Don't Overshare, Don't Undershare: a Goldilocks Solution for Inferring Sharing Preferences and Social Relationships from Phone Logs
Privacy controls are infrequently set or changed in online social sharing systems, resulting in static privacy settings that seldom match the user’s preferences. Both over- and under-sharing in these systems can have negative social outcomes. Furthermore, as social sharing becomes more passive and continuous (such as Facebook’s “frictionless sharing”), these negative outcomes are exacerbated. The goal of this work is to partially automate the process of specifying sharing preferences by inferring dimensions of the social relationships that the user has with each of her friends. In this work we show that: 1. There is a connection between relationship metrics (i.e. tie strength and life facets) and sharing preferences, 2. There is some evidence that these metrics can be inferred from a user’s communication data, and 3. A computational understanding of social relationships might improve a user’s experience across a variety of applications.