The exposure to new information and opinions, and their diffusion within social networks, are important questions in education, business, and government. However until recently there has been no method to automatically capture fine-grained social interactions between people and use the data to better model the diffusion process. In this thesis, we describe the use of co-location and communication sensors in ‘socially aware’ mobile phones to measure and model the face-to-face interactions, opinions and behaviors of the residents of an undergraduate dormitory for an entire academic year.
Political scientists have noted (Huckfeldt & Sprague, American Political Science Review, 1983) the problem of mutual causation between face-to-face networks and political opinions. For the last three months the 2008 US presidential campaigns of Barack Obama and John McCain, we find that political discussants have characteristic interaction patterns that can be used to recover the self-reported ‘political discussant’ ties within the network. Automatically measured mobile phone features allow us to estimate exposure to different types of opinions within this community. Based on exposure, we propose a measure of ‘dynamic homophily’ which reveals surprising short-term, population-wide behavior changes around external political events such as election debates and Election Day. To our knowledge, this is the first time such dynamic homophily has been measured. We show that it is possible to use exposure to other nodes in the network to predict future opinions for individuals (r = 0.8, p < 0.0001). We find that using mobile phone based features to calculate dynamic exposure increases the explained variance by up to 30%.
Similarly, it is well known that the main vehicle for contagious diseases is face-to-face networks (Elliott, Spatial Epidemiology: Methods and Applications, 2000). However it has not yet been possible for epidemiologists to quantitatively measure the likelihood of contagion as a function of contact/exposure to infected individuals in realistic scenarios (Musher, The New England Journal of Medicine, 2003). Such research requires simultaneously capturing symptom reports and social interactions between individuals over the long-term. We describe the use of co-location and communication sensors to understand the role of face-to-face interactions in the contagion process. We find that there are characteristic changes in behavior when individuals become sick. These changes are reflected in automatically captured features like their total communication, communication patterns with respect to time of day (e.g., late night, weekends), diversity of their network and in entropy of movement within and outside the university. These behavioral changes can be used to build an inference model to evaluate the likelihood of an individual being sick, based on their interactions and embedding in the network. We use a recently developed signal processing approach (Nolte, Nature 2008) to better understand the causality of signals based on temporal information flux between influenza symptoms, behavioral features and stress and depression reports.
Finally, longitudinal studies indicate that health-related behaviors from obesity (Christakis and Fowler, 2007) to happiness (Fowler and Christakis, 2008) can spread through social ties. The effects of social networks and social support on physical health are well-documented (Berkman, 1994; Marmot and Wilkinson, 2006). However, these studies do not quantify actual face-to-face interactions that might lead to changes in behavior. In our dataset, we study the propagation of these social behaviors and the relative contribution of self-reported relationships versus automatically-captured mobile phone features, and find that mobile phone features are better indicators of future behavior change.
Host/Chair: Alex 'Sandy' Pentland
David LazerTanzeem Choudhury