mOeX Seminar - Loïs Vanhée

Abstract 

In a context where artificial intelligence (AI) is at the heart of essential societal issues, debates on the use of AI systems to model human emotions (whether to imitate and/or identify and respond to them) often rely on dogma instead of scientific consensus, with societal implications that may raise questions, for example, through the legal protection provided by the AI ​​Act. 
This presentation aims to provide a broader conceptual framework, based on interdisciplinary crossover, to capture the ins and outs of AI models of human emotions. Using examples of successful work, we will cover the development chain of these systems, from psychological theories for understanding emotions, to their formalization, to their implementation through several AI methods, and conclude with practical examples of models and the many forms they can take in society. At the end of this presentation, participants will have a clear idea of ​​how other disciplines can be inspired to develop AI models of human behavior and emotions in particular, as well as broader ideas about what these models could be. A common response I receive is: "I thought I had a clear idea, but actually, there are many other ways to think about it." 
A discussion space will then be open to exchange views on these topics, whether to explore specific questions in more depth or to discuss future collaborations

Biography 

Loïs Vanhée is associate professor in ethical and responsible AI at Umeå University, Sweden, and invited professor at the University of Tromsø, Norway. He co-leads TAIGA, the Centre for Transdisciplinary AI of Umeå University and develop several projects funded by the Swedish agency for research on modeling human anxiety and the use of AI systems for psycho-therapeutic purposes. He takes advantage of various AI models, either in natural language processing, Markov decision processes and social system simulation. Loïs coordinates a Special Interest Group on Models of Human Decision.