Neural sources of prediction errors detect unrealistic VR interactions

Area of Science:

• Neuroscience
• Human-Computer Interaction
• Virtual Reality

Background:

• Neural interfaces offer implicit user experience tracking in virtual and augmented reality (VR/AR).
• Assessing immersion in VR traditionally uses subjective questionnaires, which disrupt the user experience.
• Developing objective metrics is crucial for seamless VR/AR experience evaluation.

Purpose of the Study:

• To present a complementary metric for assessing VR user experience using neural signals.
• To investigate the use of electroencephalography (EEG) and prediction error negativity for detecting disruptions.
• To validate a novel approach for measuring immersion without interrupting the VR experience.

Main Methods:

• A reach-to-tap paradigm was designed in VR to elicit visuo-haptic glitches.
• Electroencephalography (EEG) was used to record brain activity.
• Prediction error negativity features were extracted from EEG data to classify glitches.

Main Results:

• VR glitches were classified with 77% accuracy using prediction error negativity.
• Source localization identified midline cingulate and parieto-occipital EEG sources as key to classification.
• The findings demonstrate the reliability of EEG-based prediction error signatures.

Conclusions:

• Prediction error signatures in specific EEG sources serve as a robust marker for disrupted user predictions in VR/AR.
• This neural interface metric offers a non-intrusive method for adaptive user interface development.
• The study highlights the potential of EEG for real-time assessment of user experience in immersive environments.