Institute for Software Research
School of Computer Science, Carnegie Mellon University
Understandable Learning of Privacy Preferences
Jonathan Mugan, Tarun Sharma, Norman Sadeh
Currently unavailable electronically.
As mobile and social networking applications continue to proliferate, they also increasingly rely on the collection of an ever wider range of contextual attributes, location being a prime example. Prior work has shown that people's privacy preferences when it comes to sharing this information are often complex and that expecting users to spend the time necessary to tediously specify these preferences is unrealistic. Yet research has also shown that users are concerned about their privacy and that adequately capturing their privacy will likely be necessary for some of these mobile and social networking applications to be more broadly adopted. The present article reports on research aimed at reducing user burden through the development of two types of user-oriented machine learning techniques: (1) techniques to automatically generate small numbers of user-understandable privacy profiles (or "personas") that users can chose from when configuring their privacy settings, (2) techniques to turn user feedback into suggestions for incrementally modifying a userÔs existing privacy settings. We study to what extent these techniques, by themselves and in combination, can help users rapidly converge towards their preferred privacy settings in the context of location sharing scenarios where the settings control when and where a user's location is shared with different groups of recipients (close friends and family, Facebook friends, members of the university community, and advertisers). Our results, which are based on data collected from 27 subjects over a period of 3 weeks, suggest that both types of techniques can significantly help users converge towards their desired privacy settings, with the biggest improvements typically resulting from the use of privacy personas.