How to Design Successful Conversations in Conversational Agents in Healthcare?

Conversational agents (CA) applied in a healthcare setting are often designed to mimic healthcare professionals. Inappropriate design of their implemented conversation flow might impact on the achieved outcome, patient adherence and experience, and might even become a risk for patient safety. Objective of this paper is to identify factors to be considered when designing conversations of health CA to ensure that the patient-CA communication is successful. Focus is on rule-based CA-based medical interview assistants, i.e. systems that collect the medical history from a patient implemented using rules. Starting from models, guidelines and best practices for successful healthcare communication (e.g. partnership model, AIDET model and RESPECT model), I derived aspects to be considered in designing successful interactions of health CA with patients. Transferring basic concepts of these models to patient-health CA communication leads to the following conclusions: Health CA should be equipped with certain communication skills, including being empathetic, know how to listen, have respect for the patient, have open-ended skills and be able to adapt to the level of knowledge of the user. Additionally, I derived information on how to structure the conversation flow.