Assessing Personality Through Objective Behavioral Sensing

Traditional personality assessment techniques often rely on subjective report obtained from questionnaires. This work complements traditional techniques by exploring objective measures of traits at the behavior level. We explored behavior features extracted from smartphone sensing data, and used selected features to predict the traits of the Five Factor Model. The specific dataset we explored was the StudentLife dataset. We found behavior features corresponding to each trait, and were able to predict the traits with varying degrees of accuracy. The best result of each trait are: Extraversion (91.2%), Agreeableness (67.6%), Conscientiousness (70.6%), Neuroticism (79.4%), Openness(73.5%). Our results suggest that behavioral measures extracted from smartphone sensing data has potential in the assessment of personality