge level or capacity,personality and mental health,and teaching process.First,Intelligent assessment has broken through the limitations of paper-pencil test,leading to the emergence of process assessment,comprehensive assessment,and dynamic assessment of students’knowledge level or capacity.Second,self-report scale is no longer the only way,but traceless and adjoint assessments based on multi-modal data will be realized to evaluate personality and mental health.Third,intelligent analysis technology with teaching feedback makes the assessment of teaching process more direct,convenient and accurate.Intelligent assessment has played an important role in promoting the reform of educational evaluation,but there are still some problems in the aspects of multi-modal data application,the pertinence,refinement,accuracy and interpretability of the model.In the future,the collaborative analysis of multi-modal data needs to be explored,the accuracy and interpretability must be balanced,and the pertinence and refinement of the intelligent model need to be improved.Meanwhile,the discipline barriers in the field of information science and psychological measurement must be broken through,to ensure that the intelligent models can be practical,generalized and extensible.