Natural language interaction is the next frontier in the development of ubiquitous data and services. We in particular seen this in the field of chatbots, that is, machine agents serving as natural language user interfaces to data and service providers in social networks.
Already, Apple's Siri, enable easy interaction in situations where text and touch input is impractical. The Google Assistant aims to become a natural language interface to a broad range of service providers. And Facebook Messenger, with more than one billion users worldwide, already includes more than 100.000 chatbots.
For sure, we are still in the early days of chatbots. To advance the state of the art, research on human-chatbot interaction design is critical. We in particular consider two research challenges:
The conversational context challenge: Users interact with chatbots in written or spoken natural language. Hence, the potential range of input to a chatbot is formidable. Chatbot capabilities for interpreting input and providing adequate responses need to be strengthened.
The user knowledge challenge: Chatbots are to communicate with a varied group of people across gender, age, languages and preferences. However, new technologies too often create digital divides and biases across gender, age, and societal status. User models are needed to adapt interaction to individual user's requirements.
We aim to move beyond the current state of the art by combining and extending the fields for machine learning and human-computer interaction. Through unsupervised machine learning, we aim to establish the basis for chatbots to identify the conversational context, understand the user, and provide good responses. Through human-computer interaction we aim to provide the interaction design needed for helpful conversations between humans and chatbots.
Human-Chatbot Interaction Design involves researchers from SINTEF (human-centred design) and the Center for Artificial Intelligence Research (CAIR) at University of Agder. The project leader is Asbjørn Følstad at SINTEF.
The project is funded by IKT PLUSS, a program at The Research Council of Norway.