EEG-Based Machine Learning Models to Evaluate Haptic Delay: Should We Label Data Based on Self-Reporting or Physical...

Area of Science:

• Human-Computer Interaction
• Neuroscience
• Machine Learning

Background:

• Haptic interfaces enhance user experience through simulated feedback.
• Evaluating user experience in haptics is crucial for technological advancement.
• Traditional self-reporting methods for haptic feedback evaluation can be unreliable.

Purpose of the Study:

• To compare the robustness of self-report versus physical stimuli labeling for training machine learning models on Electroencephalography (EEG) data.
• To investigate the impact of data labeling methods on the accuracy of neurocognitive evaluation of haptic experiences.
• To examine the role of data labeling in evaluating haptic delay detection using EEG.

Main Methods:

• Development of a visuo-haptic task to study haptic delay.
• Recording Electroencephalography (EEG) data during the task.
• Training four machine/deep learning models twice: once with self-report labels and once with physical stimuli labels.

Main Results:

• Models trained with physical stimuli labels significantly outperformed models trained with self-report labels.
• The findings demonstrate a clear advantage of objective labeling over subjective self-reporting for EEG-based haptic experience evaluation.
• The study identified data labeling as a critical factor in the success of neurocognitive evaluation methods.

Conclusions:

• Objective labeling based on physical stimuli provides a more robust foundation for training machine learning models to evaluate haptic experiences compared to subjective self-reports.
• The choice of data labeling strategy is paramount for the effective application of neurocognitive methods in assessing human-computer interaction.
• Future research should focus on optimizing data labeling techniques for diverse haptic feedback evaluations.