You Got the Feeling: Attributing Affective States to Dialogical Social Robots

In this paper we report the result of an analysis aiming at investigating, among different virtually embodied social robots (endowed with different degrees of dialogical complexity), the perceived difference in emotion attribution and understanding by the human users interacting with them. In particular, in our case study, the most complex dialogical modality - using a emotional content to vehiculate its messages - has been based entirely on the adoption of a Large Language Model (i.e. chatGPT in our case) whilst the simplest one has been based on a manual simplification of the generated text. We report the obtained results based on the adoption of a number tests and standardized scales and highlight some possibile future directions.