Personality traits prediction based on eye movements while reading manga

Previous studies utilizing machine learning methods have demonstrated that personal traits can be predicted from eye movement data recorded in real-world situations, such as navigating a university campus or browsing one's Facebook news feed. The objective of this study was to conceptually replicate and extend these findings in a different type of visual engagement. Specifically, we aimed to predict individuals' personal traits using eye movements that reflect gaze behaviors while reading manga (Japanese comics).

We recorded the eye movements of 51 participants as they read manga and trained several machine learning classifiers to predict the levels of each of the self-reported Big Five personality traits from the eye movement features extracted from their reading behavior. The models' performance was evaluated using cross-validation, and the SHapley Additive exPlanation (SHAP) approach was employed to elucidate the classification model by identifying important features and their impacts on the model output.

Among the Big Five personality traits, only extraversion was predictable. The evaluation results demonstrated that the best model achieved comparable performance with previous literature, with a macro F1 score of 0.49. Analysis of the SHAP value plots showed that a high fixation rate, pupillary response, and blink rate were informative indicators.

The results partially replicated the previously noted associations between eye movement and personality traits. We found that gaze behaviors observed during reading manga are informative of an individual's extraversion personality trait. We also point out several potential advantages of using manga for gaze-based personality detection.