Assessing personality using demographic information from social media data

Abstract

Personality prediction has been widely studied in various fields including psychology and social media. The emergence of online social networks provides novel opportunities to obtain information about user behaviors, demographics and administer standardized surveys for studying such social and psychological phenomena in a data-driven manner. Using the data generated from a personality survey application on Facebook used by 228,343 users, this work undertakes a statistical validation of a frequently posited hypothesis: does a person's personality have any relation with their demographic information such as gender and date of birth (DOB)? Literature on astrology - though rarely scientifically tested - suggests that the one's personality is related to their sun sign which is derived from their DOB. Based on DOB we have categorized all users into 12 groups referred as 12 sun signs. We then examined the relationship between users' big five personality traits and their sun sign through Chi-square (χ2) test. Subsequently, the ordinal regression model is used to model this relationship. Our results suggest that there are certain personality traits which relate to DOB, making DOB a very quick indicator of one's personality where there exists a relationship. Interestingly, we also analyzed the impact of users' belief in sun signs and found that the degree of confidence increased for the users who believe in sun signs.