Title

Socially Aware Dialogue Management

To transition from the current, limited chatbot interactions to more meaningful and effective dialogues, it is essential to develop dialogue systems that better understand the user and the social context in which interactions occur. These systems should not only respond to user input but also actively guide the conversation, resolve ambiguities, and manage user expectations throughout the interaction.

The goal is to move beyond AI systems that merely react to inputs, towards interactive AI systems capable of engaging in more sophisticated dialogues. This includes not only responding to questions but also asking clarifying questions, checking for understanding, explaining any misunderstandings, and suggesting alternative ways to phrase inputs. As dialogue systems become more realistic and interactive, users’ expectations of the AI’s capabilities may increase, sometimes leading to unrealistic assumptions about what the system can do.

The primary aim of this project is to develop a dialogue system that gradually becomes more realistic while keeping users informed about its capabilities. This approach seeks to strike a balance between naturalistic dialogue and user expectations, ensuring that interactions remain effective and understandable.

The project will build on existing dialogue systems developed by the research group, incorporating social concepts such as social practices, conventions, and norms to enhance interactivity without the scalability challenges of entirely free-form natural language dialogues. To assess the system’s usability, the project will include at least one use case related to social robotics, providing a practical application to test the effectiveness of the developed dialogue system.