I was recently drawn towards the growing use of mobile phone sensors and online social networks in understanding social behaviors. The Human Dynamics Group at MIT Media Labs seems to have a number of interesting projects on this topic. The following papers are quite related to things we discussed in the class and are very interesting to read:
Describes a mobile-phone-based social and behavioral sensing system and some initial results on how to use them to study the connection between individuals’ social behavior and their financial status, network effects in decision making. The implementation of mobile sensing is of particular interest: they develop a very easy to use open source framework called funf with which app developers can easily develop context sensing apps. The results are mostly preliminary: One shows that there is correlation between the common apps downloaded and the people you hang out with.
This paper uses the framework from the paper above to constructed a SMS messages social network using info about more than 97,000 SMS messages sent. The aim is to predict each participants’ personal information by using their friends personal information. The results look quite impressive; the prediction accuracy for many attributes like religion, ethnicity, origin, age, etc. seems to be above 0.95. The title of the paper comes from their study on how the prediction accuracy is increased when the users sample set grows.