Personal Space and Automatically Learned Social Networks

February 22, 2011
Carson Reynolds | University of Tokyo

Even networks of very simple sensors, when the data is subjected to sufficient analysis, can reveal deep and subtle information about the behaviors of people moving around them. Social networks can be automatically learned and updated by applying machine learning to sensor network data. This talk describes a social network that is automatically learned from a sensor network which has been installed in an office environment. Describing some recent work conducted in collaboration with colleagues at MERL, this talk will cover two aspects of the social networking system. Firstly, I’ll compare its performance relative to self-reported social connections. Secondly, I’ll analyze questionnaire feedback regarding the privacy implications of automatically-generated social networks to understand how people perceive these new technologies. The talk will show that the automatic social network learning system was able to estimate 86% of social connections relative to ground truth drawn from survey data with only 0.6% false positives. The questionnaire was distributed to users familiar with the system and Internet respondents who were unfamiliar with the system. The research found that these respondents feel their privacy is invaded by many possible methods of gathering social network information. Those who had used the automatic social network learning system, however found that motion sensor logs were less invasive than other possible observation channels.

Speaker Details

Carson Reynolds is a project assistant professor in the Department of Creative Informatics of the University of Tokyo. He is co-founder of the Meta-Perception research group which investigates methods for capturing and manipulating information that is normally inaccessible to humans and machines. His research interests include sensor systems, privacy, and roboethics. Carson’s work has been discussed (online) Wired, Make, Engadget, New Scientist, Robots Podtcast, Smart Mobs, and Slashdot (in print) Boston Globe, Washington Times, the German weekly Focus and (broadcast) on the US’s National Public Radio, The Discovery Channel’s Daily Planet and Nippon TV World’s Best Lectures. He has also created and organized the Devices that Alter Perception Workshop which was held at ACM Ubicomp, the International Association for Computing and Philosophy, and the IEEE International Symposium on Mixed and Augmented Reality. He has designed a skin conductance sensor which has developed to become the first product of Affectiva, a new MIT startup.