on December 17, 2015 at 2:00 pm
“Towards natural interactions between users and artificial agents”
Recent technological advances now make it possible to interact with computer systems in a wide variety of contexts. The new generation of systems will take advantage of these new modes of interaction, which are already found in consumer products, as well as the growing capacity of systems to initiate interaction autonomously. Certainly, autonomy is still limited today, but we are gradually getting used to interact with devices that surprise us pleasantly and whose naturalness of interaction makes them directly accessible to us, without manual.
My work focuses on the interaction between a user and a virtual environment, but the fundamental principles apply to other interactive systems with a certain autonomy, properties associated with the concept of autonomous artificial agent. Our team seeks to build virtual environments that «resist» us, like the physical world, by posing as a principle that it is through interaction that the action of the system makes sense for the user. We place ourselves here in the ecological approach of cognition by considering that the behavior of agents emerges from their interactions with their environment, and in particular the user.
The autonomy of agents means that the initiative of interaction is mixed; it is shared, even negotiated, between users and agents, like our natural interactions with our physical and social environment. The mixed initiative involves flexible coordination between interactants that operates at different cognitive levels, ranging from “motor” coordination, reactive, instantaneous, to goal-driven coordination, negotiated at the task level. The coordination between autonomous entities, users and artificial agents, requires to equip agents with behaviors that are both stable, and therefore intelligible to users, and adaptive, to accommodate the disturbances generated by the actions of users.
My research focuses on the behavior models of autonomous entities inhabiting virtual environments that can interact with each other and with users. During this presentation, I will explain how we model the different modes of coordination between agents, and more specifically virtual humans. I will present how we have applied these principles to the control of pedestrian movements, the management of the turn of speech between a user and a conversational agent, or the coordination of the activity in a virtual reality environment for human learning.
Auditorium IRCICA, Haute Borne science park in Villeneuve d’Ascq
français
English