基于CD-CAT的多策略RRUM模型及其选题方法开发

在有多种解题策略的认知诊断问题情境中, 用每个Q矩阵表示一种解题策略, 由此将单策略认知诊断RRUM模型拓广为多策略RRUM模型(MS-RRUM)。随后, 在应用MS-RRUM模型的CD-CAT中开发了适用于多策略情境的MAP参数估计法和多策略香农熵(MSSHE)选题法。将MSSHE选题法与随机选题法分别在不同属性数量、不同测验长度下进行比较, 结果发现前者对被试的策略和认知状态判准率都显著优于后者, 而且都很理想。这样就顺利实现了在CD-CAT做策略诊断的目标。

Item Selection Using the Multiple-Strategy RRUM Based on CD-CAT

Cognitive diagnostic computerized adaptive testing (CD-CAT) is an application of cognitive diagnostic testing (CDT) on the computerized adaptive testing (CAT) platform. CD-CAT can provide information on the specific content areas in which a participant needs the most specific help, and it is typically engineered to tailor the test to each participant’s trait level and thus has high efficiency. The search for the best item selection method is currently one of the most pressing issues in the field of CD-CAT research. Besides, studies on CD-CAT have to address the issue on the number of testing strategies to employ: either all participants use the same strategy, or different participants use different strategies. In practice, for some tasks, all participants use the same strategy; whereas for other tasks, different participants use different strategies to solve the problem. Because cognitive strategy aptitudes (or cognitive structures) vary among participants, each strategy requires different attributes. For the latter type of tasks, a participant’s strategy must be identified before his/her knowledge state (KS, i.e., attribute mastery pattern) can be diagnosed. Strategy diagnosis procedures have not been explored in studies on CD-CAT.