A Socially Emotional Virtual Registrar Based on eBICA and Deep Learning

Alyona S. Anisimova, Aleksey O. Mikhnev, Vladimir S. Tsarkov, Anatoly A. Dolgikh, Alexei V. Samsonovich

Abstract: Virtual embodied social agents are expected to become socially acceptable and useful for emotion regulation with applications in the field of service-oriented artificial intelligence. In this study an eBICA-based model of emotional interaction is used to control behavior of a virtual hotel receptionist: the Registrar, implemented in a virtual environment (VE) and in virtual reality (VR), using a multimodal user-computer interface. The two tasks for the registrar are (1) to perceive and characterize user's personality type and the emotional state based on user's behavior, and (2) to be able to establish and maintain a social contact with the user and to control user's emotions by selecting an appropriate behavior. These functions, first implemented in a cognitive model, then need to be transferred into a neural network using deep learning on model-generated data. Here the general concept of the system is described, and preliminary results are presented. The expected long-term impact includes new smart emotion regulation technologies and new feasible means for emotional communication with and via artifacts.