In this new REU site proposal the goal is to provide interdisciplinary research experiences for promising students in building massive social networks and exploring applications of these. Specific targeted projects involve topics in the areas of distributed regular path query evaluation in social networks, privacy- preserving distributed architectures for social networks, mechanisms of 'virtual' friendship relationships, and the impact on understanding and modeling of online/offline human behaviors. The proposed REU site will recruit promising undergraduates each year and provide a structured plan of activities for the duration of the project. Special emphasis on attracting high quality students from under-represented groups in research. An eight- week summer session will be organized for all REU students.
In Suffolk University NSF REU site program, 23 undergraduate students were directed in the exploration of a number of challenges in the "technical" side of online social networks (OSNs) including evaluation of the notion of "friendship" and its psychological implications, exploration trust-based knowledge retrieval and modeling of privacy preserving distributed architectures for social networks. Working in teams, participating students were exposed to different aspects of the academic research environment while working on nine different projects designed to develop their potential as well as promote creative and independent thinking. Here are some of the outcomes of students’ work in the Suffolk NSF site: Students investigated the dynamics of Diaspora, a real-world Internet-scale decentralized online social network and performed the first such measurement study (to the best of our knowledge) of this network of independent, federated servers that are administrated by individual users who allow Diaspora users' profiles to be hosted on their servers. Growth of Diaspora can be characterized by a model similar to the Barabasi-Albert preferential attachment model previously proposed for a number of traditional scale-free networks. This work may shed light on the evolution of decentralized OSNs in practice and provide valuable empirical data and lessons to further improve their design and implementation to grant better privacy Students explored the data replication problem in the decentralized OSNs where users put replicas of their data into different servers in order to get better privacy compared with a centralized online social network such as Facebook. This work showed that if users only trust their immediate friends (social distance is only one hop) for data replication, the efficiency loss is very low. In contrast, increasing the trusts level (i.e. trusting friends’ friends) causes significant efficiency loss. These results provide significant insight into the data availability problem in a decentralized online social network, and these findings illustrate that it is possible to design a social scheme to guide users in their data replication process that can significantly improve the overall data availability in online social networks. Students investigated the relationship between publishing (blogging) and socializing in online social networks. Experimental work studying the co-evolution of both activities over time in LiveJournal suggest that publishing may be the dominant behavior in blogging social networks, and social activities, when present, complement blogging, rather than compete with it. Students investigated aspects of emotionality transfer in Reddit, a social news aggregation website whose users are responsible for supplying the site with its content, and SafeHeaven, an online forum created for people who suffer from depression and engage in self-injury. All content (posts, comments) were assigned pleasantness, activity, and imagery ratings from Whissell’s 8000 word Dictionary of Affect Language (DAL). Experimental results based on these measures show that, in general, co-occurrence and direct interaction (replying to one’s posts/comments) increase the correlation between emotionality of users in blogging networks unless the underlying networks are particularly contentious where users, who communicate with each other end up arguing and diverging on opinion more than other pairs.
