Daniel Chapsky

Div 3 Abstract :

Over the past decade, people have been expressing more and more of their personalities online. Online social networks such as Facebook.com capture much of individuals’ personalities through their published interests, attributes and social interactions. Knowledge of an individual’s personality can be of wide utility, either for social research, targeted marketing or a variety of other fields A key problem to predicting and utilizing personality information is the myriad of ways it is expressed across various people, locations and cultures. Similarly, a model predicting personality based on online data which cannot be extrapolated to “real world” situations is of limited utility for researchers. This paper presents initial work done on generating a probabilistic model of personality which uses representations of people’s connections to other people, places, cultures, and ideas, as expressed through Facebook. To this end, personality was predicted using a machine learning method known as a Bayesian Network. The model was trained using Facebook data combined with external data sources to allow further inference. The results of this paper present one predictive model of personality that this project has produced. This model demonstrates the potential of this methodology in two ways: First, it is able to explain up to 56% of all variation in a personality trait from a sample of 615 individuals. Second it is able to clearly present how this variability is explained through findings such as how to determine how agreeable a man is based on his age, number of Facebook wall posts, and his willingness to disclose his preference for music made by Lady Gaga.

http://www.thebestpictureontheinternet.com/