The PI's efforts to conduct original and practical research in socially intelligent computing - an emerging and important paradigm centered on integrating people and computers to create new forms of collaboration, communication, and intelligence previously unachievable by humans or computers alone - have been hindered, in scope, scale and quality, by the lack of a dedicated and realistic infrastructure. This proposal requests funds to set up such an infrastructure at the PI's institution, which will support integrated research and education. The infrastructure requested includes high-end computational and storage servers, desktop machines, laptops, smart phones, sensors, cameras, software and accessories for collecting, processing and extracting knowledge from large scale data arising from the daily interactions of society with the Internet and with mobile phones. The overall goal is to utilize the knowledge gained from social computing data to create a spectrum of practical services and applications benefitting society. In particular, three research projects that emphasize the close integrations of society with technology are identified in the proposal: a) Detection of depressive disorders in college settings by mining Internet usage data; b) Human "fingerprinting" by mining Internet and smart phone usage; and c) Tracking humans in the social world by fusing heterogeneous sensor data.
Intellectual Merit The planned research activities are well described and will likely significantly advance the state of the art in socially intelligent computing. The PI has pioneered the mining of real Internet data to detect depressive behavior in college students. His prior research has identified critical Internet usage features that show strong statistical differences between students with and without depressive symptoms. He next plans to design, using computational intelligence techniques, classifiers which can proactively detect depressive behavior in college students with high accuracy while being transparent and preserving privacy. He is also exploring the feasibility of mining Internet usage patterns to fingerprint humans, with applications to Internet forensics and mitigation of insider attacks. Similar techniques will be applied to mine sensor data from smart phones in order to fingerprint mobility patterns and to lay the foundation for a variety of pervasive services. While conventional tracking algorithms leverage either a network of cameras or physical sensory data or electronic signals, the PI plans to pursue an integrated approach that fuses multiple orthogonal data source and which incorporates novel feature extraction and pattern recognition techniques for human tracking in both outdoor and indoor environments.
Broader Impact This project has applications in diverse areas including mental health screening, insider attack and fraud prevention, phone and vehicle theft detection, participatory sensing etc. Research outcomes will be shared periodically with diverse stakeholders in psychology, law enforcement, forensics, business, etc. The courses taught by the PI and his team in networking, security and computer vision will be enhanced with content deriving from this project, and the infrastructure will help students learn by practical experience. Research findings, learning materials and team experiences will be disseminated periodically to a wide audience (including educators and students in HBCUs and K-12) via conferences and the Web.
