Human teamwork can be supported by agent technology by providing each human team member with an agent that monitors, supports and advices the human. The agent can, for example, monitor the human's workload, and share that information with (agents of) other team members so that work can be distributed effectively. However, though sharing information can lead to a higher team performance, it may violate the individual team members' privacy. This raises the question what type of and how often information should be shared between team members. This paper addresses this question by studying the trade-off between privacy loss and team performance in the train traffic control domain. We provide a conceptual domain analysis, introduce a formal model of train traffic control teams and their dynamics, and describe an agent-based simulation experiment that investigates the effects of sharing different types and amounts of information on privacy loss and team performance. The results give insight in the extent to which different information types cause privacy loss and contribute to team performance. This work enables the design of privacy-sensitive support agents for teamwork.