People connect each other on Social Network Sites (SNSs, e.g. Facebook) and such behavior may indicate relationship among them. This project analyzes the behavior on a company-internal social network site to determine which interaction patterns signal closeness between colleagues. A multi-linear regression model is applied, with the server-side log data is used as independent variables and users' subjective rating of their closeness is used as ground truth (see a video of the game designed to collect the subjective data). Regression analysis suggests that employee behavior on social network sites (SNSs) reveals information about both professional and personal closeness. While some factors are predictive of general closeness (e.g. content recommendations), other factors signal that employees feel personal closeness towards their colleagues, but not professional closeness (e.g. mutual profile commenting). This analysis contributes to our understanding of how SNS behavior reflects relationship multiplexity: the multiple facets of our relationships with SNS connections.

This is the work I have done as a summer intern at IBM Waston, 2009.

Predictors that distinguish professional closeness from personal closeness

Related Publications

Wu, A. DiMicco, J.M., Millen, D. R. (2010). Detecting Professional versus Personal Closeness Using an Enterprise Social Network Site. Proc. of CHI'10, 1955-1964. PDF

back to Home