This project is concerned with grounding in multi-modal computer-supported collaboration. In particular, we concentrate on two issues: how constructed visual/spatial references (drawing a diagram) are used during grounding and the role of grounding mechanisms in problem solving. The project will progress in three phases:
For more information, call or write to Pierre Dillenbourg or David Traum
This project concerns the design of artificial agents able to collaborate with a human agent which is a key issue in the area of educational science and will advance the knowledge about learning from an interdisciplinary perspective, bringing together Swiss and Romanian researchers from the areas of psychology, educational science and computer science.
A lot of research in distributed artificial intelligence has focused on explicit coordination between artificial agents. These processes cannot be applied as such to human-computer collaboration, because the cost of coordination becomes too high. Hence, the goal of this project is to model the mechanisms of implicit coordination when the task is distributed over a human and an artificial agent. This project integrates into other TECFA research projects aiming to develop flexible and opportunistic artificial agents, namely to integrate them in interactive learning environments.
The experiments and development stages will be conducted around Internet-based collaboration tools (the MOOs). One argument is to reduce the communication bandwidth to something that can be processed in knowledge-based systems. The second argument is that a long term goal of this project is the design of intelligent agents used in distance education or on-line assistance. Such agents could be made available on the net from centers with moderate computing power as it may be the case in many places in Romania.
For more information, call or write to Pierre Dillenbourg or mail to Catalin Buiu
Certain characteristics of the collaborative interaction between two persons are 'more fertile' with regard to the learning outcome. Moobservers are non-intentional agents which consist of a recognising module which information is more or less displayed in a raw graphical manner to the collaborating subjects. A major question is whether the socio-cognitive behaviour of the peer changes when it gets information about its own interaction.
For more information, call or write to Patrick Jermann .