Predicting attachment style from EEG data on the Flanker task.

Area of Science:

• Psychology
• Neuroscience
• Data Science

Background:

• Bowlby's attachment theory explains individual differences in emotional bonding.
• Attachment styles are traditionally assessed via self-report questionnaires, which can be subjective.
• Neurophysiological methods are emerging to explore the neural basis of attachment, but predictive models are lacking.

Purpose of the Study:

• To develop and validate a predictive model for inferring individual attachment profiles from electroencephalogram (EEG) data.
• To provide an objective, data-driven alternative to subjective self-report measures of attachment.
• To explore the utility of advanced machine learning algorithms for analyzing complex neurophysiological data.

Main Methods:

• Recorded EEG data from 27 participants performing the Flanker task with feedback.
• Utilized the ROCKET algorithm to extract 20,000 time-series features from EEG signals.
• Applied Principal Component Analysis (PCA) for dimensionality reduction and constructed regression models to predict attachment scores (anxiety and avoidance) from the ECR-R questionnaire.

Main Results:

• Demonstrated the feasibility of inferring individual attachment profiles solely from EEG data.
• Successfully categorized participants into the four canonical attachment styles based on their brain signals.
• Showcased the effectiveness of automated feature extraction (ROCKET) over traditional hand-crafted features.

Conclusions:

• Individual attachment styles can be objectively inferred from EEG brain signals.
• This EEG-based approach offers a less subjective alternative to traditional attachment assessment methods.
• Automated feature extraction significantly enhances the predictive power of neurophysiological data for attachment research.