Abstract
In this chapter, we discuss the concepts of agent organization, organizational model, and review some existing organizational models. Before the review, we discuss how to classify the diverse aspects of agent organizations currently captured by organizational models. These aspects are named “modelling dimensions”. We show that there are at least four basic dimensions: the structural dimension mainly composed of roles and groups, the interactive dimension characterized by dialogical interaction structures, the functional dimension formed by goal/task decomposition, and the normative dimension in which we find the concepts of norms, rights, rules, and so forth. Apart from the basic dimensions, we also identify four other complementary dimensions: environment, evaluation, evolution, and ontology. These are related to the aspects of situatedeness, measurement, adaptation, and domain specific semantics of agent organizations. Finally, we compare the organizational models reviewed and describe how the idea of modelling dimension can help in finding correspondences between organizational models.
TopAbstract
In this chapter, we discuss the concepts of agent organization, organizational model, and review some existing organizational models. Before the review, we discuss how to classify the diverse aspects of agent organizations currently captured by organizational models. These aspects are named “modelling dimensions”. We show that there are at least four basic dimensions: the structural dimension mainly composed of roles and groups, the interactive dimension characterized by dialogical interaction structures, the functional dimension formed by goal/task decomposition, and the normative dimension in which we find the concepts of norms, rights, rules, and so forth. Apart from the basic dimensions, we also identify four other complementary dimensions: environment, evaluation, evolution, and ontology. These are related to the aspects of situatedeness, measurement, adaptation, and domain specific semantics of agent organizations. Finally, we compare the organizational models reviewed and describe how the idea of modelling dimension can help in finding correspondences between organizational models.
TopIntroduction
In the last few years, a broad agreement in the area of Multi-Agent Systems (MASs) has been to consider the human organizations as a suitable metaphor to effectively assemble computational systems from a dynamic collection of heterogeneous autonomous agents (Ferber, Gutknecht & Michel, 2004; Zambonelli, Jennings & Wooldridge, 2003; Gasser, 2001). In such computational systems – often called “open MASs” –, the defining characteristics are both a variable number of autonomous agents at run-time (i.e., agents can enter and leave the system when it is in production), and the presence of agents with different interests and/or designs (i.e., agents representing different stakeholders, conceived by several designers, and/or built using different agent architectures). To cope with these characteristics, the organizational perspective proposes that the joint activity inside the MAS be explicitly regulated (moulded, constrained) by a consistent body of norms, plans, mechanisms and/or structures formally specified to achieve some definite global purpose. And this, in essence, is what “human organization” means when the autonomous agents are human beings – a dynamical collection of persons that accept to have their joint activity formally patterned and controlled, given some global goals (Scott, 1998). Inspired by the metaphor, in this chapter, we will use the term “agent organization” to denote an open MAS, or one of its sub-systems, that was designed and operates in a way similar to human organizations.
This broad agreement around agent organizations has led to the proposal of different organizational models for their engineering (incomplete list of proposals is: Ferber, Gutknecht & Michel, 2004; Lesser et al., 2004; Hübner, Sichman & Boissier, 2002; Esteva, Padget & Sierra, 2002; Dignum, 2004; Horling & Lesser, 2004; Tambe et al., 1999; Parunak & Odell, 2002; Silva, Choren & Lucena, 2004). An organizational model provides the designer with a conceptual framework and a syntax in which she can write organizational specifications for agent organizations. From an organizational specification, an agent organization can be implemented on a traditional agent platform or, more realistically, by using some organizational middleware or platform (Hübner, Sichman & Boissier, 2005; Esteva et al., 2004; Gutknecht & Ferber, 2000). In general, these organizational middleware or platforms take the organizational specifications as input, interpret them, and provide the agents with an organizational environment (agent organization) according to the specification. In order to enter, to work inside or to leave the agent organization, the agents are supposed to know how to access the services of the middleware/platform and to make requests according to the available organizational specification.
Key Terms in this Chapter
Organizational Norms: Also called “normative dimension”, it is composed of constructs to further regulate and show how organizational structure (time independent relations), organizational interaction (standardized functioning) and organizational functions are interrelated. The name “normative” is due to the fact that in the existing organizational model the concepts in this dimension are described in term of deontic norms (i.e., statements that regulate the behaviour of social entities: what they are allowed to do - direct or indirectly -, what they are obliged to do, etc.).
Modelling Dimension: A coherent class of modelling constructs found in a modelling language. The modelling dimensions are defined taking into account the main aspects in which a subject can be modelled. In the case of organizational models, we can identify four commonly occurring modelling dimensions: organizational structure, organizational interaction, organizational functions and organizational norms.
Organizational Functions: Also called “functional dimension”, it is a modelling dimension composed of constructs to represent global goals and goal decompositions (plans) to be accomplished by an agent organization.
Organizational Structure: Also called “structural dimension”, it is a modelling dimension that gathers constructs to represent what aspects of the structure of an agent organization have to be invariant through time. The main constructs found in it are roles, groups, and relationships between them.
Organizational Interaction: Also called “interactive dimension”, it is a modelling dimension that is formed by constructs to represent standardized actions and interactions involving the elements from the structural and functional dimensions. Some constructs found in this dimension are interaction protocols, scenes and scene structures.
Agent Organization: A Multi-Agent System (computational system), or one of its sub-systems, possibly open and which was designed and operates in a way similar to human organizations. Compared to human organizations, agent organizations are characterized by specific goals and formalized social structures.
Organizational Model: Modelling languages used to formally specify agent organizations. Organizational models provide a conceptualization of organizations for Muti-Agent Systems and a syntax to write specific models. In an abstract way, the syntax can be expressed by means of an organizational metamodel.