Thesis of Joseph Garnier

Modeling and simulating the somatic and cognitive process of emotion: application to virtual characters.


In works of fiction, the suspension of disbelief is a cognitive operation that allows the public to accept, when he has some doubts about the truth of a premise or hypothesis, to question all logical or scientific arguments in order to benefit from the story and the universe proposed by the author (Coleridge, 1817). It is a kind of unconscious pact between the author and the public that the latter is ready to accept to believe even though he knows that this is not the reality. This unconscious pact allows the author to recreate another reality, which to be plausible must be credible and coherent rather than realistic. For many years, research has shown that virtual characters with a set of abilities including emotional behaviors driven by goals, mental states, social knowledge and behaviors adapted to this knowledge, as well as access to natural language, contribute to making the universe credible and coherent (Bates et al., 1992, 1994).
Among these abilities, emotion plays a central role by being the only one to have some interactions and impacts on all the others, the corollary being that it cannot exist without them; it has a very strong interdependence. Since Darwin, biologists, neuroscientists and psychologists have been studying emotion to better understand this phenomenon and propose a theory. Research has shown that emotion is a somatic and cognitive process that affects both body and mind. It intervenes from the perception of the stimulus until the contraction of a muscle in constant interaction with almost all the other processes of the brain. This makes it a central and indispensable system for many biological species. Its two main functions are, on the one hand, the communication of its intentions and its state through a set of body signals (reddening of the skin, tears, smiles, leaks, etc.) and, on the other hand, the selection of behaviors during the decision-making process for a better adaptation to the social and physical context.
Because of the fundamental role of emotion in the survival and adaptation of biological species, I have undertook to model and simulate the somatic and cognitive process of emotion and then experiment with it in virtual characters. This capacity should enable them to act in a manner adapted to the context, thus appearing credible in the eyes of users and favoring the agreed suspension of disbelief. Emotion is based on many other processes, just as necessary for adaptation, which must also be modelled and simulated: sensory perception, goal hierarchy, homeostasis, autobiographical memory (semantic and episodic), updating knowledge and mental states, planning and decision-making. Although applied to virtual characters, these research works have a broader scope that extends to all artificial agents evolving in a virtual or real environment.

Advisor: Jean-Charles Marty
Coadvisor: Karim Sehaba