TOOL FOR THE EVALUATION OF INNOVATIVE THERAPIES

Multi-agent based System

Anastasiya Shtiliyanova, Fabien Feschet

ISIT, Clermont Universit

e, 28 Place Henri Dunant, Clermont-Ferrand, France

Pascal Pommier

Centre L

eon-B

erard, D

epartement de Radioth

erapie, Lyon, France

Keywords:

Innovative therapies, Modeling, Evaluation, Multi-agent systems.

Abstract:

The paper contains the main lines of the modeling of a prototype for both medical and economical evaluation of

radiotherapy centers using innovative therapies. We consider uncertainties according to patients and hospital

complexes participating in the study. The corresponding part of the tool is called recruitment model. It

evaluates a theoretical number of patients interested in treatments of one center, using spatialization into a

context of competition. Moreover we developed a scheduling model, described brieﬂy here and two shortly

presented economical models in charge of the evaluation of the average price per treatment and the comparison

between different radiotherapy strategies. We present preliminary results of the recruitment model.

1 INTRODUCTION

1.1 Context and Objectives

Innovative therapies used by centers involve un-

certainties according to the installation of facili-

ties and/or the expected therapeutic gain, after the

treatment (as hadrontherapy, stereotaxy, cyberknife).

Medical beneﬁts results are often unknown, and pa-

tients are hostile to choose between a new therapy and

a well known treatment, thus it is a hard decision to

equip a complex, using only economical methods.

We model a generic progressive prototype evalu-

ating the position of a center from both medical and

economical point of view. To proceed we ﬁrst evalu-

ate a theoretical number of patients interested in these

therapies, then its medical position regarding compe-

tition and medical politics ﬁxed in advance. A recruit-

ment model is in charge of the theoretical number of

patients. It uses a suitable scheduling model to plan

the internal organization of the hospital, giving us the

possibility to have a good response for the demand of

a patient. The general prototype consist of two eco-

nomical models, one evaluating the price per treat-

ment of a therapy in a hospital, and another one used

to compare different strategies of treatment between

them and with the cost reimbursed by the health insur-

ance of a country. Moreover we include the cost per

year gained after a treatment according to a strategy of

treatment. The economical part of the prototype is not

described here. The tool supports the development of

innovative therapies and medical structures.

1.2 Modeling Method

The tool is composed by the four independant models,

thanks to links between them they could be used al-

together (Fig. 1). We implemented them in a generic

way using the multi-agent theory (L.R.Coutinho et al.,

2005) and techniques such as (J.Ferber et al., 2004;

J.F.H

ubner et al., ; Wagner, 2003), to describe the

global system, the interactions and organizational in-

formation system. This provides us a ﬂexible organi-

zation (Nicholas, 2000).

Figure 1: Interaction between mains models.

Let us explain the execution of the global schema:

1. The recruitment model(RM) generates pa-

tients and spreads them out to suitable centers. Items

are located into the geography, thus the distance be-

tween them is known. The RM calculates a theoretical

number of recruited patients per complex. One time

step is equal to one week, according to the common

148

Shtiliyanova A., Feschet F. and Pommier P..

TOOL FOR THE EVALUATION OF INNOVATIVE THERAPIES - Multi-agent based System.

DOI: 10.5220/0003598901480154

In Proceedings of 1st International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH-2011), pages

148-154

ISBN: 978-989-8425-78-2

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

hospital organization.

2. The RM uses the scheduling model (SM) which

gives the response of acceptance or not for the current

patient, according to all constraints (functioning, ca-

pacity, medical politics). The two models are corre-

lated, and the last one could be replaced easily.

3. The use of the economical model (EM) is pos-

sible iff the evaluated center knows the number of re-

cruited patients by disease. It is estimated by the RM

model or done directly by the user. This model com-

putes the cost per treatment using the Activity based

costing method (ABC, 2011; I.Durand-Zaleski, 2005;

K.Kesteloot et al., 2000). We modeled, with medical

experts, the description of a treatment composed by

two phases (preparation and irradiation), each one is

presented by activities included in.

4. The medico-economical model estimates a

price per strategy, using a price per method of treat-

ment (protocol) included in one strategy. One strategy

is composed by multiple protocols describing the path

followed by the patient during his treatment. The user

could use the EM to get a price per treatment or give

an estimation by himself. In addition we compare it

with the cost corresponding to the French Health In-

surance System.

The global representation of the system is close

to the Information Systems As Agents (Hayes-Roth,

1995). Each model could be associated to an institu-

tional agent (Wagner, 2003), thus the communication

between agents is done by the transfer of information

or messages (Genesereth and P.Ketchpel, 1994).

The four main parts are designed as separate

agents, implemented in JAVA. They are acting in a

common environment, represented by a knowledge

dictionary and constructed especially for the tool.

This dictionary embraces all deﬁnitions for different

domains (geography, medicine, epidemiology, econ-

omy). We built up the ontology, the kernel of all

links between the four models, using the XML, XSD

properties, to control the quality of the required infor-

mation. The collective interpretation allows common

discussions and interactions. In addition the global

brute space controls the dictionary (J.Ferber et al.,

2009). Moreover the Social level, the knowledge,

is a part of the Structural Speciﬁcation of the tool

(J.F.H

ubner et al., 2002).

In the rest we present the recruitment model, rules

used to built it up and sub-processes developed in col-

laboration with medical experts. The elaboration pro-

cess is: a response with feasible results is needed for

every situation. We used the medical knowledge to

construct basic rules, which are represented like a n

node tree. A node is a rule and edges are possible ac-

tions to undertake, regarding the constraints. When

the complete tree is constructed, and not resolved

situations exist, we use the main rules again. The

process is repeated until all situations are resolved.

This mechanism belongs to the Cased-Based Rea-

soning methodology (A.Aamodt and E.Plaza, 1994;

S.Onta

non and E.Plaza, 2008). The process of col-

lecting all required data for the prototype is complex,

thus a simpler versions of some hypothesis exist.

2 THE GENERATOR OF

PATIENTS

The generation of patients is a process lead by the RM

during every time step. It depends on the simulated

geography and indications (deseases), and makes the

link with the process of research of centers. It as an

agent, composed by multiple mechanisms, described

in the two following subsections.

2.1 Geographic Position

To represent the epidemiology data the geography is

composed by different level of details.

Deﬁnition: Geography is an hierarchical structure,

composed by items. Items are the smallest non partitionable

elements (Fig. 2).

Every geographic element is referenced by an in-

dication. Epidemiological data couples with well

known statistical laws permits us to generate patients

according to an indication and an item.Thus generated

patients know their geographic position.

Geography

Figure 2: Elements of the geography and Incidences. We

subdivide the geography into indivisible zones, which are

as small as possible.One element of the geography refers to

several indications.

Deﬁnition: Incidence is the number of patients ac-

cording to one geographic element (Fig. 2). Incidences

compose the generated population for every time step.

2.2 The Disease Description of a Patient

The patient is described by a couple of [indica-

tion/protocol](IP). The protocol is the description of

the treatment according to technical constraints and

the planning of all sessions. As multiple protocols

TOOL FOR THE EVALUATION OF INNOVATIVE THERAPIES - Multi-agent based System

149

could be coupled with one indication, the chances

for the patient to ﬁnd one treatment are larger. We

suggest a method to classify them in a speciﬁc order,

which facilitates the patient to ﬁnd a center.

Deﬁnition: Priority group The couple IP is char-

acterized by a priority group, combination of the medical

beneﬁt and the proof of the innovative therapy.

The priority group deﬁnes a mutable order for

treatments. The patient initially starts the research of

centers with its best priority group.

We suggest a ﬁnal date for patient, obliging him

to ﬁnd a center for a treatment before. It allows the

patient to ask centers for a treatment multiple times if

he is not accepted immediately. We introduce in this

way the multi-demand of the patient to centers. So

the ﬁnal date for the treatment depends on the genera-

tor, which acts at the beginning of every time step. Its

discussions with the recruitment model create the pa-

tients in an environment of competition of radiother-

apy centers. This is an institutional agent operating

with biological agents.

3 THE PROCESS OF RESEARCH

OF CENTERS

The process of research of centers is based on tech-

nical requirements of the disease of the patient, his

general healthy state, his preferences to the existing

offer and geographical position of centers.

3.1 Description of the Center

The choice of the patient is inﬂuenced by possible

treatments, available facilities in centers, their geo-

graphic position and their medical politics.

1. The Mechanism of Medical Politics

Deﬁnition: Medical Politics represents preferences

of a center to the accepted indications. When it exists, it

gives priority of some treatments and/or eliminates others,

composed by several rules being a part of the planning.

Deﬁnition: Reserved Time represents reserved slots

in the hospital planning for a number of patients described,

by special characteristics as an indication and a protocol.

We suggest some rules for medical politics, de-

ﬁned as independent structures and not essential for

the functioning of the center, as well as the existence

of the mechanism of medical politics.

- A Reserved Time for a Couple of IP. The center

constitutes a group of couples, composed by reserved

slots for IP treatments. This way the center gives pri-

ority to treatments with a better therapeutic gain.

- A List of Priorities, with reserved slots for one

or multiple priority groups. The last two rules could

introduce a conﬂict situation. If the priority group and

the couple of the same IP have reserved time, the med-

ical knowledge gives priority to the rule of the couple.

- limit the Usage of Particles such as Carbone, for

hadrontherapy centers. We estimate the annual use of

the particle and control it to not exceed a ﬁxed limit.

2. The Scheduling Model

The scheduling model is a simple model, which plans

treatments and calculates the probability of accep-

tance of one patient according to its time of waiting.

We estimate the probability of acceptance with a mo-

bile mean computed for each IP accepted in a center.

A response is then given to the patient.

We constructed the SM in order to be able to test

the RM. The number of treated patients is maximized

regarding medical politics. We used linear program-

ming in 0, 1 to implement it. Constraints are not listed

here, but we illustrate the use of the model on the Fig.

3. The SM is used whenever the recruitment model

needs to have a response for one patient. It also plans

treatments for patients already in the waiting queue

and when it is not possible, the patient stays in the

waiting queue considering his preferences. The SM is

used for planning treatments for each room and dur-

ing the whole simulation, we use the RLE encoding

(RLE, 2010). It matches each day with the planning

of a room. When the scheduling model doesn’t ﬁnd

a place for one patient, not corresponding to medical

politics, he is rejected directly and he has to search

for another center. As depicted on Fig. 3 there are

some rules to use during the discussion between the

SM and the agent of the center. We can identify them

in the same way as it has been done in (Wagner, 2003;

Taveter and Wagner, 2001).

Planing of the Centr

updates

Waiting queue

scheduling model

according to

the capacity of

the centre

Day

time

rooms

external pool of patients

patients with

treatment

patients staying

in the waiting

queue

rejected

patients

Figure 3: Interaction between the SM and the RM. We make

updates during the planning of treatments, in consequence

coming patients are evaluated and accepted or rejected.

3. The Waiting Queue

The waiting queue is a consequence of the mechanism

of medical politics. When a patient being a part from

medical politics arrives with a demand, and no place

is available for him, than if he could ”wait” according

to his ﬁnal date of treatment, he goes in an waiting

queue if he agrees. The situation of this patient is

evaluated at every time step (Fig. 4). Moreover the

entry of new patients could force the exit of a patient.

A special way of ordering is applied to the waiting

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

150

time

Begining of

the simulation

process

birth

searching

for a center

evaluation of the acceptance in the waiting queue

by the center at each time step

going out of

the waiting queue

probability of

treatment 55%

probability of

treatment 40%

probability of

treatment 10%

entering in a

waiting queue

time

entering in a

waiting queue

evaluation of the acceptance in the waiting queue

by the center at each time step

treatment

probability of

treatment 55%

probability of

treatment 70%

probability of

treatment 100%

birth

searching

for a center

Begining of

the simulation

process

Figure 4: Evaluation of the state of the patient in an waiting

queue, rejection or successful issue.

queue, considering the entry date and the priority

group of the patient. The center keeps unchanged the

medical politics and the priority of patients, which

are always planned in the slots of reserved time.

The waiting queue implies waiting possibilities for

a patient. When a center does not consider medical

politics, the waiting queue does not exist and the

patient will not wait for a response.

4. Choice of Patients by the Center

We suggest a mechanism organizing demands of pa-

tients. For every time step of the simulation the mech-

anism groups demands for the same center. Thus the

center receives a group of patients and applies medi-

cal politics. The possible responses are:

- How many patients could be taken for an im-

mediate treatment. Patients evaluated to be probably

placed in the planning are part of medical politics.

When there are places available after the setting of

those patients, the other patients are set up.

- Number of patients directed to the waiting

queue.

- Number of rejected patients, when they are not

a part of medical politics and there is no place in the

planning for them.

The response by the center is a result of the interac-

tion of the agents of medical politics, the SM and the

RM. They are using the common knowledge and ac-

cording to the environment they discuss on the possi-

ble actions to undertake. The full description of the

center is a part of the mind of the center reﬂection the

decisions taken during discussions with other agents

in the system.

3.2 Characterize the Distance and the

Attractivity

As the geographic distance is an important parameter,

we deﬁne crowns of inﬂuence over each center repre-

senting the hesitation of the patient.

Deﬁnition: Crown of Inﬂuence (Zone) is composed

of one or multiple geographic elements. Over the center one

can ﬁnd multiple crowns. The inﬂuence is smaller when the

crown is distant from the center.

In consequence when the patient is in the ﬁrst

crown the center is more attractive, than if he is in an-

other crown (Fig. 5). The patient prefers centers situ-

ated nearly, thus we introduce a list of center around

him, ordered according to the geographic distance.

This model of crowns is mutable.

We mentioned the healthy state of the patient, in

consequence the medical beneﬁt of the treatment de-

pends on. This medical parameter is part of a mecha-

nism called attractivity.

Key

is a patient

Figure 5: Spatialization for patients and center.

Deﬁnition: Attractivity is a combination between the

level of proof the treatment and the priority group of the

couple IP attributed to a patient.

We use the attractivity coefﬁcient as a test if a pa-

tient wants to go to one center capable to treat him.

The result of the use of the geographic distance in the

attractivity, is a list obtained by a patient for all cen-

ters around him and capable to treat him. The mech-

anism of the attractivity is a part of the process of re-

search of centers, which is the leading agent of the

RM. All links for this process are connected with the

global environment of competition (Fig. 5).

3.3 The Patient and his Actions

We use the laws of Poisson and Gauss from the Epi-

demiological data to generate patients with a couple

of IP attributed to every one. The patient changes

his states during the simulation. He appears in the

beginning of one time step and starts searching of a

radiotherapy centers. When he founds one, he could

go in for a treatment, using the attractivity coefﬁcient.

In some cases the patient does not ﬁnd a hospital

complex, because he does not suit to the medical

politics included in the study and/or there is no avail-

able place. For well known techniques the patient

is treated out of the system, thus we evaluate only

the participation of new and innovative techniques

in the therapy center. The next ﬁgure 6 resumes the

set of states of the patient during the simulation,

related to the RM and the process of search of centers.

TOOL FOR THE EVALUATION OF INNOVATIVE THERAPIES - Multi-agent based System

151

Wait

Acceptance

treatment

time

Birth

Placed

treated out

of the system

Questioning

query

change or get a

new priority group

Figure 6: States of the patient: genesis, searching process,

waiting queue, acceptance of a treatment and the modiﬁca-

tion of the priority group.

1. The order of Couples Describing the Disease

and the Treatment, The Choice of Lower Priority

Group

One indication is coupled to more than one protocol,

thus the patient has more chances to ﬁnd a treatment.

The research of centers starts with his best couple IP.

All the couples are ordered according to their priority

group, describing a different ways of treatment. We

model the behaviour of the patient as a multiple de-

mands to all centers corresponding to the treatments

(protocols), of the given priority group. When this sit-

uation appears the patient evaluates the responses of

each center and send them back a response. The eval-

uation is based on the date of treatment done by the

center, the distance and the attractivity. When the pa-

tient is accepted by one center and he accepts to go in,

the other centers with a positive responses are alerted

by a mechanism, lead by the RM.

When the patient does not ﬁnd a center capable to

treat him, he could degrade the current priority group,

taking the next available in his list. If a lower priority

group does not exist, then he has to search with the

current one, or he has to choose to take a well known

treatment and to be treated out of our system.

To model the degradation of the priority group,

we use a decision parameter based on a probability

law, combination of the number of weeks available

for waiting and the priority group. An example of the

graph of this curve is shown on the Fig. 7. In the

example there are 51 time steps period and 6 priority

groups.

In consequence the probability law is used by the

agent of the patient as a measure to estimate his need

to change his priority group and search for new treat-

ments. Less time available implies a faster choice of

lower priority group.

2. The Waiting State, Waiting Queue and Priority

Groups

The waiting state of the patient corresponds to an

awaiting queue. This is possible when the patient is

scouring for center and he is a part of medical poli-

tics of one center. If the designed center does not ﬁnd

Figure 7: Curve of the probability function used by the pa-

tient to estimate the degradation of his priority group.

enough place for a treatment, then the patient goes au-

tomatically in a waiting queue. When multiple centers

are available for the patient, and he is a part from their

medical politics, then he goes in the waiting queues

of all centers. The ﬁrst complex that ﬁnds a place

for a treatment is referred for the patient. When this

situation appears the other centers are informed by a

message that they have to liberate the place.

The evaluation of the place of the patient at every

time step is a process described in the previous sec-

tion for the center. When the center evaluates that the

patient has no chances to be treated before his limit

date, the patient is forced to quit the waiting queue. In

addition when the patient has available time to wait,

he could stay in the waiting queue of the hospital for

several weeks. When he waits too long time, he could

decided to exit it. This mechanism makes a link be-

tween the waiting time of the patient and its priority

group. When the patient is in a waiting queue, he

could choose to stay in and start to search centers for

a less advantageous treatment. This action degrades

his priority group but gives him more chances for a

treatment (Fig. 7). A patient in an queue, is described

by a passive state, i.e. he doesn’t change its priority

group. At each time step his need to pass from the

passive state to a searching state is evaluated using

the probability law. Thus the agent uses a rule, which

validation implies the change of the state. When he

passes in an active state, i.e. keeps its current place in

the queue, he can try to ﬁnd centers with a new lower

priority group ( Fig. 8).

Waiting state and the

decision of the patient

Figure 8: The rule for changing the state of the patient.

The patient is a biological agent (Wagner, 2003))

acting into the RM with his beliefs and goal. His men-

tal components (Shoham, 1993) force him to discuss

and interact with other agents present in the global

system. The environment creates the resources and

the conditions for resolving the situations that ap-

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

152

pears. According to the context the mind of the pa-

tient decides for the best action to do (J.Ferber, 1999;

Dignum, 2009).

4 THE GLOBAL PROCESS, FIRST

RESULTS AND CONCLUSIONS

The global process of the recruitment model is an in-

stitutional agent, he guides all discussions (R.Searle,

1995). The rules are illustrated on the Fig. 9. We use

the schema introduced by (Wagner, 2003). The algo-

rithm modeled for the RM is:

- Initialization and preparation of the environment.

- Generation of patients. The patient is described

by four elements: geographic element, indication,

protocol and priority group.

- Creation of a list of centers, for every patient, ca-

pable to accept him and according to the geographic

distance and the attractivity.

- Updates for the waiting list of all centers.

- The medical politics are activated for centers.

- Choice by the patient from the available centers.

- The SM plans treatments of accepted patients.

Center

Medical Politics

Scheduling Model

requestPlace

confirmPlace

asking

Patients

responding

Center

Waiting Queue

requestUpdates

confirmUpdates

selfUpdates

rejectedPatients

planPatients

Generator of patients

Population of patients

Figure 9: Rules according to the general institutional pro-

cess of the recruitment model.

The ﬁrst results obtained by the prototype use data

from the three French radiotherapy centers: CPO (C),

NICE (N) and ETOILE (E), with corresponding ca-

pacities 489920 hours, and tree treatment rooms, for

C and E, and 16554 and one treatment room for N.

The geography included in the study is the territory of

France, divided into 23 geographic elements. There

are 24 indications and 22 protocols with average du-

ration for the whole treatment spread out over 10 ses-

sions. The couples IP are organized into 8 priority

groups. We use both, the recruitment and the schedul-

ing model for one year simulation time. Our results,

based on a week with 5 days of working time, show

that when there is more time before the start of the

treatment, the recruitment is more important. Details

are shown in the next tables. We present results re-

garding the full capacity and according to every pri-

ority group, being a part of medical politics of the

centers. The results present different situations of re-

cruitment with 0

, 3

and 20

weeks of delay before

the treatment. The tables are also ordered according

to the possible number of patients recruited for the

ﬁrst priority group. When the limit of recruitment of

patients for the ﬁrst priority group is more important,

the recruitment increase. CPO and ETOILE increase

the number of treated patients for different values of

accepted capacity of the ﬁrst priority group. We sug-

gest that medical politics have to be planed carefully,

in order to have a well structured recruitment.

Table 1: Treated patients by centers for one year. The ﬁrst

priority group is 12% of the global planning for C and E.

PG C E N C E N C E N

1 30 1 84 35 10 97 28 11 105

2 26 19 - 24 12 - 26 23 -

3 - - - - - - - - -

4 23 33 - 63 115 - 125 113 -

5 11 23 - 90 132 - 0 165 -

6 - - - - - - 70 - -

7 140 327 - 524 797 - 477 902 -

8 - - - 25 50 - 31 49 -

Table 2: Treated patients by centers for one year. The ﬁrst

priority group is 42% of the planning for C and E.

PG C E N C E N C E N

1 50 43 95 49 42 105 47 34 117

2 42 46 - 42 47 - 49 61 -

3 20 - - 18 - - 27 - -

4 24 76 - 63 123 - 127 138 -

5 12 22 - 87 153 - 97 160 -

6 - - - - - - - - -

7 152 368 - 607 917 - 522 1151 -

8 - - - 22 41 - 27 51 -

1. Conclusion

We use concepts from multi-agent systems for the

construction of a reusable and independent tool pre-

sented in this paper. The different models used for

it are presented as separate agents, thus they are

reusable and could be replaced. The link between

them is done by the ontology presented in the global

environment. The dictionary is related to multiple

ﬁelds such as medicine, economy or informatics. All

input data respect the structure of the ontology and

results were validated by medical experts.

The recruitment model, the main part of the pro-

totype, takes into account the situation of competition

between radiotherapy centers and the spatialization

of centers and patients in the geography. With our

model it is possible to evaluate the right geographic

location of a treatment center in order to optimize the

recruitment. We show preliminary results obtained

TOOL FOR THE EVALUATION OF INNOVATIVE THERAPIES - Multi-agent based System

153

by the collaboration of the recruitment model and the

scheduling model. In future works we plan to replace

the existing scheduling model with a complex one,

including deﬁnitions of staff and taking into account

critical resources.

Compare to the actual economical methods we

evaluate the efﬁciency of a center using its potential

recruitment. Our prototype could be used as a deci-

sion making tool for new innovative facilities, thus al-

lowing to test different scenarios with respect to staff

or treatment capacity as well as medical politics.

REFERENCES

(2010). RLE, http://en.wikipedia.org/Run-length encoding.

(2011). Activity based costing.

http://en.wikipedia.org/wiki/Activity-based costing.

A.Aamodt and E.Plaza (1994). Case-Based Reasoning:

Foundational Issues, Methodological Variations, and

System Approaches. AI Communications, 7:39–59.

Dignum, V., editor (2009). Multi-agent Systems: Semantics

and Dynamics of Organizational Models. IGI.

Genesereth, M. R. and P.Ketchpel, S. (1994). Software

agents.Communication of the ACM, 37(7):48–53.

Hayes-Roth, B. (1995). An architecture for Adaptive Intel-

ligent Systems. Artiﬁcial Intelligence, 72:329–365.

I.Durand-Zaleski (2005). Economic evaluation of radio-

therapy: methods and results. Cancer/Radiotherapy,

pages 449–451.

J.Ferber (1999). An Introduction to Distributed Artiﬁcial

Intelligence. Addison Wesley.

J.Ferber, O.Gutknecht, and M.Fabien (2004). From Agents

to Organisations: an Organizational View of Multi-

agent Systems. Agent Oriented Software Engineering

(AOSEIV), (LNCS 2935):214–230.

J.Ferber, T.Stratulat, and J.Trainer (2009). Towards an in-

tegral approach of organizations in mutlti-agent sys-

tems: the MASQ approach.

J.F.H

ubner, J.S.Sichman, and O.Boissier. S-moise+: A

middleware for developing organised multi-agent sys-

tems. International Workshop on Organizaions in

Multi-agent Systems: From Organizations to Organi-

zation Oriented Programming, pages 107–120.

J.F.H

ubner, J.S.Sichman, and O.Boissier (2002). A Model

For the Structural, Functional, and Deontic Speciﬁca-

tion of Organizations in Multiagent Systems.

K.Kesteloot, Y.Lievens, and E.Schueren (2000). Improved

management of radiotherapy departments through ac-

curate cost data. Radiotherapy and Oncology, pages

251–262.

L.R.Coutinho, J.S.Sichman, and O.Boissier (2005). Model-

ing Organiszation in MAS: A Comparison of Models.

First Workshop on Software Engineering for Agent

Oriented Systems.

Nicholas, R. (2000). On agent-based software engineering.

Artiﬁcial Intelligence, pages 277–296.

R.Searle, J. (1995). The Construction of Social Reality.

Shoham, Y. (1993). Agent Oriented Programming. Artiﬁ-

cial Intelligence, 60:51–92.

S.Onta

non and E.Plaza (2008). Learning Information Ex-

change, and Joint-Deliberation through Argumenta-

tion in Multi-agent Systems. OTM Workshops, (LNCS

5333):150–159.

Taveter, K. and Wagner, G. (2001). Agent-oriented enter-

prise modeling based on business rules. In Proceed-

ings of the 20th International Conference on Concep-

tual Modeling: Conceptual Modeling, number 2224

in Lecture Notes In Computer Science, pages 527–

540. Springer Publishing London.

Wagner, G. (2003). The agent-object-relationship meta-

model: towards a uniﬁed view of state and behavior.

Information Systems, 28(5):475–504.

SIMULTECH 2011 - 1st International Conference on Simulation and Modeling Methodologies, Technologies and

Applications

154