Human-ChatBot Interaction: measuring the psychophysiological reactions of chatbot users

Abstract

Affective computing and its applications have gained popularity in recent years. However, recognizing emotions is a challenge. In Computing, the artifacts used to recognize emotions are instruments of self-report or data collection through biological sensors. Interaction through chatbots, in turn, has grown and brings with it challenges related to how effective the system is and how satisfied the user is when using it. We carried out an experiment with the aim of learning more about the characteristics of human-non-human interaction. As a result, the expressive responses of users and the chatbots with whom they interacted are the focus of this human-chatbot interaction paper. We gathered face registration, body movements, self-report, and peripheral signals such as the electrocardiogram (ECG), valence, and arousal estimations of the various emotional changes during the engagement period with the Bob chatbot, a conversational recommendation system, that operates through the WhatsApp app, and recommends restaurants to the user based on information such as location. The results were encouraging; users felt comfortable and were receptive to a new conversational tool and the ECG sensors attached to the thorax, which motivates us to make improvements for future tests.