A Step Towards the Development of Socio-cognitive Agent

Abstract

The embodiment of consciousness in socio-cognitive agents play a significant role in their acceptance as co-partners. Man-machine social interaction’s success is based on the agent’s generated believable behaviors. In this regard, such an agent needs to generate realistic beliefs, intentions, goals, self-regulation, verbal, and non-verbal communication (including gestures) according to the context of ongoing interaction is very important. This study hypothesizes that the implementation of the Theory of mind (TOM) may allow the agent to change its intentions, beliefs, and desires by predicting the existing perspective and mental states of the other agents involved in the given social interaction. To study the complexity of dynamics in a social context we have taken the case study of the ‘paper-scissor-rock’ game and developed a cognitive agent capable of using gestures for non-verbal communication following the Theory of mind. This work is in progress and is being developed on the iCub robot simulation. We have used tapped delay line neural networks as the basis for reinforcement learning and strategic planning. This paper will report the cognitive model, neural network initial results obtained.