Sunteți pe pagina 1din 18

Journal of US-China Public Administration, ISSN 1548-6591

May 2013, Vol. 10, No. 5, 457-474

DAVID

PUBLISHING

Modeling Conflict Dynamics With Fuzzy Logic Inference


Janez Usenik
University of Maribor, Maribor, Slovenia

Tit Turnsek
Faculty of Organisation Studies, Novo Mesto, Slovenia

In the present article, the authors discuss a model of conflict and its interaction with the environment. The conflict
influences the environment, and the environment influences the conflict. The discussion is limited to
intra-organizational conflicts. Many conflicts are solved before they escalate. There are many definitions which
emphasize different aspects and/or types of conflict. If differences arise in a cooperative context, conflict will
probably take a constructive course and will lead to progress. This is typically cooperation. The authors standpoint
is that the cooperation and conflict are two different types of social interactions. The conflict is always destructive.
The gravity of the conflict can be measured as a level of emotional states of actors. The fuzzy logic inference is
selected to model the reaction of the management to the conflict. The focus is on the implementation of fuzzy logic
inference with system dynamics tools. Two questions are discussed. The first question is: how does management
influence the conflict? In other words, how is the fuzzy inference model connected with the conflict model? The
second question concerns delays. There is always a delay between the events in the process (conflict), the
awareness of it, and the reaction of the interested actors from outside. The question is how delays influence the
conflict, and how to build them into the model. In the last part of the paper, the authors present the results of
simulation, which show the influence of the strength of the management intervention and the influence of the length
of the delay between the conflict and the management awareness of the conflict.
Keywords: conflict, stock-and-flow modeling, system dynamics, fuzzy inference

Modeling social interactions based on system dynamics paradigms began early last century, when Lotka
(1925) and Volterra (1926) independently modeled the interactions between predators and their prey in the
form of two differential equations. This provided the roots for the development of models of different types of
social interactions. These have included models of co-evolution (reciprocal activation), diffusion, and military
battles (Epstein, 1997; Sterman, 2000; Hannon & Ruth, 2001; Liebovitch, Vallacher, Nowak, Bui-Wrzosinska,
& Coleman, 2008). The Richardson arms race model can also be classified as a model of reciprocal activation
(Richardson, 1960), and it was formulated with two coupled linear differential equations:

dx
=a ym x+r
dt

(1)

Corresponding author: Janez Usenik, Ph.D., full professor, Faculty of Energy Technology, University of Maribor; research
fields: operations research, system theory, control, fuzzy logic, neural networks. E-mail: janez.usenik@uni-mb.si.
Tit Turnsek, Ph.D. candidate, researcher, Faculty of Organisation Studies; research fields: systems dynamics, conflict theory,
management, fuzzy logic. E-mail: titturnsek@ymail.com.

458

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

dy
(2)
=b x n y + s
dt
where, x(t) and y(t) represent the level of arms, or the arms power, of each of the parties in the conflict, as well
as the perceptions of the threat.
Characteristically, the behavior of one actor depends on the behavior of the other actor, and this mutual
dependency is simultaneous (Huckfeld, Konfeld, & Likens, 1982, p. 47). This is a typical model of conflict.
Later, many studies analyzed this model, and have expanded it in many directions (Epstein, 1997; Bigelow,
2003; Scheffran, 2003; Olinick, 2008). Ruloff (1975) formulated the dyadic conflict between nations in the
form of difference equations, and translated the Strogatz model (1998) into the system dynamics
stock-and-flow model (the STELLA package). His intention was to attract people with limited mathematical
knowledge to dynamic modeling (Radzicki, 1993). Zambetta, Nash, and Smith (2007) reinterpreted and
modified the Richardson model (1967), with this modified model combined with fuzzy logic as an essential
part of an interactive computer game.
A big step forward in the mathematical modeling of conflict dynamics was achieved by the
Gottman-Murray research group (Gottman, Murray Cook, Tyson, White, and Rushe). Gottman and colleagues
developed a formal mathematical model of marital conflicts as the theory (Cook, Tayson, Whit, Rushe,
Gottman, & Murray, 1995; Gottman, Murray, Swanson, Tayson, & Swanson, 2005). Coleman, Vallacher,
Nowak, Liebovitch, and Bui-Wrzosinska developed a mathematical model of conflict between two actors in the
form of two nonlinear ordinary differential equations (Liebovitch et al., 2008). In this Coleman model, the
actors can be individuals or also groups at any level of society, or even nations. Most of these models are
dyadic models, which deal with the conflict between two actors. The further discussion will be based on dyadic
model which arises out of the Coleman model. This model is described in detail in Appendix A (Usenik &
Turnek, 2013).
More often than not, the environment of conflict is not included in a model. No one of conflicts occurs in
isolation. When a conflict grows beyond a certain threshold, it will influence the environment, and the
environment will react to it, and in turn, have an influence on the conflict.
The environment thus tends to influence (dampen, or sometimes also reinforce) conflict. Conflict and its
environment are usually but not necessary, in a balancing loop. This is valid for conflicts at any level, from
organizational, through intrastate, to international conflicts.
The question is: how the dynamical model of conflict can be extended to also take into consideration the
influence of the environment. In the present study, the focus is on organizational conflicts and the reaction of
the management.
Therefore, the model reflects the perceptions, reasoning and reactions of people in conflict situations.
Generally, in modeling social issues where human behavior and decisions are essential parts of the model,
usually vaguely defined values, linguistic terms and expressions, with other words fuzzy variables are used,
instead of crisp values. Peoples reasoning is based on linguistic variables and their linguistic values, such as,
for example, worse, bad, not acceptable, good, and better, and their knowledge about the object of
concern. Linguistic variables and their linguistic values can be represented by fuzzy sets. The relations between
these are described by fuzzy operations. Fuzzy logic is an appropriate methodology to mimic human reactions
to conflict situations. It allows a description of the desired system behavior using spoken language.
Fuzzy logic operates with fuzzy sets, fuzzy variables, fuzzy numbers, fuzzy relations, and so on.

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

459

Fuzzy sets are always functions that map a universe of objects onto the unit interval [0, 1] (Usenik, 2012;
Usenik & Bogataj, 2005). The combination of imprecise logic rules in a single decision strategy is called fuzzy
reasoning, or fuzzy rule-based inference. The reactions and the interventions of the management to conflict is
modeled in accordance with fuzzyrule-based inference.
To model linguistic variables in a context of system dynamics, table functions are usually used (e.g.,
look-up tables in Vensim). Table functions conveniently represent nonlinear relationships which are difficult, if
not impossible, to be expressed analytically (Liu, Triantis, & Sarangi, 2011). An analysis of the advantages and
disadvantages of presenting fuzzy variables with table functions in system dynamics models was carried out by
Liu (2007). There have also been several articles that have discussed fuzzy variables and operations in system
dynamics modeling, of which just to mention Karavezyris, Timpe, and Marzi (2002), Ghazanfari, Jafari, and
Alizadeh (2003), Kunsch and Springeal (2008), Youssefi, Nahaei, and Nematian (2011), among others.
In this article, the approach that was proposed by Lui et al. (Liu, 2007; Liu et al., 2011) is followed. The
fuzzy rule-based inference model and the conflict model are combined into one single model that can be
implemented with system dynamics tools (Vensim).

The Model of Fuzzy Inference Logic


Thus, fuzzy-rule-based inference is the process of the mapping a given input to an output using fuzzy logic.
The output of a fuzzy process can be the logical union of two or more fuzzy membership functions that are
defined on the universe of discourse of the output variable. The last step is the conversion of a given fuzzy
quantity to a precise crisp quantity (Usenik, 2012).
Formally, this is a process of nonlinear mapping of the crisp input vector X T = ( x1 , x 2 ,....x n ) to a crisp
output vector Z T = ( z1 , z 2 ,....z m ), based on fuzzy rules (Ross, 2007), as illustrated in Figure 1. This process
has three successive steps, as illustrated in Figure 2.

Figure 1. Rule-based inference.

Figure 2. A fuzzy-rule-based system.

Fuzzification
Fuzzification reflects the perceptions of the observed object. It maps the crisp input data vector to the
vector of the corresponding input linguistic variables. In this step, fuzzy sets for all of the fuzzy variables (input
and output) and their membership functions are defined (Usenik, 2012). For each component of input vector
i
i
X T , there is one linguistic variable (fuzzy set) x A , i = 1, 2, .......n, with linguistic values (fuzzy subsets)
Ai = { A1i , A2i ,..... Aij } that are defined on the universe of discourse of the input variables. Each linguistic value
i
is expressed by the corresponding membership functions Ali ( x ), i = 1, .....n; l = 1,...... j . Ap is the p
i
linguistic value of the linguistic variable A expressed by the membership function Ai (x ). Fuzzification
p

460

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

maps the crisp values of each component of an input vector to a set of membership values, as:

x i Ai ( x i ) ; i = 1...n; l = 1... j
l

(3)

Fuzzy Inference
Fuzzy inference is a process in which a certain conclusion is derived from a set of fuzzy statements
(Usenik, 2012). The core of a fuzzy-rule-based system is a set of rules that reflect the knowledge about the
object of concern. This knowledge is presented in the form of if-then statements as: if antecedent then
conclusion (Ross, 2007).
Such statements or rules represent the implication of the antecedent (the premise) to the conclusion (the
consequent) (Usenik, 2012). There are a few methods for the implementation of fuzzy inference, and here the
Mamdani method (Ross, 2007) is chosen, which is the method that is often most used. The rule of if-then is
now written in the form of if x is A then z is B, where A is a fuzzy set that represents the antecedent part and
B is a fuzzy set that represents the consequent part of the statement. The term x is A is evaluated to A(x).
Antecedents can often be in the multiple conjunctive form: x1 is A1p and x 2 is A2f ......and x n is Agn ,
which is evaluated to:

= min A ( x1 ), A ( x 2 ), ..... A ( x n )
1

1
p

1
p

2
f

2
f

n
g

n
g ,

or in a multiple disjunctive form x is A or x is A ......or x is A

(4)

which is evaluated to:

= max A ( x1 ), A ( x 2 ), ..... A ( x n ) , [0,1]


1
p

2
f

n
g

(5)

Generally, the antecedent can have the form of any logical expression that presents a fuzzy set A and is
evaluated to a value .
The antecedent implies the conclusion ( A

B). The membership value of the antecedent shapes the

membership function of the conclusion. There are also a few methods that show how to do this. An -cut or
clipping method was used. This method is simple and does not require many mathematical operations. The
membership function of the consequent part is cut at the value of , [0, 1]. Clipping method modifies the
output fuzzy subset (the linguistic value), as represented in Figure 3.

Figure 3. The -cut.


i

1
1
2
2
m
m
The conclusion can also be in a multiple form, as z is B p , z is Bu , ... z is Bl . Bl is the linguistic

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

461

value l (the fuzzy subset) of the output variable z i. All of the membership functions of consequent are cut at
the same value of of the antecedent.
Suppose the fuzzy inference system has r rules and m output linguistic variables. The antecedents are
evaluated to 1 , 2 , ..... r , i [0,1], i = 1,2....r . Every rule contributes one conclusion for each linguistic
variable (fuzzy set) included in the consequent. To combine all of the conclusions for certain linguistic values
into one conclusion, more often the disjunction of the values at which the linguistic value has been cut is used.
i
As an example here, linguistic values Bl have been cut in rule k at the value k , and in the rule i at the
i
value i . Finally Bl are cut at the value max[k, i]. The conclusions of all of the rules relating to one output
linguistic variable have to be combined into the overall conclusion for this variable. The same has to be done
for all of the output linguistic variables. This process is called aggregation, and it is illustrated in Figure 4. The
end result of the fuzzy inference is the combined output fuzzy set for each output variable expressed as their
membership functions ( z i ), i = 1, 2, .....m.

Figure 4. The end result of the aggregation.

Defuzzification
The last step of the rule-based inference is the defuzzification process. The results from the evaluation of
the fuzzy rules are the fuzzy output variables. Defuzzification maps these fuzzy variables to the crisp values
(the crisp output vector). There are many defuzzification methods, and these can give different results. The
most frequently used methods are: First of the Maximum, Largest of the Maximum, Mean of the
Maximum, Height Method, Center of the Maximum, and Center of Gravity (Ross, 2007). In the present
article the Height Method is used, because it is simple and easy to implement with system dynamics tools. The
i

crisp value zc of the fuzzy variable

zi

is the weighted average of maximum:

zci

(z ) z
=
( z)
l

i
Bli

i
Bli

i
Bli MAX

(6)

462

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

j
i
where l = 1......k is the index of the linguistic values of the output variable z i ; Bli ( z ) is the membership

value of the linguistic value l of the output variable z i at which the membership function is cut; z Bli MAX is
the crisp value of the linguistic value l of the output variable z i at the maximum of its membership function.

Figure 5. Defuzzification using the Height Method.

For the example shown in Figure 5, the crisp value will be:

zc =

B ( z ) z B MAX + B ( z ) z B MAX
B ( z) + B ( z)
1

(7)

Implementation of the Rule-Based Model in the System Dynamics Environment


To mimic the reaction of the management to the conflict situation, the two input, one output rule-based
inference system is implemented. The following assumptions have been made:
(1) The crisp emotional levels of both of the actors (variables x and y) are assigned to two linguistic
variables that represent the level of conflict, with the following fuzzy sets: Negligible (Nx, Ny), Small (Sx, Sy),
Medium (Mx, My), and Large (Lx, Ly), which are defined on the universe of the discourse {0, 1, 2 100}. The
corresponding membership functions for the input linguistic variable xc are represented in Figure 6. The
membership functions for the input variable yc are similar;

Figure 6. Membership functions of the input variable xc.

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

463

(2) The output linguistic fuzzy variable Z1 of the fuzzy inference represents the intensity of the
management interference in the conflict. Here, Z1 is represented with the following linguistic values: Warning
(W), Talk (T), Intervention (I) and Change (C), as illustrated in Figure 7;

Figure 7. Membership functions of the output fuzzy subsets.

(3) According to experiences with organizational conflicts, the following can be said. When the conflict is
below a certain level, it is not really seen as conflict, and the management is usually not aware of it. When the
conflict is perceived as present, but as negligible, the management merely warns all of the actors involved.
When the conflict becomes greater, but is still perceived as small, the management reacts with serious talks
with both the sides. When the conflict becomes more serious, the management usually intervenes with very
serious warnings to the actors involved, and/or by changing some of the working and communications
procedures and rules. If the conflict is perceived as large, some deeper personal and organizational changes are
undertaken. Based on the described reactions of the management to conflict situations, the following rules are
defined:
Rule 1: if xc is Negligible and yc is Negligible then z is Warning.
Rule 2: if xc is Negligible and yc is Small then z is Warning.
Rule 3: if xc is Small and yc is Negligible then z is Warning.
Rule 4: if xc is Small and yc is Small then z is Talk.
Rule 5: if xc is Negligible and yc is Medium then z is Talk.
Rule 6: if xc is Medium and yc is Negligible then z is Talk.
Rule 7: if xc is Small and yc is Medium then z is Intervention.
Rule 8: if xc is Medium and yc is Small then z is Intervention.
Rule 9: if xc is Medium and yc is Medium then z is Intervention.
Rule 10: if xc is Large or yc is Large then z is Change.
(4) Defuzzification was implemented using the Height Method.
The paradigm of system dynamics and stock-and-flow methodology to implement the fuzzy rule-based
system were used. The model is presented in Figure 8. For a more detailed description of this model presented
in Figure 8, see Appendix B.
Let the input variables xc and yc grow equally linearly from 0 to 100. The result of this simulation is
presented in Figure 9, which shows the output crisp variable Z as a function of the input crisp variables xc and yc.

464

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

xc

Sx

Nx

R1

R2

Mx

R3

Crisp input variables

yc

Ny

Lx

R4

R5

My

Sy

R9

Zo

Antecedents
Conclusions

R10

R8

R7

R6

Fuzification

Ly

Aggregation

Zr
Zp

Zs

Defuzification

AA
BB
Z

Crisp output variable

Figure 8. The rule-based inference model.

Figure 9. Nonlinear mapping of the input variables to the crisp output.

Conflict Model With Fuzzy Inference System


The task now is to connect the fuzzy inference model described above and conflict model described in
Appendix A (Usenik & Turnek, 2013). But before to do this, the following question should be answered: How
does the management intervene? There is a considerable literature dealing with this subject, for example,
Deutsch, Colman, and Marcus (2006). From the viewpoint of our model, there are two possibilities. The
management tries to convince, motivate, mediate, pressurize, change procedures, etc. All of these are done with
the one goal to calm down the conflict, but without any fundamental changes to the structure that drives the
conflict. This strategy is translated into the model by additional negative rates of changes zz Z , which
diminish the conflict; i.e., the emotional levels of both of the actors. Z (t ) is the crisp output of the fuzzy
inference system, and zz is the strength parameter. This case is modeled below.
There is, however, the second possibility that the management changes the structure of the system. In
terms of the model, this means: the change of values of parameters bi , mi , and ci .

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

465

There is always a delay between the events in the process (the conflict) and the awareness of what is really
going on. The interested outside actors (the stack-holders or management) always realize the situation with
delay.
Some time is also needed to make decisions and to act according to the decisions made. In other words,
the reaction to the event is always delayed.
In the case of ethnic conflicts and international conflicts, this delay can be large. Information delay has
been introduced between the conflict and the fuzzy inference system. All of the system dynamics software
packages implement different delay functions. There are two types of delays: the so-called material delays, and
the information delays (Sterman, 2000; Bossel, 2004). In the Vensim package as an example, nine different
delay functions are implemented (Vensim Users Guide). The third-order exponential smoothing delay, which
is implemented in Vensim by the SMOOTH3I function (Vensim Users Guide), is selected. The effect of this
function is illustrated with the model in Figure 10. Then Figure 11 illustrates the third-order exponential
smoothing delay of the step function. The model is defined by the following equations:
D = 8 delay constant;
delay x = INTEG(in x-out x, 0);
in x = STEP(1, 2);
out x = SMOOTH3I(in x, D, 0).

delay x
in x

out x

D
Figure 10. Model of the delay function.

Figure 11. Effect of the delay function on the input step function.

The conflict model with the fuzzy inference system is presented in Figure 12. The dynamic behavior of
this model depends on the values of its parameters and initial conditions. In the rest of this article, the results of
simulation runs with different values of parameter zz , which reflects the strength of the outside intervention,
are presented. It is shown how the changes in the values of different parameters can change the dynamical
behavior of the model. Sometimes these changes are surprising and unexpected.

466

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

kasnitev y

m2

c2

vh y

b2

iz y
xc

yc

rcy
Sx

Mx

dy:dt

Vx

Nx

Vy

My

Sy

R6

R7

Ny

vpliv y

ifx

vh x

kasnitev x

ify

iz x
R1

R2

R3

R4

R5

R8

R9

R10

x
dx:dt

vpliv x
P

rcx
zu
m1

b1

Zo

zz

Zr
Zp

c1

Zs

AA
BB
Z

Figure 12. The conflict model with fuzzy inference.

The intensity of conflict (the emotional levels x(t) and y(t)) is defined on the scale of 0-100. At the start the
values of the parameters were intentionally selected in such a manner that the model that is not influenced from
the outside (without the fuzzy inference system) had a maximal asymptotically stable point at
x(t MAX ) = y (t MAX ) = 100 . The values of parameters are: g = 10; h = 2; m1 = m2 = -0.03; c1 = c2 = 2.5; b1 = b2 =
0.5, and the initial conditions x(0) = 0.01; y(0) = 0.01. These values remain in the following cases unchanged.

Simulation Results
In the first two cases, there is no delay between the conflict and the fuzzy inference part of model. Only
the intensity of the outside intervention, i.e., the strength parameter zz is varied.
Case 1
The strength parameter is quite large: zz = 0.02. The conflict levels x(t) and y(t) grow equally,
approximately linearly, up to a level where the conflict is noticed (level 10, see Figure 13). The fuzzy inference
system reacts (see Figure 14) strongly enough and stabilizes the conflict at that level. The crisp output Z(t)
oscillates between zero and a specific value (approximately 30).
Case 2
In the second case, the strength parameter is set to a small value of zz = 0.009. As can be seen from
Figures 15 and 16, x(t) and y(t) asymptotically escalate to much higher levels (approximately 73).
In the following four cases, the impact of the delay constant on the dynamical behavior of the model is
demonstrated. The strength parameter is kept constant here, at zz = 0.02.
Case 3
Delay constant D = 140. The result of this simulation is presented in Figure 17.
Case 4
Delay constant D = 70. The result of the simulation is presented in Figure 18.

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

Figure 13. The conflict dynamic with immediate and strong outside intervention (case 1).

Figure 14. The dynamic of the outside intervention (case 1).

Figure 15. The conflict dynamic with immediate and weak outside intervention (case 2).

467

468

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

Figure 16. The phase diagram with immediate and weak outside intervention (case 2).

Figure 17. The conflict dynamic with zz = 0.02 and the delay constant D = 140 (case 3).

Figure 18. The conflict dynamic with zz = 0.02 and the delay constant D = 70 (case 4).

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

469

Case 5
Delay constant D = 5.5. The result of the simulation is presented in Figure 19.
These delays resulted in damped oscillations (cases 3 and 4), as it is expected. The fuzzy inference part
and the conflict part of this model form a negative feedback loop with the delay. The typical behavior of such a
structure is damped oscillations. With the growing delay time, the oscillations are greater, but the asymptotic
stable points xs and ys of the model remain the same (approximately 54) for a broad range of delays (case 2
D = 140, case 5 D = 5.5). Compare Figures 17, 18 and 19.
Case 6
Delay constant D = 5. The result of this simulation is presented in Figure 20.
When the time delay is below a certain value (the critical delay), as it is in this case (D = 5), the behaviour
of the model is changed essentially. The conflict is stabilized at a much lower level, and it oscillates slightly.
The comparison of the dynamics behaviors of the model with no time delay that is illustrated in Figure 13 with
the dynamics behaviors of the model with time delay below the critical delay in Figure 20 show that x(t) and y(t)
stabilize in both cases at approximately the same levels.

Figure 19. The conflict dynamic with zz = 0.02 and the delay constant D = 5.5 (case 5).

Figure 20. The conflict dynamic with zz = 0.02 and the delay constant D = 5 (case 6).

470

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

Case 7
Delay constant D = 70, and zz = 0.009. The result of this simulation is presented in Figure 21.
In this last case, strength coefficient zz is reduced, and time delay constant is set to 70. The comparison of
cases 7 and 4 show that the stable point of the model is higher when the strength coefficient is lower (case 7),
as it can be expected. The damped oscillations are remarkably lower.
Further numerous changes of model parameters can be done. Observations and comparisons of the
dynamics behaviors show the impact of parameter values on behavior of the model.

Figure 21. The conflict dynamic with zz = 0.009 and the delay constant D = 70 (case 7).

Conclusions
The prime aim of the generic model is to include the dynamics of a broad range of different conflicts. In
this study, the question remains open: how to determine the parameters of the model from observations of
specific conflicts; this remains a subject for further research.
In this study, a simplified model of conflict dynamics and fuzzy inference system, as a model of the
management reactions to conflict, is presented. The model is constructed according to the paradigm of system
dynamics. As mentioned, the starting point was the model developed by Coleman et al. (Liebovitch et al., 2008).
Many suppositions have been included, and just to mention in particular: parameters bi, ci, mi, g, and h are
constant, so they do not change when the conflict is ongoing, the linguistic variables are triangular fuzzy sets,
and there is only one outside actor (the stack-holder). In the case of ethnic and international conflicts there
could be a few outside actors, potentially each with different interests. Not every stack-holder is interested in
the resolution of the conflict.
The represented model can easily be extended and modified in different directions. The simulation of the
model exposes a variety of different behaviors that are in general agreement with the theories of conflict, even
if the behavior is sometimes counter intuitive. The model is simple and encourages the management to play
with it, to get a better understanding of the conflict dynamics.

References
Bigelow, D. (2003). An analysis of the Richardson arms race model. Retrieved from http://www.resnet.wm.edu/~jxshix/
math302/bigelow.pdf

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE

471

Bossel, H. (2004). System dynamics simulation modeling, analysis und simulation of complex systems. Norderstedt: Books on
Demand GmbH.
Cook, J., Tayson, R., Whit, J., Rushe, R., Gottman, J., & Murray, J. (1955). Mathematics of marital conflict: Qualitative dynamic
mathematical modeling of marital interaction. Journal of Family Psychology, 9(2), 110-130.
Deutsch, M., Colman, P. T., & Marcus, E. C. (Eds.). (2006). The handbook of conflict: Theory and practice. John Wiley & Sons.
Epstein, J. M. (1997). Nonlinear dynamics, mathematical biology, and social science. Addison-Wesley Publishing Company, Inc.
Ghazanfari, M., Jafari, M., & Alizadeh, S. (2003). An approach to solve fuzzy system dynamics problems. Proceedings from 21st
International Conference of System Dynamics, July 20-24, 2003, New York, USA.
Gottman, J. M., Murray, J. D., Swanson, C. C., Tayson, R., & Swanson, K. R. (2005). The mathematics of marriage. Cambridge:
MIT Press.
Hannon, B., & Ruth, M. (2001). Dynamic modeling (2nd ed.). Springer.
Huckfeldt, R. R., Kohfeld, C. W., & Likens T. W. (1982). Dynamic modeling: An introduction (quantitative applications in the
social sciences). Sage University Press.
Karavezyris, V., Timpe, K. P., & Marzi, R. (2002). Application of system dynamics and fuzzy logic to forecasting of municipal
solid waste. International Journal of Mathematics and Computers in Simulation, 60(3-5), 149-158.
Kunsch, P., & Springeal, J. (2008). Simulation with system dynamics and fuzzy reasoning of tax policy to reduce CO2 emission in
the residential sector. European Journal of Operational Research, 185, 1285-1299.
Liebovitch, L. S., Vallacher, R., Nowak, A., Bui-Wrzosinska, L., & Coleman, P. (2008). Dynamics of two-actor cooperation
-competition conflict models. IACM Meetings Paper. Retrieved from http://www.ccs.fau.edu/~liebovitch/NCMR.pdf
Liu, S. (2007). Three essays on travel demand management strategies for traffic congestion mitigation. Dissertation Submitted to
the Faculty of Virginia Polytechnic Institute and State University in partial fulfillment of the requirement for the degree of
Doctor of Philosophy in Industrial and Systems Engineering.
Liu, S., Triantis, K. P., & Sarangi, S. (2011). Representing qualitative variables and their interactions with fuzzy logic in system
dynamics modeling. Systems Research and Behavioral Science, 28(3), 245-263.
Lotka, A. J. (1925). Elements of physical biology. Baltimore: Williams & Wilkins Co.
Olinick, M. (2008). Stable and unstable arms race. Retrieved from http://f10.middlebury.edu/MATH0500J/Models%20Book
/CHAPTER%2002%20Arms%20Races.pdf
Radzicki, M. J. (1993). Dyadic process, tempestuous relationships, and system dynamics. System Dynamics Review, 9(1), 79-94.
Richardson, L. F. (1960). Arms and insecurity. Pittsburg: Boxwood.
Richardson, L. F. (1967). Arms and insecurity. Chicago: Quadrangle.
Ross, T. J. (2007). Fuzzy logic with engineering applications. John Wiley & Sons.
Ruloff, D. (1975). The dynamic of conflict and cooperation between nations. A computer simulation and some results. Journal of
Peace Research, 2, 109-121.
Scheffran, J. (2003). Calculated security? Mathematical modeling of conflict and cooperation. Germany: Potsdam Institute for
Climate Impact Research.
Sprott, J. C. (2004). Dynamical model of love. Nonlinear Dynamics, Psychology and Life Science, 8(3), 303-313.
Sterman, J. D. (2000). Business dynamics systems thinking and modeling for a complex world. Irwin McGraw-Hill.
Strogotz, S. H. (1988). Love affairs and differential equations. Mathematical Magazine, 61(1), 35.
Usenik, J. (2012). A fuzzy model of power supply system control. Journal of Energy Technology, 5(3), 23-37.
Usenik, J., & Bogataj, M. (2005). A fuzzy set approach for a location-inventory model. Transportation Planning and Technology,
28(6), 447-464.
Usenik, J., & Turnek, T. (2013). Modeling conflict dynamics in logistics system, logistics and sustainable transport. Proceedings
from the 10th International Conference on Logistic & Sustainable Transport, June 13-15, 2013, Celje, Slovenia.
Vensim PLE, Vensim PLE Plus Users Guide Version 5. Retrieved from http://vensim.com/docs/
Volterra, V. (1926). Variations and fluctuations of the number of individuals of animal species living together. Memorie della
Accademia Nazionale dei Lincei, 2, 31-113.
Zambetta, F., Nash, A., & Smith, P. (2007). Two families: Dynamical policy models in interactive storytelling. Retrieved from
http://www.academia.edu/ 1080484/Two_families_dynamical_policy_models_in_interactive_storytelling
Youssefi, H., Nahaei, V. S., & Nematian, J., (2011). A new method for modeling system dynamics by fuzzy logic: Modeling of
research and development in the national system of innovation. The Journal of Mathematics and Computer Science, 2(1),
88-99.

472

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE


Appendix A: Stock-and-Flow Model of Isolated Conflict
System of differential equations:

dx
1
x
= m1 x + b1 + c1 (tgh( h) + 1)
dt
2
g

(1)

dx
1
x
= m1 x + b1 + c1 (tgh( h) + 1)
dt
2
g

(2)

translated into system dynamics stock-and-flow model (SF model) is exposed in Figure A1. The programming package Vensim
PLE was used. The parameters g and h are set to 10 and 2 for both influence functions.
m2
rcy

b2

dy/dt

ifx
c2
c1
ify

x
dx/dt
rcx
m1
b1

Figure A1. Stock-and-flow model of isolated conflict.


Model equations:
Parameters:
b1 = 0.5
b2 = 0.5
c1 = 2.5
c2 = 2.5
m1 = -0.03
m2 = -0.03
Influence functions:
ifx = ((EXP(x/10-2)-EXP(-x/10+2))/(EXP(x/10-2)+EXP(-x/10+2))+1)/2
ify = ((EXP(y/10-2)-EXP(-y/10+2))/(EXP(y/10-2)+EXP(-y/10+2))+1)/2
Emotional levels; initial values: x(0) = 10, y(0) = 15
x = INTEG ("dx/dt", 10)
y = INTEG ("dy/dt", 15)
Rates of changes could be positive or negative. The conflict could escalate or deescalate, but the both emotional levels x(t)
and y(t) has to be greater or equal 0, and less or equal 100.

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE


rcx = m1*x+b1+ify
rcy = m2*y+b2+ifx
"dx/dt" = IF THEN ELSE(x>=100:AND:rcx>0, 0, IF THEN ELSE(rcx<0:AND:x<ABS(rcx), 0, rcx))
"dy/dt" = IF THEN ELSE(y>=100:AND:rcy>0, 0, IF THEN ELSE(rcy<0:AND:y<ABS(rcy), 0, rcy))
Simulation parameters:
INITIAL TIME = 1000
FINAL TIME = 200 The final time for the simulation.
SAVEPER = TIME STEP The frequency with which output is stored.
TIME STEP = 0.025 The time step for the simulation.
Appendix B: Model of Fuzzy Inference System
Model of fuzzy inference system is presented in Figure B1.
xc

Sx

Nx

R1

R2

Mx

R3

Crisp input variables

yc

Ny

Lx

R4

R5

My

Sy

R9

R10

R8

R7

R6

Fuzzification

Ly

Zo

Zr
Zp

Zs

Antecedents
Conclusions

Aggregation
Defuzzification

AA
BB
Z

Crisp output variable


Figure B1. Model of fuzzy inference system.
Model equations:
Fuzzyfication:
Nx = IF THEN ELSE(10<=xc:AND:xc<25, (xc-10)/(25-10), IF THEN ELSE(25<=xc:AND:xc<35, (35-xc)/(35-25), 0))
Sx = IF THEN ELSE(20<=xc:AND:xc<40, (xc-20)/(40-20), IF THEN ELSE(40<=xc:AND:xc<60, (60-xc)/(60-40), 0))
Mx = IF THEN ELSE(50<=xc:AND:xc<65, (xc-50)/(65-50), IF THEN ELSE(65<=xc:AND:xc<85, (85-xc)/(85-65), 0))
Lx = IF THEN ELSE(xc<75, 0, IF THEN ELSE(75<=xc:AND:xc<90, (xc-75)/(90-75), 1))
Ny = IF THEN ELSE(10<=yc:AND:yc<25, (yc-10)/(25-10), IF THEN ELSE(25<=yc:AND:yc<35, (35-yc)/(35-25), 0))
Sy = IF THEN ELSE(20<=yc:AND:yc<40, (yc-20)/(40-20), IF THEN ELSE(40<=yc:AND:yc<60, (60-yc)/(60-40), 0))
My = IF THEN ELSE(50<=yc:AND:yc<65, (yc-50)/(65-50), IF THEN ELSE(65<=yc:AND:yc<85, (85-yc)/(85-65), 0))
Ly = IF THEN ELSE(yc<75, 0, IF THEN ELSE(75<=yc:AND:yc<90, (yc-75)/(90-75), 1))
Fuzzy inference:
R1 = MIN(Nx, Ny)

473

474

MODELING CONFLICT DYNAMICS WITH FUZZY LOGIC INFERENCE


R2 = MIN(Nx, Sy)
R3 = MIN(Ny, Sx)
R4 = MIN(Nx, My)
R5 = MIN(Ny, Mx)
R6 = MIN(Sx, Sy)
R7 = MIN(Sx, My)
R8 = MIN(Sy, Mx)
R9 = MIN(Mx, My)
R10 = MAX(Lx, Ly)
W = MAX(R1, MAX(R2, R3))
T = MAX(R4, MAX(R5, R6))
I = MAX(R7, MAX(R8, R9))
C = R10
Defuzzification:
Zo = IF THEN ELSE(0<W, 30*W, 0)
Zr = IF THEN ELSE(0<T, 60*T, 0)
Zp = IF THEN ELSE(0<I, 90*I, 0)
Zs = IF THEN ELSE(0<C, 100*C, 0)
AA = W+I+T+C
BB = IF THEN ELSE(AA>0, AA, 1)
Z = (Zo+Zp+Zr+Zs)/BB

S-ar putea să vă placă și