Li Li, An Yi, Ren Jie, Wei Xiaoman
Collaborative Innovation Center of Assessment for Basic Education Quality, Beijing Normal University, Beijing, China.
Institute of Language Testing and Talent Evaluation, Faculty of Linguistic Sciences, Beijing Language and Culture University, Beijing, China.
Front Psychol. 2021 Sep 7;12:714568. doi: 10.3389/fpsyg.2021.714568. eCollection 2021.
As a new generation of measurement theory, cognitive diagnosis theory shows significant potential and advantages in educational evaluation in that it combines a cognitive process and a measurement method. The application of the theory not only reveals the potential characteristics of learners in cognitive processing, but also provides targeted remedies and strategic guidance for individuals. Given the difficulties of traditional assessment models in providing an insightful and fine-grained account for individualized and procedural learning, providing personalized learning strategies for learners of Chinese as a second language has been a new goal of teaching and measurement in Chinese listening. This study constructs a cognitive diagnosis model of Chinese listening comprehension for Chinese-as-a-second-language learners through theoretical exploration, model hypotheses, repeated verification, and model modification. The results show that the Q-matrix (Q) constructed by the experts within modification has the highest fitting degree with the empirical data. The parameter recovery rate, the accuracy of the tested attribute or mode, and the relative fitting index obtained from the simulation study are consistent with the information extracted from the empirical data. The diagnostic reliability and effectiveness of generalized deterministic inputs, noise "and" gate (G-DINA) are higher than those of DINA, deterministic inputs, noisy "or" gate (DINO), and reduced reparametrized unified model (RRUM). In the estimation of the item and subject parameters, the G-DINA model shows good convergence, and the average classification accuracy rate based on attribute level is 0.861.
作为新一代测量理论，认知诊断理论在教育评价中展现出显著潜力和优势，因为它将认知过程与测量方法相结合。该理论的应用不仅揭示了学习者在认知加工中的潜在特征，还为个体提供了针对性的补救措施和策略指导。鉴于传统评估模型难以对个性化和程序性学习进行深入且细致的描述，为汉语作为第二语言的学习者提供个性化学习策略已成为汉语听力教学与测量的新目标。本研究通过理论探索、模型假设、反复验证和模型修正，构建了汉语作为第二语言学习者汉语听力理解的认知诊断模型。结果表明，经修正后由专家构建的Q矩阵（Q）与实证数据的拟合度最高。模拟研究得到的参数恢复率、测试属性或模式的准确性以及相对拟合指数与从实证数据中提取的信息一致。广义确定性输入、噪声“与”门（G-DINA）的诊断可靠性和有效性高于确定性输入、噪声“或”门（DINA）、确定性输入、噪声“或”门（DINO）以及简化重新参数化统一模型（RRUM）。在项目和主体参数估计中，G-DINA模型表现出良好的收敛性，基于属性水平的平均分类准确率为0.861。
