Tech Reports


The Implementation of Dynamic Assignment of Rights, Responsibilities and Sanctions to External Agents in Normative Multiagent Systems (PhD Thesis)

Farnaz Derakhshan


Recently, the design and development of multiagent systems (MASs) has become increasingly concerned with the recognition that they will be used in a dynamic and open environments. In such environments, it is a very difficult and complicated task to anticipate all possible runtime situations at design time. Therefore, in order to respond to changes in this environment it is necessary to allow the system to provide dynamic responses at runtime. This thesis is concerned with one particular aspect of such responses. Our novel contribution is that we explicitly identify, clarify and address the problem of dynamic assignment of rights, responsibilities (R&Rs) and sanctions to external agents in normative MASs.

The background setting of this work deals with the topic of dynamism in normative MASs and attempts to address and combine some issues regarding dynamic resources in MASs and different types of norms in legal systems consisting of various types of legal modalities, including obligation, prohibition, permission and right; enforcement modalities, including punishment, reward and compensation; and all key elements of norms such as addressee, beneficiary, temporal notions, and preconditions.

Following this introduction, we propose two alternative methods for dynamic assignment of R&Rs and sanctions to external agents, and propose a formalism to represent a commonsense understanding of our solution. The first method is based on role hierarchies in MASs and the second method is based on conditional norms. Both methods have common features including reliance on the concept of role, using a normative Knowledge Base (KB) and sensitivity to runtime occurrences affecting the MAS. The significant differences of these two mechanisms lie in the definition of roles and normative KBs.

Furthermore, we consider aspects of implementation based on common features of the proposed methods, which we follow with a general implementation architecture for dynamic assignment of R&Rs and sanctions to external agents. Using this general architecture and guidelines, we present an agent-based auction application to demonstrate the practical feasibility of our approach and of our architecture.

The implementation of an agent auction that we present allows us to examine and compare the functionality of our two methods under various scenarios, including different runtime occurrences and various types of legal notions.

[Full Paper]