In this work, we present a robotic storytelling system, where the characters have been modelled as cognitive agents embodied in Pepper and NAO robots. The characters have been designed by exploiting the ACT-R architecture, taking into account knowledge, behaviours, norms, and expectations typical of social practices and desires resulting from their personality. The characters explain their reasoning processes during the narration, through a sort of internal dialogue that generate a high level of credibility experienced over the audience.

The paper describes an architecture for a social conversational agent. The aim is to use the agent in a serious game to improve the social and communicative skills of the players, showing the social effects of conversational choices on the emotions and behavioural changes of the interlocutors.