Cognitive tasks and combined statistical methods to evaluate, model, and predict mental workload.

Area of Science:

• Cognitive Psychology
• Human-Computer Interaction
• Neuroscience

Background:

• Mental workload (MWL) assessment is crucial for user experience and adaptive systems.
• Accurate prediction of MWL is needed to dynamically adjust task complexity.
• Existing cognitive tasks vary in their reliability for modeling and predicting MWL.

Purpose of the Study:

• To identify cognitive tasks that can reliably predict distinct mental workload classes.
• To evaluate the suitability of these tasks for real-time MWL adaptation.
• To compare the efficacy of the Corsi test and N-Back task in modeling MWL.

Main Methods:

• Administered N-Back and Corsi tests with varying complexity levels.
• Measured MWL using NASA-TLX and Workload Profile questionnaires.
• Employed statistical methods and classification algorithms to analyze task performance and predict MWL classes.

Main Results:

• The Corsi test demonstrated three distinct MWL classes with approximately 80% prediction accuracy.
• The Corsi test showed potential for MWL adaptation (over 50% accuracy) but not sufficient for online adjustments.
• Performance indicators alone were insufficient for real-time MWL adaptation, suggesting the need for physiological measures.

Conclusions:

• The Corsi test is a superior candidate for modeling and predicting MWL compared to the N-Back task.
• Current performance-based prediction models require enhancement with physiological data for effective real-time MWL adaptation.
• Further research should integrate multimodal data for robust adaptive systems.