Towards humanlike chatbots helping users cope with stressful situations

Abstract

Many researchers are studying the application of computer-generated emotional support to increase well-being in humans. In this chapter, we investigate some challenges related to the development of effective stress support bots. We developed a chatbot for Facebook Messenger that, using IBM Watson's text mining and machine learning capabilities, can carry out small dialogues with its users and recognise when they are talking about stressful daily-life events. Based on previous studies, our presented bot provides emotionally supportive text messages tailored to the stressors users share with it. Two groups of specialists have interacted with our software and provided useful insights via focus groups. Based on the results of the focus groups, a number of recommendations have been formulated to further improve stress support bots. In future work, we plan to address all the feedback obtained during this study, as well as to conduct an experiment to investigate to what extent our chatbot is able to make people cope with their daily-life stressful situations.