An architecture for collecting longitudinal social data

As social computing reaches maturity, two research problems crystallize. On one hand, longitudinal studies of social media interactions become necessary for understanding, for example, how information and behavior diffuses through social networks. On the other hand, accurate representation of a person´s social sphere requires aggregation of multiple sources of social data, such as online social networks, and interactions from email or instant messaging. This paper presents a platform for the collection, extraction and analysis of dynamic social data from multiple sources for use in observational social network research. Using an implementation of the architecture, we made daily observations of the cheating behavior of over 9 million gamers in a planetary scale online social network, finding new evidence that the spread of unethical behavior follows a contagion process and that a social penalty exists for publicly outed deviant actors.