The problem of robustly controlling a robot in a dynamic, uncertain, world remains one of the great challenges of AI. The Nursebot project, which aims to develop mobile robot assistants for elderly people, provides a unique platform to explore this challenge. Key tasks for the robot include delivering information, guiding people through their environment, and verbally interacting with them in a socially appropriate manner. To accomplish these, the robot's software systems rely heavily on probabilistic AI techniques for acting under uncertainty. In this talk, I will discuss three modules relevant to ensuring successful human-robot interaction: an automated reminder system; a people tracking and detection system; and finally a high-level robot controller that performs planning under uncertainty by incorporating knowledge from low-level modules. While previous experiments with fielded robot systems have provided evidence that probabilistic techniques work well in the context of navigation, this project shows the same to be true of human robot interaction.