The foundation of this dissertation is based on the assumption that advances in artificial intelligence will extend the capabilities of intelligent agents and create new opportunities and challenges for the interaction with them. One central aspect of this are the abilities in argumentation, logical thinking, and the consideration of contextual information. These abilities enable increasingly autonomous decision-making, proactive interactions across various modalities such as spoken dialogue or graphical user interfaces, and automated actions. While this provides the opportunity to proactively assist users in everyday tasks or even take over entire work processes in a situationally adaptive manner, it also presents new challenges: The behavioral change of the AI agent from a purely user-driven, multimodal dialogue system towards proactivity and autonomy represents a difference from current AI agents. This transition carries the risk of inducing a sense of loss of control in users due to inappropriate actions or mistrust due to a lack of transparency, which could negatively affect the acceptance of such systems.
This dissertation addresses the new advances and challenges to fully utilize the potential of autonomous, proactive agents and optimize their use in everyday life, thereby maximizing user comfort, acceptance, and efficiency in interacting with intelligent assistants. The focus is mainly dedicated to two research areas: 1) Investigating the use of context and user information for personalized and situation-based proactive and autonomous interactions and 2) the development and evaluation of interaction concepts to improve user trust and acceptance of autonomous systems.
