Chatbots are machine agents that serve as a natural language interface to data and services. The current interest in chatbots is driven by recent advances in artificial intelligence and machine learning. Chatbots are now being developed to provide customer service, health advice, work support, or playful user experiences.
To develop more useful and user-friendly chatbots, we in this project seek to combine and advance the state-of-the-art of human-computer interaction (HCI) and machine learning. For this purpose, we aim to establish human-computer interaction design as a field of research with an associated specialist circle, that is, an international researcher network with a research interest in this field. Furthermore, we work towards contributing to novel chatbot prototypes with a basis in the needs and the large text-based data sets of collaborators and data providers, as well as open data sets of text and dialogue.
In the project, we work towards establishing and maintaining an international research network on human-chatbot interaction design. For this purpose, we have ? in collaboration with researchers from the Netherlands, Great Britain, and Greece ? established CONVERSATIONS, a series of international workshops on chatbot research. In November 2021 we arranged the fifth of these: CONVERSATIONS 2021. Here >190 participants ? researchers, practitioners, and students ? were registered as participants and 18 research papers were presented. Proceedings from the workshop are published by Springer.
The research activities in the project in particular concern studies of chatbot user experience and interaction design as well as new approaches for machine learning to improve the interaction between users and chatbots. We work on analysis of dialogue data between users and chatbots, and on self-reported data from users, to better understand chatbot user experience. We also work on how the Tsetlin-machine approach to machine learning, developed at the University of Agder, may be applied to strengthen semantic understanding in chatbots, strengthen understanding of sequences of events in dialogue, and for unsupervised learning in chatbot training.
Dissemination is an important part of the project, in particular as we aim to establish 'human-chatbot interaction design' as a field of research. For this purpose, we contribute to the arrangement of the master course Interacting with AI at the Institute of Informatics, University of Oslo. The project researchers also disseminated project results through academic publication, conference and seminar presentations, guest lectures, and media.
The project research partners are SINTEF, with the research group on Human-Computer Interaction, and University of Agder, with the Centre for AI Research (CAIR).
Natural language interaction is the next frontier in the development of ubiquitous data and services. Chatbots are particularly promising, that is, machine agents serving as natural language user interfaces to data and service in social networks.
To contribute to this field of high industrial and academic interest, we will establish a research community of critical mass (specialist circle) on human-chatbot interaction design. Here, we will combine and extend the state-of-the-art in human-computer interaction and machine learning. The specialist circle will have substantial impact through future European R&I projects (H2020), industry and public sector projects, knowledge networks, and education.
A particular concern in the project is to realise the potential of chatbots to strengthen digital inclusion. Given their low threshold of interaction through natural language, and easy access e.g. through chat services, chatbots have the potential to lower thresholds for digital participation across gender, age, and socio-economic status.
We will extend the state-of-the-art by addressing the need to establish and leverage knowledge of users and conversational context in human-chatbot interaction. For this purpose, we will apply machine learning to develop generative models and libraries of interaction patterns sensitive to user types and conversational contexts. Industry and public sector actors will be invited as data providers.
The integration of machine learning in HCI also represent a needed complement to this field. Machine learning enables the development of generative interaction models, which in turn can be used to simulate interaction patterns and drive the analysis and (re)design of human-chatbot interactions. This will allow future HCI to utilize available large data sets in analysis and design of interactive systems in general, and natural language user interfaces in particular.