On the heels of my TechCrunch post on why Foursquare users check in off the grid, HuffPo intern Jake Bialer turned me on to some work being done at Carnegie Mellon about mobile social services. From CMU Associate Professor Jason Hong:


1) Rethinking Location Sharing: Exploring the Implications of Social-Driven vs. Purpose-Driven Location Sharing
// Compares Purpose-Driven Location Sharing (eg to coordinate plans) vs Social-Driven (sharing because it’s fun, not because others need to know).
// “social-driven location sharing favored semantic location names, blurring of location information, and using location information to attract attention and boost self-presentation.”
// “In one-to-one location sharing, the user’s decision is simple: is the user comfortable telling this specific person her location. For one-to-many sharing, the decision is more complex: what may have been okay sharing with one person may not be okay sharing with many people. There are three reasons why large-group sharing might differ: (1) there is a larger variance in who receives the information, (2) there is a different motivation for sharing, and (3) there is a different expectation of plausible deniability.”
// “The success of Facebook is indicative that users are relatively comfortable sharing the same status information with everyone in their online social network (i.e., people of varying tie strength), but it is unclear if the same holds true for location sharing.”


2) Modeling People’s Place Naming Preferences in Location Sharing
// “Most location sharing applications display people’s locations on a map. However, people use a rich variety of terms to refer to their locations, such as “home,” “Starbucks,” or “the bus stop near my house.” Our longterm goal is to create a system that can automatically generate appropriate place names based on real-time context and user preferences.”
// “We also present a machine learning model for predicting how people name places. Using our data, this model is able to predict the place naming method people choose with an average accuracy higher than 85%.”
// When location was shared with more intimate social groups like family members or close friends, the portion of using geographic naming method was small (<15%) and the average granularity was finer (between street level and building level). However, when the location information was shared with less intimate social groups, such as
strangers, the usage of geographic naming was much higher but the average granularity drops dramatically (i.e. as coarse as city level granularity). This observation also confirmed people’s location blurring intentions get stronger when sharing with less intimate social groups.”


3) Bridging the Gap Between Physical Location and Online Social Networks
// Can co-location result in the next social graph? “We introduce a novel set of locationbased features for analyzing the social context of a geographic region, including location entropy, which measures the diversity of unique visitors of a location. Using these features, we provide a model for predicting friendship between two users by analyzing their location trails.”
// “The co-location network has roughly 3 times the number of edges as the social network, yet the social network is better connected. The co-location network has many small disconnected components, but it has a single large and highly connected subcomponent. Despite these differences, we have shown that the co-location graph contains important information that can be used to reconstruct a portion of the social network.”
// “Social network designers may find our methodology useful for designing social applications, such as location-aware information sharing platforms, privacy control mechanisms, and friend suggestion systems.”


4) Empirical Models of Privacy in Location Sharing
// “Our results show that users appear more comfortable sharing their presence at locations visited by a large and diverse set of people. Our study also indicates that people who visit a wider number of places tend to also be the subject of a greater number of requests for their locations. Over time these same people tend to also evolve more sophisticated privacy preferences, reflected by an increase in time- and location-based restrictions.”