Information

student.uva.nl has been renewed. Read more...

What is your study programme?
Talk

Presentation Master's Thesis - Michiel Cassee - Brain & Cognition

Last modified on 23-08-2022
Predicting experienced cognitive load with an EEG-based classification model trained on working memory, a preliminary cross-task validation
Show information for your study programme
You're currently viewing general information. Choose your study programme to see additional information that's specific to your study programme, such as deadlines, regulations and contact details.
What is your study programme?
Start date
29-08-2022 08:00
End date
29-08-2022 10:00
Location

Roeterseilandcampus - Building G

Street:

Nieuwe Achtergracht 129-B

Room: 

S.04

The current study was performed to determine the applicability of a predictive model for UX research. The model was built as a convolutional neural network to classify the experienced cognitive load during task performance. To achieve this, brain activity was measured using an electroencephalogram (EEG). With this data, the model produced a cognitive load prediction output, representing the level of cognitive load during a trial. Beforehand, the model got trained on a working memory task individually. 

Besides the working memory task, the experiment included a visual search task and a reading task. A cross-task validation was performed to test the generalizability of the model to measure overall cognitive load. The validation was assessed based on the sensitivity of the model. Both experimental tasks consisted of an easy (low load) and a hard (high load) condition. The cognitive load predictions were tested for significant differences between both conditions, for each experimental task. With this, the model’s capability of distinguishing load between conditions was tested.

A within-subject analysis was performed in both experimental tasks separately. No significant differences in cognitive load prediction between conditions were found. This outcome indicates insufficient sensitivity of the model. Moreover, the statistical power was low. Concluding, the model could not be cross-task validated based on the current study. Since no results in favor of  the generalizability were found, applying the current model in UX research is not recommended. Instead, future research should focus on replicating the current cross-validation with more participants and possibly other (training) tasks.