Virtual pet powered by a socially-emotional BICA

Abstract

Cognitive architectures are used to build intelligent agents, and nowadays special attention in this area is drawn to emotion modelling. The purpose of this study is to compare two models describing social-emotional behavior, one of which is based on a traditional machine learning algorithm, and the other on a cognitive architecture supporting social emotionality. It is hypothesized that the second model will be more efficient in eliciting user's empathy to a virtual cobot based on this model. Here the object of study is a virtual pet: a penguin. Two models controlling the pet were compared: a reinforcement learning model (a Q-learning algorithm) and the emotional cognitive architecture eBICA (Samsonovich, 2013). The second approach was based on a semantic map of pet's emotional states, that was constructed based on the human ranking. It is found that the eBICA model scores higher in participant's empathy compared to the model based on reinforcement learning. This article compares strengths and weaknesses of both methods. In conclusion, the findings indicate advantages of the approach based on eBICA compared to more traditional techniques. Results will have broad implications for building intelligent social agents.