A New Method for Assessing Five-Factor Personality Model Traits: Measuring Responses to Visual Stimuli with an Eye Tracking Device

This study aims to go beyond the commonly employed self-report methods in psychological research by developing a visual inventory designed to assess personality traits through eye movement data. In this study, participants' eye responses to specific visual stimuli (e.g., pupil dilation, average fixation duration, average deviation speed) were used to predict which of the low, medium, or high groups they belonged to in the subdimensions of the Five-Factor Personality Model (FFPM) (McCrae & Costa, 1999). A visual stimulus set of 50 images, each presumed to be associated with specific personality traits, was generated using the artificial intelligence platform Midjourney. These images were evaluated for their associative relevance by clinical psychology experts and later by advanced undergraduate psychology students. The visual stimuli that received the highest ratings were presented to participants in a standardized manner, while eye responses were recorded using an eye-tracking device. Collected data were analyzed using the WEKA machine learning platform with Naïve Bayes, Logistic Regression, J48, and Random Tree algorithms to classify participants into low, medium, and high levels in the dimensions of openness to experience, extraversion, agreeableness, conscientiousness, and neuroticism. Analyzing the classification results revealed that the J48 algorithm achieved an accuracy rate above the random classification baseline for 44 out of the 50 images. Among these, the J48 algorithm yielded the most successful classification performance, achieving accuracy rates above the chance level in 44 out of 50 images. This indicates that eye movement data, when combined with machine learning, can provide meaningful signals for differentiating personality traits. Although further refinement is needed, this non-self-report attempt to assess personality through eye movements may offer valuable insights for future research.