Reality Commons

brought to you by the MIT Human Dynamics Lab

Social Evolution Dataset - Publications and Findings

Sensing the 'Health State' of a Community, A. Madan, M. Cebrian, S. Moturu, K. Farrahi, A. Pentland, Pervasive Computing, Vol. 11, No. 4, pp. 36-45 Oct 2012

Dong and colleagues [Dong et al., 2011] showed that subjects chose friends according to how often they see one another, and influence one another in participating in physical exercises, student activities, dietary behaviors, and others. They further proposed a stochastic process model to capture the co-evolution of social relationships and individual behavior with 12 events. They showed that such a stochastic process model could help researchers to interpolate between surveys with sensor data, and that it could potentially anonymize the "human behavior trails" through a resampling.

Dong and colleagues [Dong et al., 2012a] showed that it is possible to fit agent based models with the "data exhaust trails" of networked people by identifying agent based models as stochastic processes, through a case study that tells who infects whom in real-world social network.

Madan and colleagues were able to track stress, sadness, and flu by looking at how the subjects moved around, how much they talked to the others, and when they talked to the others [Madan et al., 2010a]. They were also able to explain the change in weight of a subject by the weight changes of his contacts on a weekly basis [Madan et al., 2010b].

Madan and colleagues inspected the political opinions and discussions of the subjects during the 2008 presidential election. They found that people tended to be close to those with the same political opinions during the presidential debates using Bluetooth scanning [Madan et al., 2011].