SaTC: CORE: Medium: Designing Privacy-Aware Social Companion Robots

Social companion robots can interact with people in a human-like way, using speech, gestures, and physical presence. They are designed to help families with activities like meal planning, child development, and communication. However, their advanced sensing and reasoning capabilities can lead to privacy concerns, as they can collect, infer, and share personal information about their users.

For instance, they can overhear private conversations as they move and act in the home. They also access sensitive environments, such as bedrooms, and can be designed to look like familiar objects or people, which can lead to people oversharing personal information. They might also inappropriately disclose information they learned with other home occupants or visitors.

Therefore, it is crucial to design more “privacy-aware” social companion robots that can reason about when to collect and share data, and when not to, in ways that are context-appropriate and match people’s needs for privacy. This interdisciplinary project aims to create design principles for privacy-aware social robots to improve human-robot interactions in home environments.

These principles will inform how robots are designed, used, and accepted in homes, workplaces, schools, and healthcare. Further, through outreach activities, the project will improve families’ understanding of privacy in smart environments through media coverage, workshops, and office hours. The educational activities of the project will train K-12 and college students in privacy systems, family studies, and human-computer interaction.

Finally, the project team will share code, publications, prototypes, and datasets based on the work.

This project develops, implements, and evaluates a novel privacy framework for human-robot interaction in four phases. This framework models users' privacy expectations, employs novel concepts grounded in human-robot interaction, develops new methods to help users manage their own privacy, and enables a study of privacy dynamics within families. Specifically, Phase I models users' privacy concerns and expectations regarding social robots, developing a privacy framework to identify everyday situations in which privacy awareness is most important.

Phase II instantiates the framework as a privacy controller for the robot’s actions that captures the dynamics of social privacy. At the core of the privacy controller is a customizable architecture that generates privacy-aware actions based on extracted high-level contextual factors from the robot's environment, such as location in the house and number of people present.

Phase III explores how the robot can leverage the privacy controller architecture to signal, demonstrate, or explain the robot's privacy-aware actions to users. Through deployment studies, phase IV assesses the impact of privacy awareness on the robot's management of the complex privacy trade-offs involved in participation in family life. The findings of these studies will provide an empirical basis for understanding a privacy-aware social robot's benefits, effects, and limitations on family dynamics.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.