A Natural and Unobtrusive Conversation Using a RASA-Driven Chatbot for Monitoring the Wellbeing of Elderlies

This paper investigates the development and evaluation of a chatbot for administering the EQ5D questionnaire with a focus on improving user experience and ease of completion. To achieve this, the chatbot employs an open-source conversational AI environment named RASA. The primary objective was to develop a system that leads users through the EQ5D questionnaire in a natural and unobtrusive way, while accurately assessing their current health status and maintaining a supportive interaction. The reason is that the new CALM framework within RASA integrates RASA with the successful LLM approach to conversational AI. Interestingly, there seems to be very little research on Conversational AI which explicitly employs the RASA framework. In this work we therefore demonstrate the usefulness of RASA and CALM by means of a prototype which incorporates custom actions and used log probability thresholds to manage categorization confidence and follow-up questioning, depending on the actual scores to the questions.