| Cognitive overload significantly impacts human performance in complex interaction settings, making its early detection essential for the design of adaptive systems. This paper investigates whether gaze-derived features can reliably predict overload states using supervised machine learning (ML). The analysis is based on an eye-tracking dataset collected during cognitively demanding visual tasks, incorporating fixations, saccades, and pupil diameter measurements. Five classifiers, Logistic Regression (LR), Naive Bayes (NB), Support Vector Machine (SVM), XGBoost (XGB), and Multilayer Perceptron (MLP), were evaluated using stratified training and testing splits, alongside 5-fold cross-validation, to identify the presence or absence of cognitive overload. Among them, XGB achieved the highest performance, with an accuracy of 0.902, a precision of 0.958, a recall of 0.821, an F1-score of 0.884, and an area under the ROC curve (AUC) of 0.956. The findings confirm that gaze-derived features alone can reliably distinguish cognitive overload states. The study also highlights trade-offs between model interpretability and predictive performance, with ensemble methods, such as XGB, offering superior results, which support their use in attention-aware systems. Future directions include personalization, temporal modeling, cross-task generalization, and the integration of adaptive feedback mechanisms.
|
| M. Trigka, E. Dritsas, Ph. Mylonas, "Eye-Based Cognitive Overload Prediction in Human-Machine Interaction via Machine Learning", 21st International Conference on Web Information Systems and Technologies (WEBIST 2025), Marbella, Spain, October 21-23, 2025 |
PDF] [
BibTex] [
Print] [
Back]