Predicting neuroticism with open-ended response using natural language processing.

Area of Science:

• Psychology
• Natural Language Processing
• Computational Linguistics

Background:

• Personality prediction is increasingly researched using natural language processing (NLP).
• The choice of questions significantly influences the quality of natural language responses for personality assessment.
• Neuroticism, a key personality trait, impacts various psychological outcomes.

Purpose of the Study:

• To predict neuroticism levels using responses to open-ended questions based on the Five-Factor Model.
• To evaluate the accuracy of NLP models in personality prediction.
• To investigate how the content of questions influences the prediction of neuroticism.

Main Methods:

• Recruited 425 Korean adults to answer 18 open-ended personality questions.
• Collected 30,576 Korean sentences for analysis.
• Utilized the KoBERT pre-trained language model for prediction and calculated Accuracy, F1 Score, Precision, and Recall.

Main Results:

• Questions about social comparison, unintended harm, and negative feelings showed higher accuracy in predicting neuroticism.
• Items related to negative feelings, social comparison, and emotions were superior for predicting depressivity.
• Items concerning unintended harm, social dominance, and negative feelings were most predictive for dependency.

Conclusions:

• Identified specific question types that enhance neuroticism prediction accuracy.
• Demonstrated that prediction models based on theoretically aligned open-ended questions yield superior performance in personality assessment.