面向“为学习而测评”的纵向认知诊断模型

基于“为学习而测评”的理念，以促进学生学习为目的，客观量化学习现状并提供诊断反馈的测评模式日益受到重视。相比于横断认识诊断测评，纵向认知诊断测评更有利于实现促进学生发展的目标。为使国内学者系统性地了解纵向认知诊断模型，首先，依据建模逻辑将已有纵向认知诊断模型划分为基于潜在转换分析的和基于高阶潜在结构模型的两类，并逐一介绍和说明两类模型的理论基础和应用情景；然后，通过模拟研究为读者呈现如何使用纵向认知诊断模型进行数据分析及如何解读相应的诊断结果。最后，提炼出四个可进一步研究的议题。

A Review of Longitudinal Cognitive Diagnosis Models for Learning Tracking

Individual growth/change has long been an active topic in educational,psychological and behavioral studies.During the last decades,the assessment pattern of objectively quantifying the learning status and providing diagnostic feedback has been increasingly valued,aiming at promoting students’learning,based on the idea of assessment for learning and the characteristics of cognitive diagnostic assessments(CDA).Although most of the existing studies on CDA mentioned that the main objective of CDA was to identify students’strengths and weaknesses in their learning status and to provide guidance for targeted remedial teachings,still few studies have focused on and evaluated the effectiveness of such remedial teachings.The main reason is that the cross-sectional design is currently adopted by most CDAs.This means that only one assessment is done at a specific point in time.This issue may also reflect on the development of existing cognitive diagnosis models(CDMs)which is one of the core components of CDA.Although various CDMs have been suggested by previous researchers,most of them are only applicable to cross-sectional data analysis.The longitudinal data collected from the assessments throughout the learning process provides researchers the chance to develop learning models,that can be adapted to track individual growth over time and to evaluate remedial teachings’effectiveness.Compared to the cross-sectional cognitive diagnostic assessment,the longitudinal cognitive diagnostic assessment is more helpful when aiming at promoting students’development.However,so far,longitudinal CDM is just a new rising topic in foreign countries,and still in its infancy.There are many issues for further study,and the domestic research is almost blank.Therefore,in order to make domestic scholars systematically understanding the longitudinal CDM,several longitudinal CDMs are reviewed in this article.We first divide the existing longitudinal CDMs into two types according to the modeling logic:one is based on the latent transition analysis and another is based on the higher-order latent structural model.Then,the theoretical basis and application scenarios of some representative models are introduced and explained one by one,including Li,Cohen,Bottge,&Templin(2016),Kaya and Leite(2017),Wang,Yang,Culpepper,and Douglas(2018),Huang(2017),and Zhan,Jiao,Liao,and Li(2019).Furthermore,a simple simulation study is conducted to present how to use the longitudinal CDM for data analysis and how to interpret corresponding diagnostic results,which is of certain reference value to practitioners.Finally,four future research topics are concluded,(a)Systematic comparison between different longitudinal CDMs,which can provide theoretical suggestions for practitioners to choose suitable models;(b)Incorporating process data or biometrical data into current longitudinal CDMs,which can provide theoretical support for using multivariate data to assist the evaluation of student development;(c)Extending current longitudinal CDMs to handle polytomous attributes and probabilistic attributes,since polytomous attributes and probabilistic attributes can describe the growth of students in a more refined way than binary attributes;(d)Compared with cross-sectional CDAs,the diagnostic accuracy and validity of longitudinal CDA that used to depict the learning or growth trajectories are more worthy of attention by researchers and practitioners.In addition to choose a suitable longitudinal CDM,many factors such as the quality of the longitudinal test itself,the setting of a cognitive model,students’response attitude,cheating and missing data will also affect the accuracy and validity of the diagnostic results.The influence degree of these factors on the longitudinal diagnosis and the corresponding compensation or detection methods are also worth further discussion.