A MODEL MULTI-AGENTS FOR SHARING AND EXCHANGING

KNOWLEDGE IN COMMUNITY OF PRACTICES

Kenfack Clauvice

Modeme,Jean Moulin University, 6 cours Albert thomas, Lyon, France

Keywords: Community of practices (CoPs), Sharing and exc

hanging Knowledge, Multi-Agents System (MAS), Agent

Oriented Abstraction (AOA).

Abstract: This paper attempts to show, how CoPs evolve knowledge transfer. We are focusing on the elaboration of a

frame to analyse and underlie the logics of the modalities of functioning of CoPs. We qualify this concept as

an abstract regrouping knowledge creation. By adopting a processual perspective, we will try to present the

mechanisms of sharing knowledge within a CoPs. CoPs is a collection of agents (human beings) who have

rather strong common points such as, their level of social capacity, their competences, and the cognitive

capacities. The development of the exchanges is based on abstract boundaries; the couple

knowledge/community implies that an exchange of information takes place through mechanisms of co-

operation, negotiation and through a specific communication language of community members. However,

the legitimacy of exchanged knowledge is recognized only with interpersonal confidence association that

creates for itself progressively interactions. Besides, these exchanges take place only through rules and

standards established by all the members. After having pointed out the theoretical bases of CoPs and the

sharing knowledge mechanisms, we will present an approach of simulation using the paradigm of the multi-

agents systems, sharing knowledge’s within the CoPs.

1 INTRODUCTION

The concept of community is today at the core of

many organizations which are involved in the

development of the open source software, CoPs and

Epistemic. KM (Nonaka and Takeuchi, 1995)

appears to be a highly interesting, although its

human dimension, more precisely the social aspect

of its transfer process of knowledge is often

neglected. The ability to manage and to share

knowledge could indeed fuel the competitiveness of

organizations. Recent researches suggests the

change of behaviour within a structure (Grant, 1996)

favourable in knowledge transfer. However, in our

research, by community, we mean a group of people

who interact frequently in a direct way, and in a

multi-face process, they weave between them more

or less asymmetric connections. People who always

work together evolve in the way of a community

environment. This suggests that the notion of

cooperation and not competition is one of the

fundamental characteristics of the community.

Bowles and Gintis (Bowles and Gintis, 2000)

underline the importance of the notion of

community, and not the emergence of special

community forms which, according to him, explains

the increasing interest of this concept. Our purpose

concerns the concept of CoPs, which corresponds to

an emergent space where the transfer of knowledge

can be made away from organizational constraints

(Wenger, 1998). Inside such knowledge-producing

community, the members’ behaviour is

characterized by the respect of social standards,

voluntary engagement in construction, exchange,

and sharing of common repertory cognitive

resources, experiments and storytelling. Through

their specific practices, this community can be

considered as a “core” of skills that helps the

hierarchical structures in the construction of

knowledge. This concept of distributed actors’

organization is well studied in the field of the

Artificial Intelligence (AI). More specially the

Distributed Artificial Intelligence (DAI) being

particularly interested in the modelling of intelligent

behaviours entities distributed in an environment.

Multi-agents organizations spread information on

physically distributed sites, the can serve as a

support for the design of distributed information

systems (Yu, 1999).Our research concern

implementation tool intermediation for the CoPs in

247

Clauvice K. (2006).

A MODEL MULTI-AGENTS FOR SHARING AND EXCHANGING KNOWLEDGE IN COMMUNITY OF PRACTICES.

In Proceedings of the First International Conference on Software and Data Technologies, pages 247-252

DOI: 10.5220/0001320402470252

 SciTePress

the aspect of interaction, negotiation, cooperation,

knowledge share. But this short work is strictly

focussing on share, exchange and capitalization of

knowledge. We aim to modelize CoP’s members’

behaviours while exchanging, and sharing

knowledge process. We then define the

characteristics of the agents occurring in the process

of computer adding knowledge sharing. In this step

we will describe how we mean as communication as

action between agent-agent.

2 CoPs, KNOWLEDGE SHARING

& MANAGEMENT

CoPs is a group of agents engaged in the same

practice, regularly communicating through various

mechanisms: Groupware, emails, forums, face-to-

face discussion, and meetings. The use of

information technologies and communication gives

us the possibility to work in an asynchronous or

synchronous mode. The literature on this concept is

large, (Wenger, McDermott&Snyder, 2002;

Josserand&Leger, Vaast, 2004, Soenen, 2004). We

define this concept as: a group of people that

communicate together, to exchange information and

to enrich their knowledge and know-how through

their actions to find a consensus on a subject they

are confronted with. The agents engaged in such a

process coordinate their activities to improve their

individual competences, through the exchange and

share of a common base or individual knowledge,

which is built while the practice of the community is

developing. The collective training and the

construction of new knowledge then appears as a

non-deliberated form of the common practice

(Wenger&Snyder, 2000). CoPs becomes

increasingly through the actions and the repeated

interactions that they maintain between agents which

regularly communicate their experiments and

validate new forms of common practice. It plays the

elementary role of core of skills that Wenger

describes as "Locally negotiated mode of skills"

(Wenger, 1998). The adhesion of the members is

based on cooperative process; all depends on the

type of community. There are communities in

which, the adhesion of a new member requires a

preliminary consultation with other members,

through consensual agreement. The constitution of

the common cognitive capital of the CoPs is made

through permanent sharing of experiments between

the members. The permanent comparison of

individual expertise constitutes the base of the

community and social standards base that guide the

behaviours agents (Brousseau, 2000).As the co-

operative process develops, the increase of the

common cognitive capital contributes to make

stimuli increasingly easy; the frequency of the

interactions while intensifying reinforces the

creation of the social standards and shared routines.

The implicit or explicit mechanisms of search of

legitimacy exploit the behaviour of the agents

belonging to the community. (Dupouët, Yildizoglu,

Cohendet, 2004). Thus, the context of the CoPs

refers to a range of rich behaviours of the agents

which belong to the community. Satisfaction that

they withdraw to exchange together makes it

possible for them to develop a single comprehension

(common language, and practices) in their field. The

process of negotiation is made by a diffusion of the

subjects discussed with the whole community. Roles

can be allocated to members according both their

experiments and degree of confidence. In that case,

attribution of roles can be done by vote or through

consensual way. Furthermore, members of the

community are using technological tools which can

be synchronous or asynchronous. CoPs can be

invisible or institutionalised (Wenger, Mc

Dermott&Snyder, 2002). The members share a

substrate of common knowledge. In this fact, the

concepts of the CoPs are based on three criteria’s

characterizing its operation (Lave&Wenger, 1991)

(Brown&Duguid, Lesser&Everest 2001).To meet

specific needs for a community of Knowledge

practices, we propose to equip it with a knowledge

base. In that regard, the panacea of Knowledge

Management (KM) tools can be exploited to this end

by giving the possibility to organise Knowledge

credits in predetermined conceptual classes in

ontology, by allowing them a more natural and

intuitive access to required knowledge. The KM

constitutes that an important contribution to

schedule knowledge emergent of the interactions of

the members of the CoPs, to show the interest to

mobilize the knowledge management for the CoPs

2.1 The KM Such as the Scheduling

of Knowledge Credit

The generation of knowledge within CoPs can

appear by sharing means, of exchange and

acquisition of knowledge. The more a person shares

her knowledge with somebody, the more certain

confidence grew. In addition, the knowledge

acquisition implies that agent engages in a process

of knowing and expertise research. The knowledge

sharing corresponds to a replication of knowledge

capitalized on the scale of the organization. It

consists in reproducing, exchanging to compare and

making evolving knowledge available to the

members of an organization more specifically of a

community to make a lever of value. Researchers

ICSOFT 2006 - INTERNATIONAL CONFERENCE ON SOFTWARE AND DATA TECHNOLOGIES

248

like (Roberts; Rolland and Chauvel, 2000;

Davenport&Prusak, 1998; Kramer, 1997) see in

confidence a fundamental characteristic in a process

of information’s exchange. The legitimacy of

knowledge in an exchange process is recognized if

the receiver accorded a high degree of confidence to

the sender. It consists in sending messages, request

or information in a store space. These spaces contain

a group of knowledge or cases to be solved.

Epistemological approach: The step that we

wish to adopt consists in identifying the

characteristics of the context in which the

knowledge share and exchange takes place, for that

we shall base on an initiative of modelling /

conception based on the systemic analysis (Durand

1979, Donnadieu, 2002, LeMoigne, 2002), the

socio-constructivism approach, and the collective

cognition and the constructivist knowledge design

(Avenier, 1997), The socio-cognitive conflict

(George, 2001). Finally, the approach of collective

cognition allows the individuals engaged in a

common practice to develop negotiation

competences, confrontation and argumentation. The

goal is to achieve to a group decision marker design.

In such a process, agents accumulate knowledge and

competence progressively with practice. These

mobilized theories thus aim at highlighting us in the

step of modelling of the adopted approach. What

will interest modelling consist in

including/understanding not only operation but also

obviously the motivations of the members in the

process of sharing?

3 THE EMERGENCE OF

ANALYSIS MODEL

Adhesion within a community, presupposes certain

recognition by its pairs and by the organization.

Wenger argues that one of the fundamental

characteristics to belong to a community of practice

relates to a certain mutual confidence between the

members of the community. From that point, we see

the importance of analysing here the role of

confidence in knowledge sharing to a community.

We will talk about the interpersonal confidence that

exists naturally between two individuals and whose

determinants are competence and reputation.

Organizational Model: There is a type of

community in which the control is strongly

distributed and whose members have the same skills,

and pursue multiple goals. Nowadays, after the

presentation of the theoretical and practical bases of

the communities of practices, we aboard now this

part to approach the knowledge management. Our

work consists in to trying to show how communities

of practice are managing their substrate of

knowledge by a multi-agents approach.

A multi-agents approach for Knowledge

sharing in a community of practices: The MAS uses

the social metaphor of the insects, reactive agents

(reactive SMA) or of the human organizations

cognitive agents (AI-like). According to this

paradigm, agents interact in order to accurately,

carry out actions ordered by an external agent. In

most cases, the agents were expected to coordinate

their actions and sometimes to cooperate in order to

achieve their goal although having different

motivations. The negotiation by the agents consists

to coordinate, to share limited resources or to solve a

conflict while agreeing on a solution in which their

respective interests are as well as possible satisfied.

The models of negotiation implemented generally a

language of communication and a protocol of

negotiation to conceive a diagram of the interactions

between agents, like the agents reasoning capacities,

to modelling their work procedure to carry out their

strategies. The languages of communication define

some rules to carry out the information exchange

between agents. These rules relate to aspects located

at the low level of the communication between

agents and can, for example, specify the structure of

the messages or the actions of communication (Com,

2002a). The Agent Oriented Abstraction (AOA)

paradigm was introduced into Multi-agents systems

(MAS) with an aim, to propose an agent abstraction,

i.e. not founded on extensions of the principles of

the paradigm Oriented Object. In this approach, an

agent is an autonomous entity with annotated

knowledge and a mechanism of decision based on

this knowledge. It makes it possible to approach the

KM by considering the entities of the organization,

their knowledge and their capacities of exchanges.

The exploitation of this paradigm can be considered

in a uniform way for any organization’s entity. From

a generic point of view, knowledge is a variable,

simple or complex, which can as well represent, a

collection of data, a document, an authorization, a

competence. Coordination remains as much a

significant concept because we can use them to

exploit individual interactions. Indeed, the interest to

aboard knowledge management within a CoPs based

on multi-agents approach lies in compatibility and

convergences between these two fields. If it is

considered that an agent represents an entity in the

community carrying information or knowledge, then

we can approach basic concepts characterizing

knowledge within the community: knowledge and

data distribution, autonomy of the entities and

simultaneously complex interactions between them

(negotiation, share information, coordination),

A MODEL MULTI-AGENTS FOR SHARING AND EXCHANGING KNOWLEDGE IN COMMUNITY OF

PRACTICES

249

dynamic behaviours, heterogeneity from points of

view.

Interest of an agent for capitalization of tacit

knowledge: Usually capitalization of knowledge

implies the constitution of a capital that may

produce benefits in time. Knowledge capitalization

techniques have been already developed in

distributed artificial intelligence (DAI) directly

applicable to handling knowledge. Indeed, some

agents have competencies, which may vary from

simple operations to various reasoning. Those agents

may be have in various way procedural behaviours

and contain much knowledge. Cognitive agent

seems to be an interesting support to realize the

social system that we mobilize. Those cognitive

agents can be specialised by insufflating new

knowledge for a restricted field. He can also be

associated with other specific agents which purposes

are ranging from keeping knowledge to memorising

the tasks. The developed agent is able to manage

knowledge, according to its expertise field. This

gives the possibility to agents to control the

development of its knowledge base. Instead, to have

only one KM, we allocated to each agent her own

knowledge base. This can be cooperating if faced

with a complex problem solving. For a specific

domain, it will be necessary to create concept

models to represent this field. Such systems of

concepts are called ontology. We showed, the

knowledge or exchanges within a community are

done through roles, which are allocated to each

member. This section describes initially how roles

are formed, and in the second place, how roles can

be configured and be allocated in terms of actions.

Roles, agent, and attribution of the roles:

Several roles can be defined within the community,

in a general way these roles are allocated by a

regulator jointly indicated agreement with the whole

of the members see Agent-group-role model and role

attribution (Parunak, Odell, Fleischer, 2003). A role

is a class, which defines a normative behavioural

repertory of an agent (human or artificial). It

provides modules for the social systems of agent and

the conditions by which the agents act together.

Each agent communicated to other agents according

to functional conditions of the system. Several

methodologies of design agents were proposed, we

studied two of them: the proposal of (Ferber, 1999),

and (Jennings and Wooldridge, 2000). A role in

Aalaadin (Ferber, 1999) is an abstract representation

of the function, service or quite simply the identifier

of an agent within a group. Each agent can hold

several roles, but each role is local with a group. The

communication between the agents is not possible

through the roles, which they assume, and

consequently, the group carries out control on the

communications. Roles can be defined

independently of the groups and played within these

same groups. Moreover, the roles can have

associations of knowledge with other roles,

indicating this interaction. Our proposal is mainly

limited to two major aspects. The first one is what

we called Role formation and the second one is

called Role configuration

Role formation: in this case roles can be

assigned to the agents in a multi-agents system in

two manners: endogenous (by self-organization

emergent like system), and exogenic (by the

originator of the system when the system is built or

modified). The endogenous self-organization is a

phenomenon spread in the normal systems. While

agents act together, structures and models on the

emergent level system can adapt and be robust to the

changes of the environment system. At the

beginning, the behaviour could have been the

actions of a simple individual. With time, however,

the actions of the individual could be identified as

together useful behaviour which can be used by

other individuals to produce similar results.

Role configuration: can be considered

according to two dimensions: Horizontal

specialization that addresses the number and

horizontal complexity of the actions supported by a

role. Contrary to the vertical specialization, which

separates the execution from the actions of the

administration, in another word, it takes place

according to the degree of command, which an agent

can have above its actions and the actions of other

agents. Specifically, horizontal roles specialization

requires an end, a role for agent can require handling

any kind of request, and for vertical specialization it

relates to managing the action of the agent, more

precisely deals with the degree of control of the

actions of the system.

3.1 The CoPs like Support of

Knowledge Creative

We define a community of practices as a creative

community of knowledge formed of a grouping of

agents carrying a common interest for a specific

subject and exchanging knowledge in bond with this

subject. This definition is well situated in the

knowledge management of an organization through

the AOA approach. Gathering agents implies

existing dynamic capacities of adhesion and

participation in the communities (Calmet & Maret,

2000). At least, an agent must have dynamic

capacities to create communities. The idea of a

common interest for a specific subject implies the

evaluation of this subject by each agent, taking into

ICSOFT 2006 - INTERNATIONAL CONFERENCE ON SOFTWARE AND DATA TECHNOLOGIES

250

consideration its objective. In the same way, the

concept of knowledge bonds with the subject of the

community requires the evaluation of a distance

between knowledge and a subject. The knowledge

exchange (consists in sending messages in a store

space, a request or information, these spaces contain

a group of knowledge or cases to be solved), imply

that agents can acquire and transmit knowledge.

They must only have the capacity to interiorise

received knowledge and to exteriorise/diffuse own

knowledge: from where the idea to equip each agent

with a knowledge base. The primitives that we can

associate to the members of our community can be:

created/finished a community (creates, delete) if the

need for creating a under-community is made feel

(in the case of the reorientation of the objectives of

the community), to join/leave a community (Join,

Leave), to send a knowledge or to require a

knowledge (inform, request).We consider several

models and process of operation of the CoPs. The

principal constraints that we wish to impose are to

ensure that the knowledge obtained by the agents

during their interactions is not centralized, the

autonomy of the agents, and the safeguarding of the

opening of the system of agents, i.e. the possibility

of entry and exit of agents without damaging the

system. We wish to use an approach of "one to

several" to implement the knowledge sharing. Each

agent would have, in addition its own knowledge

base, a list of agents if it wants to transmit

knowledge or requests. Agents can everytime join

communities (which he knows). We wish to develop

prototypes (Java, Jade platform or others) of the

simulation programs as well as applications with

user interface: agent responsible for the knowledge

management, agent responsible for the diffusion of

information or the allocation of the functions,

system of information’s exchange. In term of

knowledge we propose: Creation and enrichment

knowledge, research centres and pro-active

dissemination of knowledge service, presentation or

visualization knowledge service, evaluation

knowledge service, maintenance knowledge service,

Administration -knowledge service.

The agents that we can elaborate in our

Proposition are:

Safety agent: dynamically, he manages to

control access to the database of the system. This

knowledge is imported from consensual decisions

made by the agents.

Moderator agent: ensure the coordination and

the diffusion of the tasks, the follow-up the

realization of these tasks, as well as the integration

of knowledge in the database of the community. Lay

down the objectives of the group, the topics of

discussion and definite a scenario of collaboration of

the agents. It has moreover a list of the agents of the

system.

Community initiator agent: dedicated to the

agent chosen as leader, initiation consists in creating

a subject of debate, sending messages and making

known the community. This action is done within a

space dedicated for this purpose. All the agents of

the system are members of the community. These

messages consist of inform (transmission of

knowledge) and request (request for knowledge).

The evaluation of the contents of the messages is

specific to each agent. No agent centralizes the

exchanges.

Evaluator agent: ensures the evaluation of

knowledge likely to be stored in the knowledge base

dedicated to the community, and then transmits it to

the regulator. Supports the self- evaluation and the

motivation to be shared, evaluates if the objectives

were achieved. The agent, which wishes to insert

new knowledge, can require the authorization to the

regulator of the community.

Interface agent: are agents who belonging to

several communities can transfer knowledge from a

practice to another, we can consider them as experts

playing an advisory role.

Profiles agent: in charge to manage user

profiles (name, firstname, competences,

experiments....), works in close cooperation with the

moderator agent. In order, without neglecting the

problems of the interoperability of information

systems, the model presented positions in the

context defined by the CoPs (Wenger, 1998). Those

diffuse knowledge while integrating environments

supporting training intention. And, they can be a

group of individual authors and users of knowledge,

an industry of conceptual tools, a consortium of

diffuser of formalisms or a community of free

software developing the components of a system of

remote formation. While referring to the

fundamental characteristics of the communities of

practices we could say that a community is defined

by a triplet < C, D, P >, where C is the community

of the actors, D is the field of competences, and P

contains the questions raised or prone to discussions

of the community from which new practices

emerges from. For the agents, a CoPs bases itself on

the structure of communication which exists

between the agents, we can advance that the

behaviours of the agents can consist of two large

shutters: an engagement in a practice and an

engagement in social exchanges. However, these

two shutters are dependent insofar as the

communication informs the practice and the practice

feeds the exchanges.

The communication as action: Following

upon the works of Austin and Searle (Austin 1962,

Searle 1969), we lean on a pragmatic frame, where a

A MODEL MULTI-AGENTS FOR SHARING AND EXCHANGING KNOWLEDGE IN COMMUNITY OF

PRACTICES

251

statement is considered as an action - due to the

change which it entails in the world -and where its

meaning bases on the knowledge shared by the

interlocutors and on the intention which each has

and gives. At the conclusion of a phase of

negotiation, a consensus is found; the negotiated

concepts belong to the knowledge shared by the

interlocutors. It is thus clear that the communication,

which bases on the shared knowledge, also

contributes to spread this knowledge, via the

mechanisms of negotiation.

4 CONCLUSION

We have presented the methodological and

technological choices to conceive and implement the

concepts relating to our study. The multi-agents

systems through their models and technologies offer

a support that we can use to conceive and establish

in relevant manner the social system of which we

showed the characteristics previously. Readings

concerning CoPs are showing interests of such

practice, but also difficulties faced by then members.

A CoPs is a social system as well as agents in MAS

paradigm. MAS’s models and technologies fulfilled

in relevant manners, CoPs characteristics.

The idea is, then, to propose a platform

intermediation tool, based on MAS, which

reproduce a social system in order to facilitate Cops

functioning. From this step, we ha to implement the

intermediation tool, on for example JADE or

MADKIT platform. This system will help member

of community to communicate together or with

members of another community. The

communication system which we will used is based

on the act of language developed by Searle and

Austin (Searle 1969, Austin, 1962).

REFERENCES

Austin, 1962. How to do things with words. Oxford

University Press, UK.

Avenier, M-J., 1997.

Une conception de l'action

stratégique en milieu complexe: la stratégie

tâtonnante. La stratégie chemin faisant. Economica.

Gintis 2000. Social Capital and Community

Governance”.Working Paper n° 01-01-003, Santa Fe

Institute.

Brousseau E. 2000. What Institutions to Organize

Electronic Commerce? Private Institutions and the

Organization of Markets, Economics of Innovation

and new Technologies, 9 (3), July-September, pp. 245-

279.

Communication F.T., 2002a. FIPA ACL Message

Structure Specification. Proceedings Report.

Connelly C. & Kelloway K., 2000. Predictors of

Knowledge Sharing in Organizations. MSc Thesis

Queen’s University, Kingston ON.

Davenport T. H. & Prusak L., 1998. Working knowledge:

how organizations manage that they know », Harvard

Business School Press.

Donnadieu G., Karsky M., 2002. La systémique penser et

agir dans la complexité. Editions Liaisons. Université

du Maine, France.

Durand D., 1979. La Systémique, Puf.

Ferber, J. & Gutknecht, O., 1999. Operational semantics

of role-based agent architecture. In Proceedings of

ATAL'99.

George S. 2001. Apprentissage collectif à distance.

SPLACH: un environnement informatique support

d'une pédagogie de projet. Thèse de doctorat,

Université.

Grant R.M., 1996. Toward a Knowledge- based theory of

the firm. 17: 109-122.

Josserand E. & Saint Léger B. 2004. Les difficultés

pratiques des communautés de pratique. 13e

conférence de l'AIMS. Normandie. Vallée de Seine 2,

3 et 4 juin.

Lemoigne J.-L., 2002. Le Constructivisme - TOME I –

Les enracinements. Collection Ingénium, L'Harmattan

2002.

Maret, P., Calmet, J., 1996. Un modèle multi-agents pour

la gestion des connaissances.

Nonaka & Takeuchi, 1995. The Knowledge-Creating

Company. Oxford Univ. Press.

Odell, J.J., Van DykUser e Parunak, H., Fleischer, M.,

2003. Modeling agent organizations using roles.

Springer-Verlag .

Roberts J., 2000. From know-how to show-how?

Questioning the role of information and

communication technologies in knowledge transfer

Technology. Analysis & Strategic Management.

Rolland N. & Chauvel D., 2000. Knowledge transfer in

strategic alliance.Butterworth Heinemann, Boston,

MA, p 225-236.

Searle 1969. Speech Acts. Cambridge University Press,

UK.

Soenen G., 2004. Effective communities of practices: a

comparative study. Working Paper, EM Lyon.

Wenger E. & Snyder W. M., 2000. Communities of

practice: the organisational frontier. Harvard Business

Review, Vol 78, N°1.

Wenger E. 1998. Communities of Practice: Learning as a

Social System. Systems Thinker, June 1998.

Yu, E. 1999. Strategic Modelling for Enterprise

Integration. Proceedings 14th World Congress of the

International Federation of Automatic Control, July 5-

9, Beijing, China.

ICSOFT 2006 - INTERNATIONAL CONFERENCE ON SOFTWARE AND DATA TECHNOLOGIES

252