Development of a Multidimensional Cognitive Ability CAT Using Neural Network Graded Modeling

• In Event: Virtual Poster Session
  In Virtual iPresentation Only: Division D - Section 1: Measurement, Psychometrics, and Assessment Virtual Poster Session

Abstract

This study developed a multidimensional self-report cognitive ability assessment based on the Cattell-Horn-Carroll (CHC) theory. To overcome limitations of existing self-report tools, such as incomplete dimensional coverage and low efficiency, we constructed an item bank of 84 items across 18 dimensions within a Computerized Adaptive Testing (CAT) framework. A neural network-based Multidimensional Graded Response Model was integrated to enhance estimation accuracy and testing efficiency. The resulting system achieved high measurement precision (average RMSE = 0.3281, r = 0.9479) while significantly reducing the number of items administered (average = 27.81). This personalized, theory-driven tool supports efficient and reliable cognitive profiling for both research and applied contexts.

Authors

• Wen Sun, Beijing University of Chinese Medicine

• Zichen Liu, Beijing University of Chinese Medicine

• Wenjie Xu, Beijing University of Chinese Medicine

• Lei Quan, Beijing University of Chinse Medicine

• Zongshuo Sha, Beijing University of Chinse Medicine

• Yixing Liu, Beijing University of Chinese Medicine