个性化测评:下一个测评前沿  

Personalization:The Next Assessment Frontier

作　　者：Randy E.Bennett 金婧(译) Randy E.BENNETT(ETS Research Institute,Princeton,NJ 08541,USA)

机构地区：[1]美国教育考试服务中心研究院,普林斯顿NJ 08541 [2]教育部教育考试院,北京100084

出　　处：《中国考试》2025年第5期1-9,共9页journal of China Examinations

摘　　要：传统意义上严格的标准化测评已是属于过去时代的理念,个性化测评才是未来的发展方向。之所以得出这一论断,是因为随着各国人口多样性的增加,标准化测评非但无法保障公平,反而可能会削弱其实现。实际上,教育测量领域内早已有一些个性化测评的实践先例。从这些先例中,可以提炼出个性化测评的通用方法,其中若干方案已在当前的实践中被广泛采用。不过,无论是现有方法还是更具前瞻性的人工智能测评方法,都存在一些问题,引起了考生、考试项目设计方与测评使用者等的顾虑和关注。本文简要回顾了标准化测评的起源与发展变化,探讨了两个教育测量领域的个性化测评实践先例,并详述了三类个性化测评的通用方法,最后阐述了笔者对于这一测评前沿方向的若干顾虑。In their strict form,standardized assessments treat all examinees equally by administering to each of them the same content in the same format under the same conditions,with responses scored using common rules.Personalization violates this equality principle in one or more ways.The article argues that strict standardization is an idea of the past and that personalized assessment is the future.That argument stems from the observation that,as population diversity increases,standardization may undermine fairness rather than facilitate it.Several precedents exist for personalizing assessment from within the field of educational measurement.These precedents include the field's technical standards,which grant conditional approval to departures from standardized procedures,and the practices of operational testing programs,which frequently make such departures based on examinee exceptionality,native language,interests,personal identity,and preference for test location.From these precedents,several general approaches to personalized assessment can be derived,instances of which are widely used operationally today.These approaches include machine-driven artificially intelligent ones,examinee-driven ones,and combinations of the two.The existing approaches,as well as the more futuristic artificially intelligent ones,raise substantial concerns for examinees,testing programs,and test users alike.Among those concerns are the absence of conventionally accepted design frameworks,the heavy dependence on examinee choice in the operational personalized assessments,and the possibility that personalization will lead to overestimating proficiency.Concerns related to machine-driven approaches include the high requirements for such systems to work effectively,their potential for bias,and the challenges black-box algorithms pose for defensibility.

关 键 词：个性化测评 标准化测评 测评方法 教育评价与测量 教育评价改革 

分 类 号：G405[文化科学—教育学原理]