AN INTERMEDIATION SYSTEM BASED ON AGENTS

MODELLING TO SHARE KNOWLEDGE IN A COMMUNITY OF

PRACTICES

Clauvice Kenfack and Danielle Boulanger

MODEME, Jean Moulin University, 6 cours Albert Thomas, Lyon, France

Keywords: Community of Practices, Knowledge Emergence, Intermediation System, JADE Agent.

Abstract: This paper presents an intermediation multi-agent system to manage the distributed collaborative design

environment like a CoPs. The JADE-based intermediation system (JAIS) uses community enactment

mechanism and agent integration mechanism. The community enactment mechanism is the system kernel

and follows the specifications of the CoPs reference model. The system kernel supports four agents

(moderator, user, expert and newcomer agents) to manage the community, whereas the integration

mechanism supports an intermediation agent to interact, coordinate and monitor the activities between

agents. JAIS facilitates the team interaction in a collaborative and distributed environment.

1 INTRODUCTION

We describe an intermediation system able to design

distributed and collaborative environment in a

community of practices.

Agent modelling is a good candidate to highlight

emerging knowledge coming from the CoPs’

different members.

JADE (2006) is used as the agent platform for

linking the heterogeneous system in a distributed

environment like CoPs.

This paper is organized as follows: Section 2

presents the state of art and section 3 describes our

proposal. Section 4 defines the intermediation

system framework, the communication model, and

the agent intermediation model. The last section

summarizes contributions and provides some

suggestions for future works.

2 LITERATURE REVIEW

A community of practices is defined as a group of

agents (human) having quite strong common points

such as their social aptitude, skills, and cognitive

capacities and which share a set of problems on a

given subject, and look further into their knowledge

emerging (Nonaka, 1995) and experiments by the

daily interactions they maintain. A CoPs provides a

“forum” aiming at sharing ideas, solving problems,

disseminating best practices, and organizing

knowledge (Wenger, 2002). CoPs can be defined as

a group of agents which share a substrate of

knowledge related to their professional skills,

interacting via virtual spaces. The actions performed

by the members to reach a consensus on a subject

are confronted to enrich agents’ knowledge and

know-how. Through their actions, the bases of

common and individual knowledge are built, and the

practices of the community are developed. The

context of CoPs refers to a range of rich agents’

behaviours which belong to the community

(Yildizoglu and al, 2004).

The CoPs organisational model. In a classical

manner the terminology used in a CoPs is:

A domain, as defined in (Wenger, 2004), is the

area of knowledge that brings the community

together, gives it its identity and defines the key

issues that the CoPs members need to address. It is

the “focus” of the CoPs and evolves over its life

span in response to new, emerging challenges and

issues (Henri, 2006).

A field: It is the “context” of the CoPs; it can be

referred to as the “discipline” or the “branch of

knowledge” of the CoPs members.

The practice represents standards, rules, ideas,

frameworks, languages, accounts, and documents

278

Kenfack C. and Boulanger D. (2008).

AN INTERMEDIATION SYSTEM BASED ON AGENTS MODELLING TO SHARE KNOWLEDGE IN A COMMUNITY OF PRACTICES.

In Proceedings of the Third International Conference on Software and Data Technologies - SE/GSDCA/MUSE, pages 278-283

DOI: 10.5220/0001885102780283

 SciTePress

shared by the members. We can affirm that the

practice represents knowledge which the community

creates, shares and maintains (Wenger and al, 2002).

Objective (Activities): related to the CoPs as a

whole, or to a part of it (a group, a project, a team,

depending on the CoPs organisation and functioning

modes), an objective can be permanent or

temporary.

CoPs is Characterised by: Membership and

Cultural Diversity (from homogeneity to

heterogeneity): the nationality, the profile and the

organisational culture (Langelier and Wenger,

2005).

Policies: The policies are expressed by

standards, rules defined within the community; for

example policies to leave/join a community.

Role Member: A role is a behaviour identifier,

which represents a whole of actions and the

constraints of their appearance. In short in UML a

role is a stereotype of class that expresses a

collection of operations. It can include constraints on

the operations.

The benefit get through the share of information

makes possible for the members to develop a single

comprehension (common language, and practices) in

their field. The process of negotiation is made by a

broadcast of the subjects discussed within the whole

community. Roles can be allocated to members

according to both their experience and level of

confidence. In that case, attribution of roles can be

done by vote or through a consensual way.

Furthermore, members of the community use

technological tools which can be synchronous or

asynchronous.

Generally, within a community following

process occur from the interventions of the members

(Deale, 2006):

Exchanges occur when a participant asks a

question or proposes an observation made at his

workplace or a problem. Furthermore exchanges can

lead to experiences sharing where participants

develop their observations of their own context.

Agent technology is a good candidate to model

CoPs because he offers a great flexibility concerning

development of complex and distributed systems.

An agent is a software entity that can autonomously

perform routine tasks with a level of intelligence

(Boudriga 2004) Wooldridge (Wooldridge and

Jennings, 1999). Nwana (Nwana, 1996) divides

agents into five types: collaborative agents, interface

agents, mobile agents, intelligent agents, and smart

agents. Goal driven agents typically possess three

key characteristics which are autonomy,

cooperation, and learning (Etzioni, 1995) (Liang

2002) (Nwana 1996). They are able to acting

autonomously, cooperatively, and collectively.

In the field of multi-agent system for knowledge

dissemination and management, some systems have

already been realized.

Most of such systems are specialized for

information retrieval from heterogeneous databases

such as SIMS (Arens et al., 1996), InfoMaster

(Genessereth and al., 1997) RETSINA (Decker and

Sycara, 1997) and InfoSleuth (Nodine and al.,

2000). They are composed of agents that wrap these

information repositories, combine and translate

information through mediation techniques.

Another important set of systems such as SAIRE

(Odubiyi and al., 1997), UMDL (Weinstein and al.,

1999), CASMIR (Berney and Ferneley, 1999) is

specialized to facilitate information retrieval and to

gathering information. Such systems help the user

supporting retrieval of the relevant information from

one or more information repositories and adapting

the interaction with the system to the user’s

preferences.

Roda (Roda and al, 2003) proposed an agent-

based system designed to support the adoption of

knowledge sharing practices within communities.

Hammond (Hammond and al, 2004) proposed an

approach for virtual communities based on JADE.

3 OUR PROPOSAL

3.1 System Model

This specification is based on the modus operandi of

the practices presented in the previous parts. The

generic model proposed is used to define the models

of the entire system. It also takes into account the

composition of the community, the types of

interactions, knowledge treatment, themselves and

the internal structure of the agent.

This specification has identified the models of

our system; it defines the organizational model for

the community of practices and some others models

like cooperation model, interaction, coordination,

and agents models.

As for the models of intermediation system, it is

composed of the cooperation model which includes

the agent model and the coordination model

(communication structure, Knowledge base).The

interaction model specifies in the interaction

structure the agent activities in terms of agreements

between the roles played by members; moreover it

AN INTERMEDIATION SYSTEM BASED ON AGENTS MODELLING TO SHARE KNOWLEDGE IN A

COMMUNITY OF PRACTICES

279

describes the operational interactions results as well

as the interaction protocol and social standards

applied to the interactions between agents.

The interaction structure controls the interactions

and reinforces the policies within the community.

It therefore describes in our model all the

interactions that represent the activities of the

participants through their agents in the community

of practices and the activities of agents in the

intermediation system. These interactions require

coordinated action of several roles in the task

resolution. This interaction structure gives the

sequence of events which specifies the intent of the

interactions between the roles that take place

according to the norms, rules and policies. The

interaction model can be used to compose an

interaction protocol consisting in a set of messages

based on the communication acts.

Figure 1: System models.

As we see in the figure 1, the macro level

corresponds to the organizational modeling aspect of

the community of practices i.e. its offers a

description of the themes of the community, to his

field, protocols resolution of the tasks, activities,

members of the community. When at the micro level

it models the cooperation progress between

members (cooperation model) while describing the

process of cooperation, active in the community, the

roles played by members, their skills, as well as the

characteristics of agents in the cooperation, as well

as their roles. The process of cooperation of all the

agents of the system will be modeled through

models of coordination and interaction

3.2 Intermediation System

At first we provide here a definition of the concept

of intermediation system.

By drawing on the work of E. Rigaud (Rigaud,

2003) which relates to the application of multi-agent

systems for virtual organisations for the risks

management of SMEs, we define the intermediation

system as "a system that allows all members of the

CoPs to create interactions between them, even in

geographically dispersed locations, in order to

promote the co-construction of meaning, enrich their

common knowledge base, improve their skills, share,

exchange and the acquire knowledge. To do this, the

system must be able to provide the mechanisms

needed to manage constraints to be imposed by the

functioning of the community”. These constraints

can be tasks allocation, the profiles management, the

control the access to the knowledge base, and the

processing requests.

The key issues in multi-agent systems are

communication, cooperation, and coordination

(Papazoglou, 2001). Specifically, communication

enables an agent to exchange messages and

coordinate activities. Communication allows

cooperation and coordination among agents

involving in a conversation.

The following figure represents the three layers

of our prototype.

Figure 2: intermediation System.

4 SYSTEM COMPONENTS

Our aim is to design a system based on agent

technologies able to respect the constraints imposed

by the CoPs environment. These choices are

motivated by the sociological approach claiming that

the social behaviour of the agents is vital to model

all the interactions that occur within a real

organization (Lindemann et al, 2001). Our system is

composed of three components (figure 1), but we

ICSOFT 2008 - International Conference on Software and Data Technologies

280

focuses on the communication model. First of all we

defined some agents of our system.

4.1 Definition of Agents

The agents that we use have the capacities to treat

their own tasks or to solve problems; that mean’s

agent’s have the explicit capacity of knowledge

representation and communication (Huhn, 1999).

Such agents can be adapted to the preferences,

the members’ needs of the community, based on past

experiences and information on the interaction with

these members.

Relatively to the models defined by Chun-Che

and Huang Lai Gu-Hain (Chun-Che Huang and Gu-

Hain Lai, 2004), we built our JAIS agents ( Jade

Agent Intermediation System) based on AOD

models (‘acquiring’, ‘organizing’ and ‘distributed’)

and those which are interactions objects like

‘profiles’ and ‘external members’.

In case of acquiring we define:

(1) An agent CompagnonModerator (AgCM)

whose role here is to collect knowledge from various

sources (documents, databases, other agents).

(2) Task manager agent (AgTM): in charge of

updating the tasks and knowledge. Broadcasted

response to ActivMember collected from the

community to the knowledge base.

In case of organizing we define:

(3)CompagnonModerator agent: ensures the

coordination and dissemination of tasks, monitoring

the finalization of these tasks. When the task is

resolved it transmits retroactively the obtained

answers to the ActivMember. It reproduces the

moderator’s features. It is the mediator between the

intermediation system and ActivMember of the

community. It supports the knowledge integration in

the community knowledge base (this process is

performed with the assistance of the agent

knowledge manager.

(4) Task manager agent: in charge of the queries

execution. It manages interactions throughout

solving tasks.

(5) Domain manager Agent (AgDoM): he has the

features and capabilities to identify the knowledge

base questions depending on the field; it works

closely with the Agent Task Manager.

(6) Knowledge manager Agent (AgKM) took

over the treatment of the knowledge base of the

CoPs and the Agent.

In case of distributing we define:

(7) Dialog manager Agent (AgDM): in charge to

manage upstream exchanges between different

agents of the system (via the technological tools

used in the CoPs). It play interface role, which help

him to treat messages sent from the

CompagnonModerator agent, it sends and receives

messages; he oversees the cycle of discussion in

collaboration with the moderator.

(8) Interface / Expert Agent (AgI): its role is to

interact both with other communities as

ActivMember of the community in which it transfers

knowledge from one practice to another (community

to another community) it also plays an advisory role.

Agent responsible for the agent Profiles:

(a) ActivMember profile Agent (AcMP):

it stores

the information concerning the relationship between

end users and specific knowledge (e.g., Thomas is

interested in multi-agent system).

(b)Agent Domain expert profile (ADExP): it

stores the information concerning the relationship

between the domain expert and specific knowledge.

stores the information concerning the relationship

between the Knowledge Storage Agent and the

knowledge; e.g., the knowledge about informatics is

in the database.

External Members: (a) ActivMember (AcM):

AcM within the community of practice.

In order to allow that these agents interact and

share, knowledge, communication mechanisms are

necessary to implement these processes.

5 COMMUNICATION MODEL

The Intermediation system is composed of different

types of agents. Each of them performs a single task.

Collaboration of agents is necessary. To facilitate

multi-agent coordination and collaboration, it is vital

that agents exchange information via communication

about goals, intentions, results, and status to other

agents. It is crucial that agents agree on the format

and semantics of these messages. Jade follows FIPA

standards so that ideally Jade agents could interact

with agents written in other languages and running

on other platforms.

So, for our case the agent communicate as

following:

(1) The AcM sends a message to request the

AgCM to have access to the knowledge base. The

AN INTERMEDIATION SYSTEM BASED ON AGENTS MODELLING TO SHARE KNOWLEDGE IN A

COMMUNITY OF PRACTICES

281

message requests the data required and the analysis

technique to be used.

(2) The AgCM requests the AgKM via the

AgDM to collects the data from the Knowledge

base.

(3) The AgCM receives the collects data and

send it to the AcM.

(4) The AgDoM requests the ADExP containing

the information on the relationship between the

knowledge and domain expert.

(5) The ADExP replies to the AgDoM with the

information.

(6) The AgDoM sends the analytical information

to the domain expert.

(7) When the domain expert receives the

information, some comments are added based on

expert knowledge. The domain expert submits the

expert knowledge to the AgDM

(8) The AgDM sends the expert knowledge to the

AgCM.

(9) The AgCM sends the expert knowledge to the

Knowledge Storage Profile and requests the

Knowledge Storage Profile to obtain information

about the relationship between knowledge storage

and expert knowledge.

(10) The AKsP replies with the information to

the AgCM.

Figure 3: Agent conversation process.

The AgCM is an interface agent that manages the

member’s access authority and communication with

other agents via the dialog manager. When users

want to belong to the community, the system

presents an interface based on role and authority.

The AgCM is the core agent of JAIS. The members

defined in JAIS can also ask the AgCM to provide

predefined services. The following figure 3 shows

how the different agents interact to exchange

information in the system.

6 SYSTEM IMPLEMENTATION

The Java Agent DEvelopment Framework (JADE,

2006) was developed by TILab. This software

framework uses the agent communication language

(ACL) specifications proposed by the Foundation

for Intelligent Physical Agents (FIPA) and provides

a set of graphical tools that supports the debugging

and deployment phases. JADE supports two types of

agent containers, the main container and the normal

container.

The JADE main container consists of a message

transport system (MTS) used for communicating

with other agents or agent platforms.

Figure 4: class diagram of the intermediation system.

The agent management system (AMS) is used

for managing the agent life cycles such as starting

and stopping, and the agent directory facilitator (DF)

is used to record the services provided by an agent.

Each JADE agent registers itself using the remote

method invocation (RMI) provided by the JADE

main container. Jade was chosen to implement the

JAIS prototype. Agents are created by simply

extending the jade.core.Agent class. The main class

of our system is the JAISAgent class (see figure 4). It

is an abstract class that all different types of agents

must extend. We propose two extensions of the

JAISAgent class: The CommunityAgent and the

IntermediationAgent. These two different types of

agent have different behaviours. The

CommunityAgent represent all the members of the

community of practices include in the organizational

model (Macro Level) and the IntermediationAgent

are the agents include in the Micro level our system.

ICSOFT 2008 - International Conference on Software and Data Technologies

282

7 CONCLUSIONS

In this paper we presented the modelization of our

intermediation system. This model is composed of

organizational model, cooperation model which

includes coordination and interaction model. We

focused on communication model involved in the

coordination model. Interactions between

intermediation agents require the implementation of

the model belonging to the micro level. The result of

our work deals with the implantation of the

intermediation system for CoPs. To do this, the

design of the system includes definition of the agent

communication, the agent behavior, and interaction

protocols.

REFERENCES

Arens, Y., Knoblock, C.A. Shen, W. 1996. Query

reformulation for dynamic information integration.

Journal of Intelligent Information Systems, 6(2):99-

130.

Berney B. and Ferneley E. 1999 CASMIR: Information

retrieval Based on Collaborative User Profiling.

Boudriga, N. and Obaidat, M. 2004. Intelligent Agents on

the Web: A Review, Computing in Science and

Engineering, July/August, 35-42.

Deale, A. 2006. A model for representing professional

development through the participation in a virtual

CoP: uses for developing enhanced services.1st

International Workshop on Building Technology

Enhanced Learning Solutions for Communities of

practice, 2 October, Crete, Greece.

Decker, K. S. Sycara, K. Williamson, M 1997. “Intelligent

Adaptive Information Agents. Journal of Intelligent

Information Systems, 9: 239--260.

Etzioni, O. and Weld, D.S. 1995. Intelligent Agents on

the Internet: Fact, Fiction, and Forecast, IEEE Expert,

10(4), 44-49.

Genesereth, M., Keller, A. and Duschka M.. 1997.

Infomaster: An Information Integration System. In

Proc. Of the ACM SIGMOD Int. Conf. On

Management of Data, pp. 539-542, Tucson, AR.

Hammond M., 2004. Virtual Knowledge Communities for

Distributed Knowledge Management: A Multi-Agent-

Based Approach using JADE. Institut für Algorithmen

und Kognitive Systeme Universität Karlsruhe (TH)

SS.

Henri, F. 2006: CoPs: Social Structures for the

Development of Knowledge. PALETTE Kick-off

Meeting, Lausanne, March 13-15. Also in

http://www.licef.teluq.uqam.ca.

Huhn, M. Stephens. 1999. Multiagent systems and

Societies of Agents. In Weiss, G (1999). Multi-Agent

Systems. MIT Press.

JADE, 2006. (Java Agent DEvelopment framework),

Available at: “http://jade.tilab.com/”, cited 5

November.

Langelier, L., Wenger, E. (eds.), 2005. Work, Learning

and Networked, Quebec, CEFRIO.

Huang, C.C and Liang, W.Y. 2004. The Agent-based

Collaboration Information System of Product

Development, International Journal of Information

Management. 22, 211-224.

Lindemann, V.T. Münch I. 2001.The Role Concept for

Agents in Multiagent Systems. MASCHO’01, Vienna.

Louis, R., 1999. Software agents activities. In ICEIS’99,

1st International Conference on Enterprise Information

Systems. ICEIS Press.

Nodine, M. Fowler, J. Ksiezyk, T. Perry, B. Taylor, M.

Unruh, A. 2000. Active Information Gathering in

InfoSleuth. International Journal of Cooperative

Information Systems, vol. 9, no. 1/2, pp. 3-28, 2000.

Nonaka & Takeuchi, 1995. The Knowledge-Creating

Company, Oxford Univ. Press.

Nwana, H.S. 1996. Software Agents: An Overview, The

Knowledge Engineering Review, 11(2), 205-244.

Odubiyi, J.B. Kocur, D.J. Weinstein, S.M Wakim, N.

Srivastava, Gokey, S., Graham, C.1997. SAIRE – A

Scalable Agent-Based Information Retrieval Engine,

Proc. Autonomous Agents 97, Marina Del Rey, CA.

Papazoglou, M. 2001. Agent oriented technology in

support of ebusiness: Enabling the development of

intelligent business agents for adaptive, reusable

software. ACM. 44(4) 71– 77.

Rigaud E., 2003. Définition et opérationnalisation d'une

organisation virtuelle à base d'agents pour contribuer à

de meilleures pratiques de gestion des risques dans les

PME-PMI. Thèse Informatique temps réel, robotique

et automatique, Pôle Cindyniques, école des Mines de

Paris à Sophia Antipolis [ENSMP].

Roda C, Angehrn, A, Nabeth T and Razmerita I. 2003.

Using conversational agents to support the adoption of

knowledge sharing practices. Interacting with

computers 15(1), 57-89.

Wenger, E. McDermott, R. Snyder, W.M. 2002.

Cultivating Communities of Practice, Harvard

Business School Press, Boston.

Wenger, E. 2004. Knowledge management as a doughnut:

Shaping your knowledge strategy through

communities of practice. Ivey Business Journal, 68, 3.

Wooldridge, M. Jennings. N. 1999. Software Engineering

with Agents. Pitfalls and Pratfalls, IEEE Internet

Computing 3(3), pp. 20-27.

Yildizoglu, M. Dupouët, O., & Cohendet. P. 2004.

Morphogenèse des communautés de pratiques, Revue

d'Economie Industrielle, 103, 91-110.

Weinstein, C. Birmingham.W.P., Durfee, H. 1999. Agent-

based Digital Libraries: Decentralization and

Coordination. IEEE Communication

Magazine.pp.110-115.

AN INTERMEDIATION SYSTEM BASED ON AGENTS MODELLING TO SHARE KNOWLEDGE IN A

COMMUNITY OF PRACTICES

283