Email Network Analysis as an Early Indicator of Organization Stability

 

RONALDO MENEZES

FLORIDA TECH

 

Large social networks tend to be very robust in their structural characteristics. If we look at patterns of email activity in an organization, we expect the network to be kept very stable overtime. This assumption can be confirmed by looking at the fact that the stream of email for people do not change very often except when there are special circumstances. Even when there is such an increase in message flow, the pattern of these messages tend to be the same, meaning that people send messages to the same people they send message every day (on average). The question then becomes whether changes in the network characteristics be serve as an early indicator of unrest in the organization body. In our study we construct a network of the Enron email logs and show that anomalies can be detected at the network level in properties such as clustering coefficient and homophily. What we argue is that these anomalies are an indication of unrest that may lead to problems in the organization. In our study we correlate the anomalies with events in the history of Enron from a time of high moral (when the stock price was very high) to a period of almost desperation (when all employees lost their jobs). The argument in our study is that the anomalies in the network precede the events because people radically change their behavior. For instance, social sciences indicate that in hard times, people tend to cling more to their friends and friends (increase in homophily). Hence one can build tools to detect such changes and provide early warnings to interested parties.