Social sciences are becoming increasingly important for robotics research as work goes on to enable service robots to interact with inexperienced users. This endeavor can only be successful if the robots learn to interpret the users' behavior reliably and, in turn, provide feedback for the users, which enables them to understand the robot.

In order to achieve this goal, the thesis introduces an approach to describe the interaction situation as a dynamic construct with different levels of specificity. The situation concept is the starting point for a model which aims to explain the users' behavior. The second important component of the model is the expectations of the users with respect to the robot. Both the situation and the expectations are shown to be the main determinants of the users' behaviors.

With this theoretical background in mind, the thesis examines interactions from a home tour scenario in which a human teaches a robot about rooms and objects within them. To analyze the human expectations and behaviors in this situation, two main novel methods have been developed. In particular, a quantitative method for the analysis of the users' behavior repertoires (speech, gesture, eye gaze, body orientation, etc.) is introduced. The approach focuses on the interaction level, which describes the interplay between the robot and the user. In the second novel method, also the system level is taken into account, which includes the robot components and their interplay. This method serves for a detailed task analysis and helps to identify problems that occur in the interaction.

By applying these methods, the thesis contributes to the identification of underlying expectations that allow future behavior of the users to be predicted in particular situations. Knowledge about the users' behavior repertoires serves as a cue for the robot about the state of the interaction and the task the users aim to accomplish. Therefore, it enables robot developers to adapt the interaction models of the components to the situation, actual user expectations, and behaviors. The work provides a deeper understanding of the role of expectations in human-robot interaction and contributes to the interaction and system design of interactive robots.