Cognitive Modeling of Learning Using Big Data From a Science-Based Game Development Environment

Abstract

The purpose of this study was to identify the underlying cognitive attributes used during the design and development of science-based serious educational games. Study methods rely on a modification of cognitive diagnostics, item response theory, and Bayesian estimation with traditional statistical techniques such as factor analysis and model fit analysis to examine the data and model structure. A computational model of the cognitive processing using an artificial neural network (ANN) allowed for examination of underlying mechanisms of cognition from a server-side data set and a 21st century skills assessment. ANN results indicate that the model correctly predicts successful completion of science-based serious educational game (SEG) design tasks related to 21st century skills 86% of the time and correctly predicts failure to complete SEG design tasks related to 21st century skills 78% of the time. The model also reveals the relative importance of each particular cognitive attribute within the 21st century skills framework.