Conference on the Human Impact and Application of Autonomic Computing Systems (CHIACS2)

Autonomic Policy-Based Collaboration for Cross-Agency Human Teams

Mark Cornwell, James E Just, and Lalana Kagal

April 21, 2004

We describe our implementation of a set of agents that facilitate the formation and execution of ad hoc cross-organization human collaboration teams. Such teams are often formed to deal with actual, imminent, or potential crises that cross geographic or political jurisdictions or require coordinated expertise from organizations with different responsibilities. Distributed infrastructures to support ad-hoc collaboration require an extraordinary degree of flexibility. Effective collaboration between people and software resources that make up cross-agency teams is inhibited by differences in software technologies, organizational policies, security policies, and personal styles of working. In the first year of a multi-year effort, we have developed a policy-based collaboration infrastructure intended to overcome such obstacles. This infrastructure is based on composable software agents that interact with legacy infrastructure. These agents are guided by policies expressed in Rei. The Rei language enables reasoning about rights, responsibilities, obligations, and dispensations in various domains through specific ontology extensions for each domain that map to its core domain independent ontological terms. The agents are organized as a drop-in framework that augments, rather than replaces, an organization's existing email and office automation applications. The framework utilizes personal agents, agent-based task monitoring and workflows, and Web services to deliver its functionality. Preliminary results for use of the framework show a reduction in team formation time of two orders of magnitude. Experiences with the prototype development, lessons learned, and plans for future extension are discussed. The framework is being enhanced to facilitate on-going collaborations of team members and workgroups.

BibTeX OWL Tweet Scholar

Type: InProceedings

Organization: IBM Almaden

Downloads: 4014 downloads