PhD project In order to move from current, limited chatbot interactions with automated systems towards more useful dialogues we need to develop dialogue systems that have a better understanding of the user and the social context in which the dialogues take place.
In the other hand the dialogue system should use its knowledge of the user and the social context to direct the dialogue, resolve ambiguities and create expectations of the interaction.
This will lead us from are active AI system that only reacts to the input of a user to an interactive AI system that can incorporate question answers series, but also returning questions asking for explication, checking whether information is understood, explaining which user input is not understood or should be formulated in a different way, etc.
However, where a user has a pretty good (and simple) model of the AI system when the chatbot is very limited (e.g. limited to filling in numbers or yes/no questions), this becomes more complex when the dialogues become more realistic and interactive.
Thus, the expectations of users about the capabilities of the AI system will increase and perhaps become unrealistic. Therefore, the aim of this project is to develop a dialogue system that can gradually get more realistic while keeping the user informed about its capabilities.
This should provide a good balance between naturalistic dialogues and user expectations at all times. We will start from dialogue systems already developed by members of the group and incorporate specific social concepts such as social practices, conventions and norms to achieve the desired interactivity without the problems of scalability of completely free natural language dialogues.
In order to test the usability of the developed dialogue system we aim to have at least one use case related to social robotics.
The project is part of the Wallenberg AI, Autonomous Systems and Software Program - Humanities and Society (WASP-HS). The program is initiated and generously funded by the Wallenberg Foundations. The vision of WASP-HS is to realise excellent research and build competence on the ethical, economic, labor market, social and legal aspects of Artificial Intelligence.
For more information about the research and other activities in the program, please visit WASP-HS.