Mining digital data for smarter mental healthcare

Human interactions with Computers (and especially the Internet) are ever increasing, As a consequence, the field of CyberPsychology, which studies the thinking, behavior and attitudes of the person using the computer has assumed immense importance. In this paper, we report our findings and conclusions from several experiments conducted among college students to demonstrate the feasibility of assessing behavior (especially, mood disorders) from monitoring Internet usage. The core novelty of our work is that these experiments are believed to be the first to use real Internet data, collected unobtrusively and preserving a high degree of privacy. Specific contributions are a) Deriving new Internet usage features after mining NetFlow data that are indicative of negative mood; b) Demonstrating statistically strong correlations among many Internet usage features and negative mood; c) Demonstrating the feasibility of designing classification algorithms to predict negative mood and changes in mood from passive Internet usage monitoring.