Policing Commitments in Virtual Organisations

Alun Preece


Commitment management and policing are key issues in the management of
agent-mediated virtual organisations (VOs). A service-provider - which
may be a single agent acting within an organisation, or the VO acting as
a collective whole - manages particular resources, and commits these
resources to meet specific goals. Commitments can be modelled as
constraints on resources. Such constraints are often soft: they can be
broken if necessary. When an agent breaks a commitment, it will be in
breach of contract. Service-provisioning domains are often not fully
observable, featuring incomplete logging of events, or unreliable
sensors. In such domains, the process of identifying which agent is to
blame for breach of contract is essentially one of argumentation: agents
are able to advance arguments on their views of the state of the
environment. When a disagreement occurs, they may be able to probe the
environment state. By associating utility with both probing actions and
winning the argument, a strategy for argumentation can be defined.

In this talk, I will present an integrated framework for representing
and reasoning about commitments, and monitoring contracts in
partially-observable domains using argumentation and subjective logic.
This framework has been applied in various "e-domains": e-business,
e-science, and e-response.
Tuesday 5th December 2006, 14:00
Robert Recorde Room
Department of Computer Science