Privacy in ubiquitous computing environments is primarily considered as a problem of protecting personal information from unauthorized access and misuse. Additionally it can also be seen as a process of interpersonal communication where not hiding but selective disclosure of personal information is the central issue, i.e. how users can practice privacy intuitively and dynamically in computerized environments similar to the analog world. In this work we discuss the management of private information concerning interpersonal privacy implications in smart environments. Existing work mostly does not match the intuitive and dynamic aspects of privacy in context of interpersonal communication. As an alternative we suggest an ad hoc approach to privacy management which uses learning techniques for an in situ disclosure assistance and present user interaction models for this disclosure assistance.