Classifying creativity: Applying machine learning techniques to divergent thinking EEG data.

Area of Science:

• Neuroscience
• Cognitive Psychology
• Artificial Intelligence

Background:

• Prior research links higher electroencephalography (EEG) alpha power (8-13 Hz) to increased creativity in individuals and tasks.
• Creative thinking involves divergent processes that may be reflected in distinct neural activity patterns.

Purpose of the Study:

• To investigate the feasibility of using machine learning (ML) to classify brain states associated with varying levels of creativity.
• To determine if ML can reliably distinguish between brain activity during common versus uncommon uses of objects.
• To assess ML's ability to differentiate between individuals with higher and lower creative potential.

Main Methods:

• Participants engaged in an Alternate Uses Task, generating normal or uncommon uses for everyday objects.
• EEG data was recorded, focusing on alpha power (8-13 Hz) during task performance.
• Spectrally weighted common spatial patterns were employed for EEG feature extraction.
• Quadratic discriminant analysis was utilized for classification of brain states and individuals.

Main Results:

• Alpha power was significantly higher during the generation of uncommon (more creative) uses compared to normal uses.
• ML achieved a mean classification accuracy of 63.9% for distinguishing between common and uncommon use conditions.
• ML demonstrated higher classification accuracy (82.3%) in differentiating between more and less creative individuals.

Conclusions:

• EEG alpha power reliably differentiates creative task conditions and can be used to classify creative brain states.
• Machine learning offers a promising tool for objective assessment and classification within creativity research.
• These findings support the broader application of ML in understanding the neural underpinnings of human creativity.