Online ocular measurement of cognitive load

Abstract: Five different ocular measurements were investigated as a means of predicting cognitive load during a visuospatial memory game. Eye data was collected from the participants during a non-related n-back task, trained on both classification and regression models, and investigated in relation to their ability to make predictions regarding how the participants were performing on an n-back task. All the algorithms performed well in their ability to make predictions on the data and a second task was created to examine how stable the pre-dictions were over different tasks. During the second task, a number of participants from the first part completed a visuospatial memory game while information about their ocular reactions was collected. The information was continuously sent to the learning algorithms previously investigated, making predictions about the participants’ performance online. Based on the predictions, changes were made in the level of difficulty of the game. The results of the study show that it is possible to use a combination of several different ocular measures, collected from several participants during one task, to predict the performance of individual agents during a second task. The tasks in this study were not related and the data used for prediction were not individually adapted, in contrast to previous studies.