16:00 - 16:15 | Thu 10 Nov | Maya I - II | ThBT1.2
One in five Americans is lonely and loneliness disproportionately affects senior citizens and international students. In this paper, we propose Socialoscope, a smartphone app that passively senses user loneliness from their communication and interaction patterns (e.g. calls, SMS, browsing patterns and social media usage), while factoring in different personality types. Data was gathered from 9 international students over 2 weeks to train machine learning classifiers for loneliness. Using smartphone-sensed data, we show that of the big 5 personality traits, extraversion and emotional stability features were strongly correlated with smartphone-sensed loneliness. We synthesized machine learning classifiers that classified user smartphone interaction and communication features into ranges of loneliness with an accuracy of 98%, while factoring in user personality types.