Look who’s talking? Developing and evaluating chatbots for health

Digitalisation of society has led to the creation of artificial conversational agents with which humans can communicate about a myriad of topics, including health. Chatbots are a popular example of conversational agents and, in essence, they are computer programs with humanlike communicative skills. They rely on dialog systems that process textual input known as Natural Language Processing (NLP) and can generate adequate textual responses known as Natural Language Generation (NLG). Chatbots have many potential affordances: they are always accessible, cost efficient, scalable and able to provide personalized advice to many people simultaneously, in ways that human partners can never do. 

Unfortunately, despite their promise, the current generation of chatbots has substantial problems acting as human-like conversation partners. They are not capable of generating a broad variety of responses, often only take into account the last contribution of their human communication partner, and lack a ‘theory-of-mind’ a reflection on the human user’s perceived knowledge, goals and intentions, derived from the textual input in the preceding interaction. 

As a result, the potential of chatbots has not materialised yet in complex domains where multiple, longer human-chatbot interactions are required, such as for health behavior change applications. The main aim of the ‘Look Who’s Talking’ consortium is to create and fieldtest chatbots designed to develop and foster longterm engaging and humanlike conversations for two technologically and linguistically diverse application areas (smoking cessation and safe sex promotion), whilst at the same time answering important ethical questions about the design and use of those chatbots. 

The consortium blends key expertise to reach this aim, such as computational linguistics, health behavior change, and social artificial intelligence, and collaborates with important societal partners (Trimbos, Soa Aids), and is made possible due to a grant from NWO Digitalization. The four-year project hires three PhD candidates. 

Research team ACHC
Divyaa Balaji, PhD candidate
Dr. Gert-Jan de Bruijn, (ACHC coordinator, co-applicant, co-promotor)
Prof. Dr. Reinout Wiers, University of Amsterdam and Prof. Dr. Tibor Bosse, Radboud University (promotores)

Status: ongoing since 01 September 2020
Funding: NWO Digitalisering
Link: Website under construction