Thesis of Alexandre Galdeano

Developmental learning of behaviors suggesting empathy in heterogenous social robots

Abandoned thesis: 30/01/2020

Advisor: Salima Hassas
Coadvisor: Mathieu Lefort


This Ph.D. thesis aims to study, conceive, and develop mechanisms that enable social robots to have more empathy, in order to enhance Human-Robot interactions. Empathy has a been one of the keys of the human evolution and of the social interactions and some parts of the human brain are specialized in the understanding of others’ emotions, goals, knowledge, and beliefs. Because of the need for Human-Robot interactions, robots specialized in social interactions have been developed. However, as of today, the available robots on the market only have a narrow spectrum of social interaction skills. These limits are due to various factors: noisy environment, not or only a bit structured, poor vocal recognition, poor perception in general, unexpected events, etc. 

In this thesis, I want to use Developmental Artificial Intelligence (DAI) to build the core of an interactional engine able to generate interactions that seem to be more empathic. The DAI is an AI’s paradigm inspired by Jean Piaget’s work on developmental psychology and which has recently been subject to a growing interest in the scientific community, especially in AI. Indeed, since it enables a long life learning of concepts from interactional regularities with the environment, the DAI can address a wide range of problems encountered in robotics. Unlike most approaches that require prior knowledge or a lot of data to constitute a learning base, the DAI has good results in short times with less data-points than big data and deep learning, and in unknown and changing environments. In this thesis, I will be focusing on the creation of interactions, that are the most empathic possible, using DAI.