Anmol Madan, Alex 'Sandy' Pentland

Abstract

As citizens of the information age, we are leaving pervasive digital traces of our idiosyncratic behavior– in emails, on online social networking sites, in mobile phone call logs, in ATM machines, in metropolitan train systems. Researchers in the emerging field of Computational Social Science (CSS) are attempting to create quantitative models of large-scale human social systems from these digital traces, as well as address key questions like privacy, data-ownership and sharing of interaction data. Recent work has focused on online techniques, e.g. mining corporate emails and crawling social networking sites to build maps of social interactions. In contrast, mass-market mobile phones allow sensing and analysis of the building blocks of social interactions – those that occur face-to-face. Eagle & Pentland (2005) and later Dong & Pentland (2006) used Bluetooth proximity information to recognize social patterns in daily activity infer relationships, identify socially relevant locations and model organizational rhythms. In this paper we describe a mobile platform accessible to CSS researchers, devised by augmenting a commodity mobile phone with the ability to mine face-to-face interactions and user activity patterns over long-term durations. As an example of its potential, we also describe in-progress work to study social influence, viral diffusion of media, and the correlation between social ties and privacy