Virtual reality-based assessment and training platforms proffer the potential for higher-dimensional stimulus presentations (dynamic; three dimensional) than those found with many low-dimensional stimulus presentations (static; two-dimensional) found in pen-and-paper measures of cognition. Studies have investigated the psychometric validity and reliability of a virtual reality-based multiple errands task called the Virtual Environment Grocery Store (VEGS). While advances in virtual reality-based assessments provide potential for increasing evaluation of cognitive processes, less has been done to develop these simulations into adaptive virtual environments for improved cognitive assessment. Adaptive assessments offer the potential for dynamically adjusting the difficulty level of tasks specific to the user’s knowledge or ability. Former iterations of the VEGS did not adapt to user performance. Therefore, this study aimed to develop performance classifiers from participants (N = 75) using three classification techniques: Support Vector Machines (SVM), Naive Bayes (NB), and k-Nearest Neighbors (kNN). Participants were categorized as either high performing or low performing based upon the number items they were able to successfully find and add to their grocery cart. The predictors utilized for the classification focused on the times to complete tasks in the virtual environment. Results revealed that the SVM (88% correct classification) classifier was the most robust classifier for identifying cognitive performance followed closely by kNN (86.7%); however, NB tended to perform poorly (76%). Results suggest that participants’ task completion times in conjunction with SVM or kNN can be used to adjust the difficult level to best suit the user in the environment.
