=Paper=
{{Paper
|id=Vol-2577/paper19
|storemode=property
|title=Choice Manipulation in Multicriteria Optimization Problems
|pdfUrl=https://ceur-ws.org/Vol-2577/paper19.pdf
|volume=Vol-2577
|authors=Hryhorii Hnatiienko
|dblpUrl=https://dblp.org/rec/conf/its2/Hnatiienko19
}}
==Choice Manipulation in Multicriteria Optimization Problems==
https://ceur-ws.org/Vol-2577/paper19.pdf
234
Choice Manipulation in Multicriteria Optimization
Problems
© Hryhorii Hnatiienko1[0000-0002-0465-5018]
1 Taras Shevchenko National University of Kyiv, Kyiv, Ukraine
g.gna5@ukr.net
Abstract. The problems of manipulating the choice of decision options in
situations of peer review are considered. The description of the task of expert
evaluation in the form of a tuple. The possibilities of manipulation of choice are
presented. Particular attention is paid to the problems of multicriteria
optimization. The formulation of the CSF is described with emphasis on its
components and heuristics, which can be used for manipulation. The
classification of selection manipulation problems in expert evaluation situations
that are formalized in the class of multicriteria optimization problems is
proposed.
Keywords: expert evaluation, decision making, manipulation, heuristics,
multicriteria optimization.
1 Introduction
In today's world, the problem of choice manipulation is becoming more urgent. After
all, the problems of choice and decision-making arise in a wide variety of fields, and
the possibility of choice is the basis of democracy. At the same time, it is well known
that manipulation is a type of hidden management that provides ample opportunity to
influence the distribution of resources in a wide variety of fields of human activity.
Therefore, the study of manipulation opportunities for different decision-making
models is currently a relevant and promising area of research. The study of
manipulation problems is the most important because in modern conditions it is often
not the case that the use of scales in measuring some phenomena turns into a trivial
assignment of numbers. And not at all, declared by the researchers to measure the
subjective components of the decision-making problem.
The manipulation of collective choice problems (voting theory) is well researched
and the theory of the agent [1] is developed for such problems as manipulation of the
"agenda", that is, the sequence of application of selection procedures.
Even natural, well-founded and logical voting systems are not protected from
deliberately influencing election results, that is, from electoral manipulation. It is also
known [2] that the choice rule is protected from manipulation if and only if it is
dictatorial.
Copyright © 2019 for this paper by its authors. Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
235
In addition, in many peer review situations, it is advisable to set not one scalar quality
assessment of the objects, but a set of quality indicators and consider the problem in a
vector formulation.
In this article we will consider the choice manipulation problems that arise when
formalizing practical problems in a class of multicriteria optimization problems. The
criterion approach is a common approach in solving decision-making problems. This
approach to describing poorly structured models is associated with the assumption that
each object can be estimated by a specific number, which is a criterion value, so
comparing the objects is a comparison of the corresponding numbers.
Often, multicriteria is a way to increase the adequacy of goal description. At the
same time, as will be shown below, multicriteria offers ample opportunity to manipulate
choices.
Manipulation can be in the interests of different individuals and can be carried out
by different participants in the process of selection and decision making:
– the voting participant does not use false criteria, but false criteria in order to
achieve a collective decision that is more favorable to him by his true criteria [1];
– the organizer of the vote influences the partitioning of the voting participants into
favorable subsets or splits the set of variants into such subsets which as a result will
lead to the desired choice: due to their structure or sequence of their presentation [1];
– the decision maker (ODA) or ODA group can initiate or perform the manipulation;
– the expert or coalition of experts can significantly influence the results of the
selection and manipulate the choice of options;
– the voting party may sometimes be able to manipulate the election, affecting in
different ways each of the categories of voting participants mentioned above.
Given the existence of different sources of manipulation, we will refer to the person
or group of persons who tries to influence or influence the results of the choice by
applying some non-standard approaches, the term manipulator.
The purpose of manipulation can be different: choose the desired alternative, select
the subset of objects, remove the unwanted alternative, remove the unwanted subset of
alternatives, make some alternative a priori dominant, get rid of the influence of the
dominant alternative, etc.
2 Expert evaluation tasks
Its symbolic representation in the form of a motorcade is used to describe the expert
judgment (SEA) [3]
,
where A is the set of objects (variants, alternatives, parameter sets), S is the set of
constraints, R is the set of criteria, measurement scales by criteria, mapping the set of
acceptable alternatives into the set of criterion estimates, the convolution of criteria, E
is the set of formal characteristics of experts, C is the set of goals that the researchers
face, P is the system of preference for the deciding element: the decision maker (ODA),
several ODAs in collegial decision-making, or a group of experts.
Suppose that a set A, consists of n objects that are compared with each other:
236
ai ∈ A, i ∈ {1,..., n} = I .
Objects are described by m parameters, that is, each object is a point of some
m
parametric space Ω
ai = ( ai1 ,..., aim ) , a ∈ A, i ∈ I , A ⊂ Ω m ,
i
ai , i ∈ I , j ∈{1,..., m} = J ,
j
where is the value of the i-th object j-th parameter.
From the set A , the deciding element is the best-fit object and the choice is justified.
It is clear that already at the stage of substantiation of the approaches to the selection
and the actual selection procedure, there are opportunities to manipulate the results of
solving the problem.
C, A ,
SEA means a couple where C − the principle of optimality, A − is given
C, A
by a set of objects. A solution to the SEA is a subset of objects A0 ⊂ A, obtained
using the principle of optimality: 0 A = C ( A). In the theory of operations research [4], a
C, A
couple is called:
– the decision-making problem, if C and A they are not a priori specified and can
vary;
– the task of choice if the set of objects A is given and the principle of optimality
C − varies;
– is a general optimization problem when C and A is given.
Of course, in situations where variations in the principle of optimization of variants
or multiple objects are allowed, there is an opportunity for manipulation. In this paper,
we will explore the problems of multicriteria optimization problems and the possibility
of manipulating the definition, justification and interpretation of the solutions obtained
in such problems.
3 Formulation and solution specifics of multicriteria
optimization problem
The problem of multicriteria optimization is formalized in the following formulation
[3, 5]:
y i = f i ( x ) → max, i ∈ L1 ,
(1)
y i = f i (x ) → min, i ∈ L2 ,
(2)
x ∈ A, A ⊆ E m ,
(3)
where A − the set of objects, characterized by m parameters, that is, belongs to space
E m ; y ( x ) = ( f1 ( x ),..., f k ( x )) - the vector of object or criterion estimates, which is
given by the mapping
f : A → Ek , L = {1,..., k} −
of indexes of criteria,
L1 = {1,..., k1 }, L2 = {k1 + 1,..., k }
the sets of indexes of criterion functions that are
237
maximized and minimized, respectively, in some problems the set of objects stands out
from the wider set by the constraints most often given by the inequality system.
( )
Vectors y = y1 ,..., yk and y = ( y1 ,..., yk ), y ≠ y , are relative ">" , if i
′ ′ ′ ′ y ≥ yi′,
∀i ∈ L, and at least one inequality is strict.
An evaluation of an object
( ) is said to be effective (Pareto
y 0 = f (a0 ), y 0 = y10 ,..., y k0 ,
optimal, Pareto optimal, non-perfect, majoritarian, non-dominant), unless ">" , there is
another estimate
y, that would strictly y 0 , outweigh the relationship y > y 0 .
Sometimes effective object evaluations are called compromise, but we will use this
term for the final solution of the problem obtained from the effective set after applying
heuristics.
It is known [5] that two effective objects are either equivalent or incomparable by
many criteria.
The problem of determining the Pareto region is strictly objective and is solved
without the use of any heuristics [5]. But the area of compromise is the set of points
from which in most cases one has to choose. Narrowing down the area of effective
objects and, moreover, choosing one of them basically requires the use of additional
information from experts, since effective sets of parameters cannot be compared with
each other formally.
Most practical problems of object estimation have different dimension, because they
characterize different physical properties of objects, so it is advisable not to consider
the absolute values of the parameters of the objects i
a j , i ∈ I , j ∈ J , and their
ωij ( aij ) , i ∈ I , j ∈ J , −
corresponding normalized values monotonous
transformations, which bring the parameters to a dimensionless form and allow to
compare them between yourself. As a rule [3], the following types of parameter values
transformations are used in practice:
ωi(1) = ωi(1) ( ai ) = ( ai0 − ai ) / ( ai0 − aiГ ) ,
z (4)
ωi( 2) = ωi( 2) ( ai ) = ( ai0 − ai ) / ai0 ,
(5)
ωi(3) = ωi(3) ( ai ) = ωis ( ai ) , i ∈ J , s ≥ 2,
(6)
ωis ( ai ) , i ∈ I ,
where, respectively, is the optimal and worst value of the i-th
ω ( a ),i ∈ I ,
i
s i
parameter. In form (6), the values can be determined by relations (4),
(5),(1) andi the figure s is an integer. Obviously, for the transformation (4) the values
ωi ( a ) , ∀i ∈ I ,
always lie in the interval from zero to one, and for (5) the values may
not lie in this interval.
238
In addition to these, there are several ways of normalization of parameters [3], the
purpose of which is to eliminate the problem of incomparability of parameters in the
formal relation. Among the most common methods C
of quantification are:
– rationing average, if the mean values
a i , i ∈ J , of the i-th parameter on the set of
alternatives are significantly different, and in their content they should be the same:
ωi( 4) = ωi( 4) ( ai ) = ai / aiC , i ∈ J ,
– if not only the average but also the variance should match
σ i ,i ∈ J :
ωi(5) = ωi( 5) ( ai ) = ( ai − aiC ) / σ i , i ∈ J ,
(7)
aE ,i ∈ J , −
– if i the reference (normative, true, perfect, measured, known, desired,
etc.) value of the th parameter is known:
ωi( 6) = ωi( 6) ( ai ) = ai / aiE , i ∈ J ,
ωi( 7) = ωi( 7) ( ai ) = ai / ai0 , i ∈ J .
i
Sometimes the following method of normalization is used a , i ∈ J [3]:
( )
ωi(8) = ωi(8) ( a i ) = ( a i − aiC ) / σ i − aiГ / ( ai0 − aiГ ) , i ∈ J ,
which is derived from methods (6) and (7) and results in the distribution of parameter
values between zero and one.
There are many approaches to solving the CSF today [2]. Let us dwell on one of
them [3]. Heuristics are introduced to describe this method of resolving STDs.
Heuristics E1. The type of monotonous transformation to translate the values of the
object parameters to a dimensionless type is made by an expert using formulas that
must satisfy the following requirements:
– take into account the need to minimize deviations from the optimal values for each
objective function (CF);
– to maintain the ratio of preference to the set of objects being compared, across the
set of CFs, and thus not to change the set of effective objects.
A CSF solution may not be optimal for any CF of the form (1), (2), but at the same
time be in some sense the best solution for all the criteria functions at the same time.
Heuristics E2. The best object when solving the CSF should be considered to be the
one for which the deviations from the best values at each estimate are minimal.
If the smallest deviation values for each criterion are not reached at the same time
on any object, then there is a need to compare these deviations with each other due to
the need to attract additional heuristics from experts. To constructively determine the
best object, we present another theorem.
It is known that for every object
a ∈ A, such that in the space of transformed values
0 < ω ( a ) < 1, ∀i ∈ L,
i
of the CF of fair inequalities there exists a vector
ρ = (ρ i , i ∈ L ), that satisfies the normality ratio:
239
ρ i > 0, i ∈ L, (8)
ρ = 1,
i
i∈L (9)
k0 ,
and the number such that the object a ∈ A, satisfies both k equals
ρ iωi (a ) = k 0 , ∀i ∈ L.
(10)
We will introduce three heuristics.
Heuristics E3. The vector of the weight coefficients of the CF ρ = (ρi , i ∈ L), which
satisfies the relations (8), (9), will be interpreted as the ratio of preference between
different CFs, given in quantitative form on the set of CFs.
Heuristics E4. By deciding the CSF for a given vector of weights ρ = (ρi , i ∈ L), we
mean a compromise object that belongs to the set of effective objects and is in the
direction determined by the vector ρ = (ρ i , i ∈ L ), or closest to the beam ρ = (ρi , i ∈ L),
for discrete problems.
Heuristics E5. The compromise solution a0 ∈ A of the STCF should provide the
ρ ω (a ), i ∈ L,
same minimum weighted relative deviations i i 0 for all criteria at the same
time.
It has been proved [5] that if a0 ∈ A it is an effective object for a vector
ρ = (ρi , i ∈ L), then this object corresponds to the smallest value of the parameter k0 ,
at which system (10) is executed simultaneously for all CFs. ∗
It is also known [5] that in order for an object a ∈ A, to be effective
ωi (a ∗ ) > 0, ∀i ∈ L,
at a given vector of weights ρ = (ρ i , i ∈ L ), it is sufficient a ∈ A to be
∗
the only solution of the system of inequalities
ρ iω i (a ) ≤ k 0 , ∀i ∈ L,
(11)
∗
for the minimum value of the parameter k0 , at which this system is compatible.
Thus, the EBS heuristic compromise solution determined by the E5 heuristic can be
found as the only solution of the inequalities of the form (11) for the minimum value
k ,
of the parameter 0 at which this system is still compatible. In the space of the relative
values of the object parameters, the point of intersection of the compromise object
corresponds to the point of intersection of which the directing cosines are determined
by a given vector of relative importance of the CF
ρ = (ρ i , i ∈ L ), with the set of
effective objects. If no such point exists, that is, no vector corresponding to the effective
object lies on the beam determined by the vector
ρ = (ρ i , i ∈ L ), , then the compromised
object is considered to be the one for which the inequality system (11) holds and the
nearest point corresponds to this object. to a given beam. If the compromise object is
not unique, that is, there is some subset of effective objects equivalent to an accuracy
of some sufficiently small number by the value of the parameter k0 , the choice of the
compromise object on this subset is made by another criterion.
It should be noted that alternatives to the CSPS (1) - (3) may include, in particular,
the ranking of objects [3] or objects or phenomena of a different nature. That is, the
tasks of object ranking can be formalized in the class of CSF [3].
240
4 Classification of choice manipulation tasks when solving
multicriteria optimization problems
In accordance with the general description of the problems of peer review and the
approach described above for solving the CSF, we will present the classification of
problems to manipulate solution choices in the problems of multicriteria optimization
developed by the author.
4.1. Tasks of manipulating multiple alternatives
There is ample opportunity for manipulation of choice when there are means of
influencing the initial set of alternatives to the task. Many alternatives can be changed
in different ways:
− ODA voluntary approach;
− imitation of democratic procedures that result in a change of the initial set A;
− introducing additional restrictions and changing the conditions for participation of
alternatives in further solving the multicriteria task;
− deliberately ignoring some restrictions in order to expand the initial set of
alternatives;
− other approaches that change the initial set of alternatives in order to manipulate
and “reasonably” select the ODA required for the ODA.
4.1.1. Supplementing the initial set of alternatives A by some subset
A* , A* ∩ A = ∅
containing the alternative a ∈ A that, from the point of view of the
* *
manipulator, should be chosen as a solution as a result of the decoupling of the CSP of
form (1) - (3). Then a ∈ A = A ∪ A .
* 1 *
−4.1.2. Removing from the initial set of alternatives A some subset
A , A\ A = A − 1
containing an alternative a ∈ A that, from the point of view of
* −
the manipulator, should not be chosen as a solution as a result of solving the BCO
problem (1) - (3).
4.1.3. A combination of approaches 4.1.1 and 4.1.2: simultaneously adding an initial
set of acceptable alternatives A to a certain subset A and removing a subset
* A− ,
that contains undesirable alternatives. That is A = A ∪ A \ A .
1 * −
4.1.4. Change the weight of the parameters that characterize the objects of the initial
set A . To explore this approach to manipulating choice, the author developed a group
of methods for adaptively determining the weights of object parameters [3].
In addition to purposefully changing the weight of object parameters, more drastic
variations of this approach can be applied.
4.1.5. Supplementing the initial set of parameters of alternatives A by some subset
of additional parameters J * = (m + 1, m + 2,...). That is, increasing the dimension of the
m m1 m1 = J ∪ J *
parametric space from Ω to Ω , where . Such manipulation may
241
make it possible to select an alternative in the newly expanded parameter space that is
desirable for the manipulator.
4.1.6. Removal from the initial set of parameters J of some subset J ⊂ J . That is,
−
m m2
the reduction of the dimension of the parametric space from Ω to Ω , where
m2 = J \ J − .
In this case, there may be conditions for choosing the alternative for the
manipulator to be chosen as a solution as a result of solving the problem of BCO (1) -
(3). Or, at least, to remove from the subset of winners such alternatives that are
undesirable for the manipulator.
4.1.7. A combination of approaches 4.1.5 and 4.1.6: simultaneously adding
additional parameters to the initial parameter space and removing from the set of
parameters undesirable for the manipulator. That is, changing the parametric space
m m3 m = J ∪ J * \J − .
from Ω to Ω , where 3 Such manipulation may allow the desired
manipulator to be the winner or at least to remove from consideration the subset of the
manipulator undesirable.
4.2. Manipulation of CSF limitations and measurement scales
4.2.1. Supplementing* the initial set of constraints S by some subset of the new
S , S ∩S =∅
*
additional constraints .
4.2.2. Removing a constraint S subset from the original set of constraints
S − , S \ S − = S1
.
4.2.3. Combination of approaches 4.2.1. and 4.2.2: simultaneously adding
restrictions to the original set of constraints and removing some unwanted subset of
restrictions from the initial set S .
4.2.4. Manipulation of scales when measuring the parameters of alternatives:
replacement of scales of measurement, use of unacceptable operations in operations on
results of measurement, etc.
4.2.5. Manipulation of measurement scales in determining weights of parameters,
relative weight of criteria, coefficients of competence of experts (weight of information
sources [6]), etc.
4.2.5.1. When measured in ordinal scales, the organizers of the examination may
invite the experts to determine their preferences in the space of strict preference
relations, thus denying the participants of the expert group the opportunity to express a
situation of equivalence or indifference, which can often arise due to peculiarities of
the subject area and psychological properties of a person. That is, it may be deliberately
narrowed the possibility of experts to determine their true advantages, inclining them
to choose in an artificially narrowed space.
4.2.5.2. Limitations can be set in cardinal scales: for example, in the range from 1/9
to 9, which significantly disrupts transitivity already at the stage of initial examination
and does not a priori allow to reach a situation of supra-transitivity of relations between
parameters, criteria or competence of experts, depending on the formulation and
interpretation of the task .
242
4.2.5.3. Establishing a requirement for a fixed setting of parameter values or
preference ratios between objects, their parameters, criteria or expert competence,
depending on the purpose of the examination. This often leads to a significant violation
of the adequacy of modeling the subject area and the deliberate loss of information
about the phenomenon being modeled.
4.3. Manipulation of many applicable metrics, criteria and methods of convolving
criteria
To achieve the goal of manipulation, it is almost imperceptible for the untrained
observer to select from the initial set of acceptable alternatives A practically any
alternative that is desirable for ODA.
4.3.1. When solving a problem in ordinary scales, there are some of the most popular
metrics - Heming, Cook, Euclid, etc. [3]. In some cases, applying each of the metrics
produces results that do not overlap with the solutions obtained in the other metrics.
Therefore, the manipulator just needs to choose the solution by which metric suits it the
most and to set the requirement to unblock the CSF using the metric that is
advantageous for it.
4.3.2. When applying formulas to switch to a dimensionless type of criteria, the
limits of changing the criterion functions are essential. By choosing these boundaries
hypothetically, you can choose the values you need to manipulate the value Hjunction
ω ,i∈ J,
and significantly influence the final decision. The manipulation of the lower i
and upper
ω iВ , i ∈ J , bounds of the dimensionless values of the criteria defined by
formulas (4) - (7) is that in this way the order of change of deviationsH from the optimal
values by those criteria for which variations of the boundaries i
ω , i ∈ J , and / or
ω iВ , i ∈ J , deviations are carried out ω i , i ∈ J .
4.3.3. During the preparation phase for the multicriteria optimization task in the
previous steps, one or more rounds of sequential analysis can be applied. That is, to
pre-establish such criteria that will screen out undesirable alternatives from the initial
set A for ODA, as being in advance unpromising given the “legitimately” established
criteria.
4.3.4. Changing the class of tasks in which the initial model of multicriteria
optimization should be formalized. This automatically entails the sound application of
the methods required to achieve the goal of manipulating the methods of solving the
problem and applying the appropriate types of convolution to the criteria of the task.
Such manipulations are unlikely to remain unnoticed by a person skilled in the art,
hidden by complex formulas, supplemented by formal and comprehensive
justifications.
4.3.5. Supplementing the initial set of criteria by some new subset or one, even
insignificant criterion k1 > k. It is thus possible to make an effective alternative, which
in the previous model, when the number of criteria was equal, was dominant. Such
manipulation can significantly affect many effective solutions to the initial CSF. And
then to determine the compromise solution to problem (1) - (3).
243
f ( x) = ( f1 ( x),..., f k ( x))
4.3.6. Removing from the original set of criteria some
subset of criteria or even one of them. That is, the narrowing of the criterion space of
the problem: 2
k < k . Such an interference with the modeling process can significantly
affect the multicriteria model of the phenomenon under study.
4.3.7. The combination of 4.3.5 and 4.3.6 may have a synergistic effect over the
autonomous application of each of these approaches. In particular, it can be deduced
from the choice of a non-advantageous ODA dominant alternative under some
reasonable excuse or initially transformed into an effective alternative.
4.4. Manipulation by influencing the expert group
4.4.1. Replacement of a dissatisfied expert when a single examination is appointed
and the expert selected does not satisfy the manipulator's wishes.
4.4.2. Supplementing the initial pool of experts with some subset of experts affected
by ODA or other decision makers interested in manipulation - to prepare and justify the
required decision.
4.3. Removing from the initial pool of experts those who are uncomfortable with
ODA and have their own personal opinions.
4.4. A combination of approaches 4.4.2 and 4.4.3 is the simultaneous addition of an
expert team to new team members and the removal from the initial expert group of
some of its members.
4.5. Manipulating many of the goals declared when solving a problem
4.5.1. Increasing the subset of choices in the course of solving a problem, which will
introduce to the subset of winners the alternatives that are desirable for ODA.
4.5.2. Achievement of the desired ODA solution: with the help of a powerful tool,
original mathematical support, you can achieve both the harmonization of solutions at
different levels, and provide wide and unnoticeable opportunities for malicious
manipulation.
4.5.3. In the case when the prepared and justified decision does not satisfy the ODA,
the choice can be made invalid for various contrived motives and formal reasons.
4.6. Manipulation of the system of benefits of experts
4.6.1. Minimal changes in the preference given by the experts can be achieved by
changing the initial compromise solution by changing the elements in the MPP at
n >> 3.
4.6.2. Achieving the goal of manipulation may be to replace only some of the
individual preferences of the expert in order to smooth his preferences or to agree on a
group expert decision.
4.6.3. The manipulation of expert competence weights (the weight of information
sources [6]) can be achieved by the desired ODA solution or at least blocking the
decision that is most unacceptable to the manipulator.
244
4.6.4. The exclusion of some elements of the WFP from consideration - that is,
transferring it to the category of incomplete matrices. This changes not only the
information but also the methods of solving the peer review task. This may entail the
choice of a compromise solution that is necessary or acceptable for the manipulator.
4.6.5. Adjusting the values of the element of the pairwise comparison matrix that
most significantly affects the result is a way of indirectly influencing the final solution
of the CSF.
5 Prospects for further research
The problems of choice manipulation have long and comprehensively been studied by
domestic and foreign scientists. At the same time, the possibilities of manipulation in
multicriteria formulation of the problem have wide prospects for development. In order
to improve the approach described, the possibilities of manipulating the GCS solutions
in the following directions should be considered and explored:
- formalizing the problem of manipulation for a class of object ranking tasks - strict
and non-strict;
- study of resistance to manipulation of tasks that are formalized using different met-
rics and criteria of different kinds;
- generation and comprehensive consideration of examples that clearly demonstrate
the ability to manipulate the described task classes;
- implementation of a series of computational experiments to determine the relation-
ships between the methods of manipulation, identify the most effective of them and
find options for preventing manipulation;
- development of approaches to visualization in the problems of choice manipula-
tion, which is a significant evidential factor and a way of demonstrating the capabilities
of the method, in particular the detection of manipulation;
- developing and investigating manipulation procedures or preventive detection of
manipulation capabilities at the model development level or in the early stages of solv-
ing the problem;
- exploring the possibilities of using combinations of different approaches to manip-
ulation in order to minimize the impact on each element of the model;
- study the sensitivity of the initial elements of the model and change the solution of
the problem with a slight variation of the data and minor changes in the elements of the
model;
- Investigate the possibilities of using system optimization problems to model and
develop methods of decoupling in CSS.
6 Conclusions
Various aspects and possibilities of manipulation of choice in expert evaluation tasks
in the case of formalization of them in the form of multicriteria optimization are inves-
tigated.
245
In general, various aspects and sources of possible manipulation in peer review tasks
are considered.
The choice can be justified or voluntary. With well-organized manipulation, the
ODA can be adjusted to the desired selection for the manipulator. That is why it is
especially important for expert judgment to have a clear representation of the heuristics
that researchers rely on when deciding on a compromise solution, documenting it, and
disclosing it explicitly.
References
1. Lezina, Z. M.: “Manipulation of option choice (agenda theory)”, Avtomat. i Telemekh.,
1985, no. 4, 5–22; Autom. Remote Control, 46:4, pp. 411–426 (1985).
2. Voloshyn, O. F., Mashchenko, S. O.: Decision theory: tutorial. Kiev: Publishing and
Printing Center "Kyivskyi universytet", 366 p. (2010).
3. Hnatiienko, H. M., Snytyuk, V. E.: Expert Technologies for Decision Making. – K.:
Company ”Maklaut”. – 444 p. (2008).
4. Makarov, I. M., Vinogradskaya, T. M., Rubchinsky, A. A., Sokolov, V. B. The theory of
choice and decision-making. Moscow: Nauka. – 328 p. (1982).
5. Mikchalevich, V. S., Volkovich, V. L. Computational methods of research and design of
complex systems. Moscow: Nauka. – 286 p. (1982).
6. Dodonov, A., Lande, D., Tsyganok, V., Andriichuk, O., Kadenko, S., Graivoronskaya, A.:
Information Operations Recognition. From Nonlinear Analysis to Decision-Making.
Lambert Academic Publishing, (2019).