Resolving multi-user privacy conflicts with computational theory of mind

Online Social Networks (OSNs) serve as digital platforms for users to share information and build relationships. These networks facilitate the sharing of diverse content, which may disclose personal information about users. Some of these contents pertain to multiple users (such as group pictures), with different privacy expectations. Sharing of such content may lead to multi-user privacy conflicts. Decision-making mechanisms are crucial to managing conflicting privacy preferences among users, reducing their effort in conflict resolution. Various mechanisms are proposed in the literature, most of which demand significant computational resources. We propose a novel approach based on computational modeling of Theory of Mind (ToM), the human ability to understand others’ mental states (e.g., their beliefs, preferences, goals, etc.), to portray users’ privacy expectations.We argue that leveraging computational
ToM modeling allows the design of agents capable of accurately capturing
users’ behavior and reasoning about other agents’ privacy understanding,
making them effective tools in multi-user privacy conflict management.
To illustrate our ideas, we consider a content-sharing scenario and point
out potential benefits of using our agent-based computational ToM approach
in resolution of privacy conflicts.