MULTI-AGENT ORGANIZATIONAL MODEL FOR
ECONTRACTING
Benjamin G
ˆ
ATEAU, Djamel KHADRAOUI, Eric DUBOIS
CRP Henri Tudor
29, Av. John F. Kennedy, L-1855 Luxembourg
Olivier BOISSIER
SMA/SIMMO/ENSM Saint-Etienne
158, cours Fauriel, F-42023 Saint Etienne Cedex 02
Keywords:
Multi-Agent System, Organizational Model, eBusiness, eContracting, Sanction, Normative Specification.
Abstract:
This paper covers development and analysis tools, software and system architecture engineering and devel-
opment methodologies. It introduces the MOIS E
+
(Model of Organization for multI-agent System) model
seen under three points of view : structural, functional and deontic. In practical, this model is available as a
JAVA component. The original contribution of the paper is the extension of the model to take into account
the notion of sanctions. These ones are necessary in order to control the respect of normative specifications
(obligation, permission, prohibition) of behaviors. The results of the generalized model are implemented on
an eCommerce application dealing with eContracting.
1 INTRODUCTION
Nowadays only conducting eBusiness is still not ob-
vious because of numerous obstacles. To assist users
in conducting these transactions, we often associate
them with software. In our works, an agent-based sys-
tem is proposed due to the natural decomposition and
abstraction of intelligent agents that mimic a human
organizational structure and behavior. We can say
that the need for intelligent agents has never been as
greater as pressures to make business grow. Research
and development activities in this area increased due
to the market demands as well as systems needs to
solve various and complex problems related to the do-
main.
This paper offer a complete organizational model,
called MOISE
+
(Model of Organization for multI-
agent System) (Hubner et al., 2002), that considers
the structure, the functioning, and the deontic rela-
tions among them to explain how Multi-Agent Sys-
tems (MAS) organization collaborates for its purpose.
The objective is to control agents behaviors and re-
spect of rights and duties described by deontic rela-
tions. This model practice is represented by a JAVA
component which provides an empty definition of an
agent able to organize itself regarding the society and
others agents.
2 MULTI-AGENT SYSTEM
DOMAIN
Within Multi-Agent Systems (MAS) domain, an
agent is a real or virtual entity which interacts with
a shared environment or directly with others agents
by perceiving and acting. Its behavior can be consid-
ered as autonomous because of its ability to have a
local self-control (Jennings et al., 1998). The capac-
ity of each agent to determine its own goals expresses
the concept of agent autonomy. These goals can be
determined according to environment configuration,
agents motivations, agents included in the system or
contracted engagements.
Multi-Agents System can be defined as a dis-
tributed system composed by a set of agents interact-
ing according to cooperation or competition modes.
Nevertheless the agents of a MAS do not compose a
collection of juxtaposed entities but a structure of en-
tities able to work together to solve a problem too dif-
ficult for an agent left alone (Chaib-draa et al., 2001).
Consequently we can say that there are different kinds
of dependencies networks between agents.
Interfering managements and conflicts are the main
problems of this domain research due to agents au-
tonomy. These lasts relate to shared environment and
common resources between the agents and to agent
striving towards personal goals that can be incom-
patible, complementary or in competition. Sophis-
489
Gâteau B., Khadraoui D., Dubois E. and Boissier O. (2004).
MULTI-AGENT ORGANIZATIONAL MODEL FOR eCONTRACTING.
In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 489-492
DOI: 10.5220/0002657404890492
Copyright
c
SciTePress
ticated models of such conflicts management have
been elaborated in the domain. These models re-
veal themselves for example by protocols of negoti-
ation (Zlatev, 2002), organizational models (Dignum
et al., 2002), etc.
Our work deals with the development of models
that are common to the whole agents. These models
define and structure the issues of autonomous agents
overall activity. The objective contains two steps:
the first one is the construction of contract models,
scenarios or institutions in which the common rules
governing the operation of the agents could be im-
plemented in a declarative way. The second one is
the construction of models and architectures of agents
able to conform to these rules. These problems are ap-
proached in a context of open and dynamic systems.
The objective of our research activity takes into ac-
count the development of a set of models in order
to constrain and control the behavior of autonomous
agents while making the agents reasoning and pre-
serving their autonomy towards constraints interpre-
tation. We present in the next section an extended
model of MOISE
+
to introduce the notion of sanction.
It is necessary in order to control the respect of nor-
mative specifications (obligation, permission, prohi-
bition) of behaviors.
3 MOISE
+
ORGANIZATIONAL
MODEL DESCRIPTION
MOIS E
+
is composed by two core notions: an Orga-
nizational Specification (OS) and an Organizational
Entity (OE). An OE is a population of agents func-
tioning under an OS. We can see an OE as an instance
of an OS, i.e., agents playing roles defined in the OS
(role instance), aggregated in groups instantiated from
the OS groups, and behaving as normalized in the OS.
Following this trend, a set of agents builds an OE by
adopting an appropriate OS to easily achieve its pur-
pose. MOISE
+
OS is formed by a Structural Speci-
fication (SS), a Functional Specification (FS), and a
Deontic Specification (DS). Each of these specifica-
tions will be presented in the following subsections.
3.1 Structural Specification (SS)
In MOI SE
+
three main concepts, roles, role relations,
and groups, are respectively used to build the individ-
ual, social, and collective structural levels of an orga-
nization.
As shown on the Fig. 1, the SS of a MAS organi-
zation is formed by a set of roles, a set of group root
specifications (which may have their sub-groups, e.g.
the group specification team), and the inherited rela-
tion on the set of roles.
back
leader
goalkeeper
player
middle
attacker
coach
soc
1..1
3..3
0..1
0..1
4..4
3..3
1..1
1..1
1..2
1..1
acq
com
aut
aut
aut
acq
aut
com
compat
intra-group
key
min..max
composition:
inheritance:
role
Abs Role
inter-group
links
1..1
sub-groups scope:
group
defense
attack
team
Figure 1: A structure example of a soccer team
3.2 Functional Specification (FS)
The FS in MO IS E
+
is based on the concept of mis-
sions (a set of global goals) and global plans (the
goals in a structure). These two concepts are gath-
ered in a Social Scheme (SCH) which is a goals tree
where the root is the SCH goal and where the respon-
sibilities for the sub-goals are split into missions.
Two types of SCH are necessary to express the var-
ious missions:
The social diagram of execution which defines the
missions to be used for the achievement of the root
goal.
The social diagram of sanctions which defines the
missions to be used in the sanction part of the de-
ontic specification (see next section).
3.3 Deontic Specification (DS)
The DS explains the way the organization contributes
to achieve common goals. The DS is the link between
the SS and the SF by defining relations between roles
and missions. The deontic specification of the con-
tract expresses rights and duties of the participants
in terms of obligations, permissions and prohibitions
with time constraints. For instance with a deontic ex-
pression, we can express the fact that a role has the
obligation to execute a mission by meeting a deadline.
In order to avoid agents lack of deontic expressions
respect coming from their ability to reason and to de-
cide autonomously, each deontic expression is asso-
ciated to sanctions (penalties and rewards). An agent
which does not respect an obligation is punished and
ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS
490
if the MAS provide such service its reputation could
be prejudiced. These sanctions constitute tools to
guide the agents behaviors and to obtain a quite stable
society.
agent
role
mission
scheme
goal
group
FS
SS
DS
OS
OE
Figure 2: Organizational Entity
Within the MOISE
+
extension we achieved a sanc-
tion is a special deontic expression with a condition.
A sanction could be positive or negative. As said in
the previous section, we could have a social diagram
of sanctions because these special deontic expressions
are linked between them. The root deontic expression
is generally an obligation. According to the respect
of the obligation or not there is a positive or negative
sanction. According to the result of this sanction there
is another sanction, and so on.
The structural, functional and deontic specifica-
tions (SS, FS and DS) constitute an organizational
specification (OS). An organizational entity (OE) is
an OS instance and the result of agents playing roles.
Fig. 2 summarizes the actual organization model.
4 CASE STUDY :
ECONTRACTING
APPLICATION
Our aim goal is to apply our model on an eCommerce
platform used to trade intangible assets. It is basically
an eContracting application dedicated to translation
services, as described in (Khadraoui and Dubois,
2003)). The model of contract presented hereafter
is based on the organizational model MOISE
+
.
Within this framework a contract is seen like an
Organizational Entity i.e. like the assignment of an
Organizational Specification with a set of agents.
The OS could be seen as a template of contracts. It
defines constraints on groups, roles and missions.
A contract is made up of:
an unique identifier;
an header which gathers:
a list of agents’ identifiers participating in the
contract and playing roles appearing in the SS;
a list of services involved in the contract with
attributes specification defined during the con-
tract negotiation. Services improvement appears
in the FS as goals;
a list of products involved in the contract with
attributes values defined in the contract defini-
tion phase. The action of producing these goods
appears as goal in the FS.
some general conditions;
The contract general conditions represent the
whole requirements relating to the provided
services or products. The negotiation step of the
contract introduces partly the definition of these
attributes.
a Structural Specification (SS) :
The structural specification of the contract ex-
presses the organization in term of roles (Employe,
Employer or Arbitrator for instance), links be-
tween roles and groups.
a Functional Specification (FS) :
The contract functional specification describes
missions to carry out for the services or products
realization. The FS consists of social diagrams.
a Deontic Specification (DS) :
The DS is a set of deontic expressions which can
be an obligation, a permission, a prohibition or a
sanction
1
:
hDSi ::= ‘(’ hdeonticExpi*‘)’
hdeonticExpi ::= ‘(’ :id hidDEi ::’[hDConditioni ]
hdeonticReli hroleIdi
‘(’hdeonticActi hrelationi hdeadlinei‘)’‘)’
hDConditioni ::= hDConditioni &&hDConditioni |
‘Respect(’ hidi ‘)’ | ‘NonRespect(’ hidi ‘)’ |
Valid(’hactionIdi‘)’ | ‘NonValid(’hactionIdi‘)’
hdeonticActi ::= ‘do’‘(’(hmissionIdi hargumenti|
hactionIdi hargumenti)‘)’
A deontic expression hdeonticExpi consists in a con-
dition hDConditioni and in an action that is realized if
and only if the condition is met.
1
We use the BNF meta-language (Backus-Naur Normal
Form).
MULTI-AGENT ORGANISATIONAL MODEL FOR E-CONTRACTING
491
The field hdeonticReli defines the deontic relation that
is activated if the condition is satisfied. It may take
the values ’O’ for an obligation, ’P’ for a permission
and ’I’ for a prohibition.
The hroleIdi refers to the role defined in the SS :
the agent playing this role will have to behave
according to the deontic relation that precedes this
role with respect of the concerned deontic action.
This last is defined with reference to the missions
(hmissionIdi) or to the action (hactionIdi) detailed in
the FS. We added a temporal expression to this
deontic relation. For the moment we only used
the precedence operator < but an extension is
possible.
This deontic specification enables us to specify and
structure the contract execution. The setup of a con-
tract is done by defining the header, the general condi-
tions and the structural, functional and deontic speci-
fications. In other words we describe the elements of
the contract, the organization of the participants, the
clauses and finally the contract execution methods.
Thanks to the structure of MOISE
+
our model pro-
posal introduces independently the structural part and
the functional part of it. The structural specifica-
tion defines the roles and the obligations linking these
roles within the contractual group. The FS defines
the whole missions which will be involved in the con-
tract. Missions are organized in the social diagram
of execution which represents the contract execution
structure. Other missions are located in the social di-
agram of sanction which specifies the ones related to
the sanctions. This allows distinction between normal
and exceptional scenarios of the contract (in case of
non-respect of engagement). All necessary elements
to the implementation of the contractual relation was
defined beforehand in the header and the general con-
ditions of the contract. Finally the deontic specifi-
cation establishes the link between the structural and
functional aspects to define the various scenarios of
execution with an adequate representation that per-
mits the course of the sanctions and to reason on the
consequences of the non-observance of a clause.
5 CONCLUSION
We presented in this paper a new approach that pro-
vides an organizational model for Multi-Agent Sys-
tems. The concept of organization has been described
through three views: an structural view (group, role
and relations between them), a functional view (plans,
goals and missions) and a deontic view (association of
roles with missions involving a deontic operator and
a concept of sanction in case of non-respect of com-
mitments).
The notion of sanction has been added to the initial
MOIS E
+
model introduced in (Hubner et al., 2002).
This constrains the agents behaviors and controls their
autonomy. Actually an agent that has to execute its
missions can loose commitment because of others
missions allowing as well the achievement of local
goals and takes priority. In order to avoid a commu-
nity of airy agents, a reputation could be built on the
sanctions associated with validation or not of an exe-
cuted execution.
REFERENCES
Chaib-draa, B., Moulin, B., and Jarras, I. (2001). Agent
et syst
`
emes multiagents. In et Y Demazeau, J. B.,
editor, Principes et architecture des syst
`
emes multi-
agents. Hermes, Lavoisier.
Dignum, V., Meyer, J., Weigand, H., and Dignum, F.
(2002). An organizational-oriented model for agent
societies. In International Workshop on Regulated
Agent-Based Social Systems: Theories and Applica-
tions (RASTA’02) at AAMAS, Bologna, Italy.
Hubner, J., Sichman, J., and Boissier, O. (2002). Moise+:
towards a structural, functional, and deontic model
for mas organization. In Proceedings of the first in-
ternational joint conference on Autonomous agents
and multiagent systems, pages 501–502, Bologna,
Italy. ACM Press. ISBN 1-58113-480-0. URL:
www.lti.pcs.usp.br/moise/.
Jennings, N., Sycara, K., and Wooldridge, M. (1998). A
roadmap of agent research and development. Au-
tonomous Agents and Multi-Agent Systems, 1(1):7–
38.
Khadraoui, D. and Dubois, E. (2003). B2b econtract so-
lution for teleservices. In International Conference
on Intelligent Agents, Web Technologies and Internet
Commerce (IAWTIC2003), pages 12–14, Vienna, Aus-
tria. ISBN 1740880692.
Zlatev, Z. (2002). Examination of the negotiation domain.
Technical report, EEMCS department, University of
Twente.
ICEIS 2004 - ARTIFICIAL INTELLIGENCE AND DECISION SUPPORT SYSTEMS
492