Classifying the difficulty levels of working memory tasks by using pupillary response.

Area of Science:

• Cognitive Science
• Biomedical Engineering
• Educational Technology

Background:

• Task difficulty assessment is vital for optimizing learning outcomes.
• Pupillary response data offers a potential, non-invasive measure of cognitive load.
• Inter-subject variability in pupil responses presents a significant challenge for classifier development.

Purpose of the Study:

• To classify the difficulty level of memorization tasks using pupillary response data.
• To investigate the efficacy of various machine learning classifiers for this task.
• To identify key pupillometric features indicative of task difficulty.

Main Methods:

• Collected eye-tracking data from students performing low-, medium-, and high-difficulty memorization tasks.
• Performed statistical analysis on pupillometric features like peak dilation and pupil diameter change.
• Employed a wrapper method to select optimal features for Support Vector Machine (SVM), Decision Tree (DT), Linear Discriminant Analysis (LDA), and Random Forest (RF) classifiers.

Main Results:

• Significant statistical differences were observed in pupillometric features across different task difficulty levels.
• A Random Forest classifier, trained with five selected features, achieved the highest F1-score of 82%.
• The study demonstrates the potential of pupil size features for difficulty classification.

Conclusions:

• Pupillary response data can be effectively utilized to classify memorization task difficulty.
• The developed method offers a promising approach for evaluating cognitive load non-invasively.
• This research contributes to personalized learning and adaptive educational systems.