King's College London

Multiparty privacy conflicts (MPCs) occur when the privacy of a group of individuals is affected by the same piece of information, yet they have different (possibly conflicting) individual privacy preferences. One of the domains in which MPCs manifest strongly is online social networks, where the majority of users reported having suffered MPCs when sharing photos in which multiple users were depicted. Previous work on supporting users to make collaborative decisions to decide on the optimal sharing policy to prevent MPCs share one critical limitation: they lack transparency in terms of how the optimal sharing policy recommended was arrived at, which has the problem that users may not be able to comprehend why a particular sharing policy might be the best to prevent a MPC, potentially hindering adoption and decreasing the chance for users to accept or influence the recommendations. In this paper, we report our work in progress towards an AI-based model for collaborative privacy decision making that can justify its choices and allows users to influence them based on human values. In particular, the model considers both the individual privacy preferences of the users involved as well as their values to drive the negotiation process to arrive at an agreed sharing policy.
We formally prove that the model we propose is correct, complete and that it terminates in finite time. We also provide an overview of the future directions in this line of research.