Exploring a Machine-Learning Approach to Assess the Construct and Criterion-Related Validity of Game-Based Cognitive Ability Assessment: A Secondary Study
Abstract
Game-based assessments (GBAs) are increasingly applied in various fields, making studying their validity crucial. Traditional methods of testing the validity of psychological assessments, such as structural equation modeling and correlation analysis, rely on statistical assumptions that real-world data often violate. Thus, those traditional approaches might not be suitable for GBAs due to the amount and complexity of GBAs data. A machine learning approach, emphasizing accurate predictions and requiring few assumptions of data, might offer an alternative to studying the validity of GBAs. This study explores a machine learning approach to assess the psychometric properties of Cognify, a game-based cognitive ability assessment, using secondary data from a study conducted by Landers et al. (2021). We used machine learning techniques like random forest regression models and variable importance calculation to support the predictive and construct validity of the GBA. The results showed the promising potential of utilizing a machine-learning approach to solve psychometric problems in game-based assessments.
Subject Area
Psychology|Artificial intelligence|Cognitive psychology
Recommended Citation
Deng, Xinyue, "Exploring a Machine-Learning Approach to Assess the Construct and Criterion-Related Validity of Game-Based Cognitive Ability Assessment: A Secondary Study" (2023). ETD Collection for Fordham University. AAI30631468.
https://research.library.fordham.edu/dissertations/AAI30631468
