Toward Harmonizing Self-reported and Logged Social Data for Understanding Human Behavior

Arushi Jain

Abstract: While self-reporting remains the most common method to understand human behavior, recent advances in social networks, mobile technologies, and other computer-mediated communication technologies are allowing researchers to obtain detailed logs of human behavior with ease. While the logged data is very useful (and accurate) at capturing the structure of the user's social network, the self-reported data provides an insight into the user's cognitive map of her social network. Based on a field study involving 47 users for a period of ten weeks we report that combining the two sets of data (self-reported and logged) gives higher predictive power than using either one of them individually. Further, the difference between the two types of values captures the level of dissonance between a user's actual and perceived social behavior and is found to be an important predictor of the person's social outcomes including social capital, social support and trust.