ijcai-99 Home Page

A3 - Collaborative Multiagent Systems

Monday, AM

Barbara Grosz & Charlie Ortiz

Systems that are able to act as collaborative partners on joint tasks have the potential to greatly improve human- computer interactions and productivity. Such collaborative systems are within reach thanks to progress in our understanding of rationality, both collective and individual.

This tutorial will describe both the major theoretical advances that can support the principled designs of such systems as well as describe implementations based on these theories.

The tutorial will begin with an overview of rationality: What it means for an agent to be rational and how this can be reflected in agent designs. This will include a brief review of models of mental state: for example, the representation and role of intentions and the relation of intentions to other attitudes such as that of belief.

Then we will consider information flow within agent architectures, emphasizing considerations of resource-boundedness and the ways this affects formalizations and system designs.

The tutorial will then examine a range of approaches to modeling the collaborative behavior of a group of agents on a joint task. Several formal computational models will be presented and examined in the light of major philosophical approaches. The formal models require the introduction of new notions of intention, ability, and helpful behavior. These new notions will be examined, as will ways to model stages of partiality in joint planning processes. The relationship of this work to work in distributed AI will be discussed briefly. Finally, applications to human-computer communication and planning will be discussed.

Prerequisite knowledge:
This tutorial is suitable for a general AI audience. Knowledge of AI planning would be helpful. It should be of interest to: researchers in distributed AI; those interested in the theoretical aspects of collaboration; and those interested in designing and building collaborative information systems, user interfaces, and planning systems.

Barbara Grosz is Gordon McKay Professor of Computer Science at Harvard University and is a past president of AAAI. Her research addresses fundamental problems in modeling collaborative activity and in developing computer systems able to collaborate with each other and their users. She is one of the developers of the SharedPlans model of collaboration. She is extending this model and using it to construct collaborative interfaces and computer agents that work together in teams.

E-mail: grosz@eecs.harvard.edu
Home page: http://www.eecs.harvard.edu/grosz/

Charlie Ortiz is a computer scientist in the Artificial Intelligence Center at SRI International where he has also served as director of the Applied AI Technology Program. While a postdoctoral fellow at Harvard, he conducted research on collaborative planning systems and rational agency. He holds an S.B. in physics from MIT and an M.S. in computer science from Columbia University. His Ph.D. in computer science from the University of Pennsylvania was for his work on causation.

E-mail: ortiz@ai.sri.com
Home page: http://www.ai.sri.com/~ortiz/

Webmaster: Sven Olofsson, sveno@dsv.su.se
Last modified: Mar 13, 1999