=Paper=
{{Paper
|id=Vol-2392/paper21
|storemode=property
|title=Network-centric Approach to Destructive Manipulative Influence Evaluation in Social Media
|pdfUrl=https://ceur-ws.org/Vol-2392/paper21.pdf
|volume=Vol-2392
|authors=Sergiy Gnatyuk,Marek Aleksander,Petro Vorona,Yuliia Polishchuk,Jamilya Akhmetova
|dblpUrl=https://dblp.org/rec/conf/coapsn/GnatyukAVPA19
}}
==Network-centric Approach to Destructive Manipulative Influence Evaluation in Social Media==
https://ceur-ws.org/Vol-2392/paper21.pdf
Network-centric Approach to Destructive Manipulative
Influence Evaluation in Social Media
Sergiy Gnatyuk1,2 [0000-0003-4992-0564], Marek Aleksander3 [0000-0003-2619-1063],
Petro Vorona4 [0000-0002-6755-9023], Yuliia Polishchuk1 [0000-0002-0686-2328]
and Jamilya Akhmetova2 [0000-0003-3054-5987]
1
National Aviation University, Kyiv, Ukraine
2
Yessenov University, Aktau, Kazakhstan
3
State Higher Vocational School in Nowy Sacz, Nowy Sacz, Poland
4
Institute of Personnel Training of the State Employment Service of Ukraine, Kyiv, Ukraine
s.gnatyuk@nau.edu.ua, aleksandermarek4@gmail.com,
voron67@ukr.net, polishchuk.yu.ya@gmail.com, jamy88@mail.ru
Abstract. The mass media phenomena investigation are equipping a mankind
with a broader and deeper understanding of society and culture due to the fact
that the texts produced by the mass media are the most socially significant mes-
sages and have a priority in the society over all other types of texts. Moreover,
in recent years, the concepts of mass consciousness influence and mass behav-
ior influence had become increasingly popular. Frequently, mass media is the
subject or instrument of such manipulation. Today Internet is digital media
technology including web-sites, e-mail, Internet-radio and TV, as well as social
networks and other services (so called social media). Destructive manipulative
influence evaluation in social media is actual and important scientific task
whichever quantitative its assessment. In this paper advanced approach to de-
structive manipulative influence evaluation in social media was proposed. This
approach allows calculating the quantitative parameters that characterize the
level of destructive manipulative influences on public opinion using channels of
social media. Using this approach specialized software tools can be developed
for manipulative influence automatized detection and quantitative evaluation.
Keywords: Social Media, Destructive Manipulative Influence, Quantitative
Evaluation, Public Opinion, Information Psychological Security.
1 Introduction
Currently, the mass media has become the main tool for information dissemination
that affect individual consciousness and subconscious and, as a consequence, the pub-
lic consciousness. Public consciousness should be considered as a social phenomenon
whose main function is to influence the audience through semantic and evaluative
information transmitted by mass media channels. Moreover, public consciousness is a
mechanism for actualizing information through various means of communication.
Therefore, mass media influence does not always have a negative effect, since in the
globalized world it is difficult to imagine human life without newspapers, social net-
works and other mass media. However, on the other hand, some individuals, for gain-
ing their own advantages, use mass media as an instrument of influence on society.
2 Analysis of Modern Researches Devoted to Manipulative
Influence Evaluation in Social Media
The research in mass media manipulative influence field was engaged by large num-
ber of Ukrainian as well as foreign scientists.
In research papers of A. Peleschyshyn [1,2] the description of informational environ-
ment of the virtual community in social networks was described; a model of internal
informational environment and the discussion model of the virtual community were
developed and detailed; on the basis of a formal model of the informational environ-
ment of the virtual community, taking into account the quality of information content,
the indicator of information threats is determined. These models allow organizing the
detection and counteraction of the information threats and providing the information-
psychological security of a person in a social network. However, a significant disad-
vantage of these models is the impossibility of quantity evaluation.
A human activity model and human joint activities model in the information space
was proposed by A. Shiyan [3]. On the basis of proposed models, a human protection
method from negative informational and psychological influence on the basis of the
peculiarities of its activity was presented. However, directly, the formal protection
model against manipulative influence was not highlighted, as a consequence, is not
possible to evaluate mathematically the effect.
In research study of D. Gubanov [4] an overview of the influence models in social
networks was presented, the main classes of models in social networks were consid-
ered and the correspondence between classes of models and parameters of the model-
ing object was established. An analysis of these models allows to study the relation-
ships between objects in a social group and the dependence of information influence
on persons and public opinion.
The formalized model for detecting PR-influence through Internet publications was
developed by O. Khatyan in paper [5]. According to this work, PR-influence is the
mass media manipulative influence. However, this model allows to identify the im-
pact, but do not assess it. Therefore, the developed model can be used as an initial
stage for the development of a further evaluation model. Proceeding from the above, a
detailed analysis of manipulative influences models was conducted.
Given results of the analysis show that existed models and methods have certain limi-
tations as to practical application for solving various kind of tasks of information and
psychological security (also it can be combined with information-technical influ-
ence [6]). It is found that the evaluation method of media manipulative influence is
only then effective and adequate in case it meets all the specified criteria like taking
into account methods of manipulative influence or formal its description.
The main task of this paper is to develop an evaluation method of destructive manipu-
lative influence, which will allow calculating the quantitative parameters of their in-
fluence on public opinion in social media.
3 Theoretical Background of Proposed Approach to
Manipulative Influence Evaluation
After the definition of existed approach new approach based on network-centric con-
cept [7,8] was developed. The proposed evaluation approach of mass media manipu-
lative influence on public opinion is realized in following 8 stages:
1. Evaluation the financial costs;
2. Ranking of reasons by the degree of their danger;
3. Determination the objectives of the manipulation campaign (hereinafter – KMp);
4. Determination of the KMp tasks;
5. Determination of strategies for the KMp implementation;
6. Selection the mass media for manipulation;
7. Selection the manipulation method;
8. Estimation of manipulative influence.
These stages will be detail described after defining input and output data. The practi-
cal realization of this approach could allow evaluating the level of media manipula-
tive influence on public opinion.
Input data: the set of reasons for conducting KMp; the set of dislocation places; the
set of goals for KMp; the set of focus groups; set of criteria for estimating parameters;
the set of strategies for KMp; the set of mass media; the set of manipulation methods
and the set of «weight by criteria».
Output data: list of selected causes, goals, criteria, tasks, strategies, methods, mass
media; values of magnitude and effectiveness of manipulative influence.
Evaluation the effectiveness of manipulative actions is the preparatory stage for con-
ducting the campaign. This stage is only necessary for the customer – to evaluate the
financial efficiency of KMp. At this stage, from the whole mass media database it is
necessary to select those mass media which, in the opinion of the expert, are suitable
for the implementation of manipulative influence. First of all, it is necessary to deter-
mine the economic effect of advertising in selected mass media in accordance to (1):
Т D HT
E U P U D , (1)
100
where E is the economic effect of advertising; Т D is additional trade turnover under
the advertising influence; НТ is trade margin on goods (measured in % to the realiza-
tion price); U Р are advertising costs; U D are additional expenses on the growth of
goods turnover.
Z P Z P ZO
The next formula . gives an opportunity to evaluate the effectiveness of
ZO ZO
advertising in relation to the profits of the company, obtained from advertising ( Z P )
to profit before advertising ( ZO ).
3.1 Stage 1 – Evaluation the financial costs on KMp
Step 1 – Choosing the reasons for conducting KMp.
At the first step of Stage 1 it is necessary to input a set of reason ( I ):
P
I { Ii } {I1, I 2 ,..., I p }, Ii I , i 1, P
i 1 ,
and experts choose those reasons (using voting method), which, in their opinion, re-
quire a KMp. After that experts go to step 2.
Step 2 – Ranking of reasons by the degree of their danger.
That is, for each possible i reason (for the period of preparation and implementation
of the KMp) in r regions, it is necessary to evaluate the financial cost ( ui ) for a ma-
nipulative attack using formula (2):
R J T
ui (t ) c , r 1, R; j 1, J ; t 1,T
r 1 j 1 t 1
r
j
r
i (2)
,
where rj (t ) is the amount of mass media of j type (social networks, newspaper, TV
etc.), for the period t in region r , cir is the cost of one mass media of j type in re-
gion r .
Step 3 – Definition of financial constraints.
At this step, it is necessary to determine financial constraints of conducting KMp
using following expression:
R T
ui d (t ), i 1, I ,
r t 1
r
j (3)
where d rj (t ) is the amount of possible financing of the KMp during t period in r
region. As a result of Stage 1, will be received a financial statement about expenses
on the KMp. Based on the report, it is possible to analyzed if KMp should be carry
out from a financial point of view.
3.2 Stage 2 – Ranking of reasons by the degree of their danger
During the ranking of reasons by the degree of their danger, each expert are conduct-
ing pairwise comparison of reasons by the degree of their danger for KMp (step 1)
and create the matrix H r (step 2) for each r region. The expert’s mark define as hijr ,
where i is the expert number, j is the reason number, r is the region number:
1, if I i equal to I i 1;
hijr 2, if I i more dangerous than I i 1;
0, if I less dangerous than I ; (4)
i i 1
h11, h12, h1b
H r hij (i 1, a ) h21, h22, h2b ,
( j b ) ... ... ...
ha1, ha 2, hab
(5)
Furthermore, the system agrees the expert score with one of the well-known harmoni-
zation computer algorithms. As a result of the Stage 2 implementation, will be ob-
tained a matrix of reasons, according which it is expedient (or inexpedient) to conduct
KMp.
3.3 Stage 3 – Determination the objectives of the KMp
At this stage, a set of objectives ( С ) must be created by experts:
c
C { Ci } {C1, C 2 ,..., Cc }, Ci C , i 1, c
i 1 .
The set of objectives is a possible consequence of the reasons for conducting the KMp
(step 1). Moreover, in step 2 by experts are creating the set of focus group ( G ) who
are the target of influence:
g
G { Gi } {G1, G2 ,..., Gg }, Gi G , i 1, g
i 1 ,
r
and criteria ( Kr ): Kr { Kri } {Kr1, Kr2 ,..., Krr }, Kri Kr , i 1, r , that are the param-
i 1
eters by which the objectives and tasks of the KMp will be selected.
As a result of the Stage 3 implementation, will be formed objectives, the focus groups
who are the objects of influence are identified and the criteria (for choosing objectives
and tasks) are selected.
3.4 Stage 4 – Determination of the KMp tasks
At the step 1, experts form a set of tasks ( A ), which are connected to the objectives
of the KMP (the objectives were identified in step 3), for assessing the KMP's tasks
by using (6):
y
A { Ai } { A1, A2 ,..., Ay }, Ai A, i 1, y. (6)
i 1
At the step 2, the proportion of agreed positive scores x vAi to the criteria of each v
task of each agent Ai according to all the criteria is calculated by (7):
xvAi
v
Ai Kri
, i 1, v.
a v
Ai Kri
(7)
As a result of Stage 4, the tasks to be achieved during KMp are formed.
3.5 Stage 5 – Determination of strategies for the KMp implementation
The objectives of manipulating different groups are made at different times ti and
with different benefits. In accordance with the formulated objectives and tasks their
strategies ( L ) Ci Li , Ai Li are selected and implemented. At the step 1, from the
set of strategies experts select the strategy that require of the KMp conducting:
m
L { Li } {L1, L2 ,..., Lm}, Li L, i 1, m.
i 1
At the step 2, the list of strategies is ranking by following expression (8):
3, if al a p ,
flp 2, if al a p ,
(8)
1, if al a p.
As a result of Stage 5, it is formed a number of strategies (possibly more than 1) on
which KMp will be realized.
3.6 Stage 6 – Selection the mass media for manipulation
At the stage 6, it is necessary to choose mass media ( Z ) which will be used during
manipulative influences [9-11].
Then, it is necessary to conduct an analytical study of mass media with further evalua-
tion of the information influence effectiveness. At the step 1, a set of mass media is
formed as follow:
z
Z { Zi } {Z1, Z 2 ,..., Z z }, Zi Z , i 1, z
i 1 .
Step 2 – evaluation the effectiveness of manipulative actions can be presented using
following formula (9):
aZRiG
cZRiG i i
i i
bZRi (9)
i
,
where aZRiG is a score of z mass media in 𝑔-th focus group in r region, z 1, Z ,
i i
R R
g 1, G , r 1, R, bZ i is a score of z mass media of population in r -region, cZ iG is a
i i i
compliance index of z mass media in 𝑔-th focus group in r region.
Thereafter, the mass media price matrix (10) is created and stored in database of ma-
nipulation system:
d1Z d1P
Dzp ,
(10)
d Z1 d P1
where d zp is publication price (unit of standard area, standard broadcast time etc.) in
z mass media on р page (in р broadcast time etc.).
After that, an important action is the calculation of «cost per thousand» readers:
d Zi p
eZi Gi pRi .
bZi Ri cZi Gi Ri
According to the different price in different cities (regions), the «cost per thousand»
quantity will difference. So, ind index can be presented as indZGR and matrix IND
converted into cube, but in our example (11) it is necessary to leave just two indices
Z and G :
ind1Z ind1P
INDZP ,
(11)
ind Z 1 ind P1
where ind zg is the influence of z mass media on 𝑔-th focus group or its segment.
The ratio of advantages in 𝑔-th focus group for each studied above j criteria of
p j (k , h) for pair of mass media alternatives Ak , Ah can be presented as
rkjg rhjg
, if rkjg rhjg
p gj k, h m j ,
(12)
0 if not.
where m j is the rating scale according to the j criteria, rkjg , rhgj are the value of the
Ak , Ah options according to the j criteria in 𝑔-th focus group.
The ratio of advantages over a pair of alternatives ( Ak , Ah ) taking into account all
considered criteria (12) can be defined by following expression (13):
rkjg rhjg
J
J
, if rkjg rhjg ;
p g
k, h p gj
k, h m j
j 1 j 1 (13)
0 if not.
The ratio of the dominance Ak alternative over Ah in 𝑔-th focus group can be pre-
sented by membership function (14) which characterizes the dominance intensity of
k over h mass media:
p g k , h p g h, k , if p g k , h p g h, k ;
Dg k , h (14)
0 if not.
The best alternative using (12) – (14) corresponds to the following expression:
* Ak* max Dk 1,,m * Ak 1 min max p k , h l 1,,l p h, k .
k 1,.., m
l k
As a result of Stage 6, the mass media which according to the criteria has the highest
rating will be chosen. Moreover, isn’t necessarily it will be one mass media for all
disposition places. It is also possible to choose different mass media for different
disposition places.
3.7 Stage 7 – Selection the manipulation method
In database the manipulative influence methods ( M ) are stored (for example [12]).
These methods (15) can be presented as following complex set:
x x e
M { Mi } { { M ij }} {{M11 , M12 ,..., M1e },{M 21 , M 22 ,.., M 2 e },
i 1 i 1 j 1
,...,{M x1, M x 2 ,..., M xex }}, (i 1, x, j 1, e)
(15)
.
In database all methods are separated by certain characteristics and used by different
mass media. Furthermore, each agent chooses the most effective method from the
database that, in agent’s opinion can use by mass media. After that, the list of criteria
(16) for which the methods selecting are conducting will be formed.
M ij k11,i j k2 2,i j k3 3,i j , i 1, I , j 1, J , (16)
where k1 «the value» of l criteria ( l 1, 2, 3 ), iji is criterion value of j operative
manipulative influences method according to i strategy for l criteria.
As a result of Stage 7, manipulative influence methods which will be implemented by
the mass media chosen in step 6 will be chosen.
3.8 Stage 8 – Estimation of manipulative influence
If the initial evaluation of 𝑔-th focus group is presented as bg , then the task of manip-
ulation is to change the group's opinion to
bg qg , qg Qg , g N ,
where qg is the change of initial opinion, thus qg is the controlling influence.
As a result of the sequence of operational manipulative influences, the final opinion in
some focus groups is formed:
Bq T (b q),
where qg is the initial opinion of focus group, q q .
gN
g
The target function of public opinion modification can be represented as follow:
Ф Bq , q max H ( Bq ) C q , (17)
where H () are received benefits (profit, votes in elections, change in certain officials
or government etc.) by an organization in whose interests the manipulative influence
is conducted, which depends on the change of public opinion, C () are costs for ma-
nipulative actions.
Using (17) the task of management is to choose an acceptable management method
that maximizes efficiency Ф Bq , q max . The value of operational influence can be
determined by following expression:
x x ,
Di f ik k
i
k 1
i
where f ik is a function (table or algorithm), that reflects the difference in scores
xik xik 1 in the amount of operational influence, for example, the volume and num-
ber of compromising newspaper articles or news releases on the TV channel.
If xik xik 1 , then Di f ik xik 1 xik . The effectiveness l of influence El in peri-
od k
is determined in different ways (18), in particular, by the following formula:
wlk wlk 1
k k 1
, if ( wlk wlk 1 ) ( xlk xlk 1 )
l
x x
El k
l
, (18)
k 1
wl wl
k
k 1 k 1
x k 1 x k , if ( wl wl ) ( xl xl )
k k
l l
where wlk is influence assessment in the period k , xlk are the magnitudes (levels) of
the produced influence of l type in the period k , l 1, L is influence type.
4 Experimental Study and Discussion
The specialized experimental technique was developed for mentioned network-centric
approach and also implemented for quantitative assessment based on real examples in
social media sphere. In practice the task of destructive manipulative influence causes
changing opinions, views and behavior of focus groups. If primary assessment of 𝑔-th
focus group is 𝑏𝑔 , then task of destructive manipulative influence is changing to
𝑏𝑔 + 𝑞𝑔 , 𝑞𝑔 ∈ 𝑄𝑔 , 𝑔 ∈ 𝑁, where 𝑞𝑔 is changing of opinion and 𝑞𝑔 is controlling in-
fluence. As result of the sequence of manipulative influences the finishing opinion of
focus group is forming. Next by using expressions (17) – (18) the effectiveness of
influences can be calculated for all Internet mass media (see last stage of proposed
approach). Considering the network-centric concept involves data analyzing and
gathering from different regions for obtaining results on manipulative influences, it is
proposed to calculate the influence of the mass media and its efficiency as an average
of 3 Ukrainian regions for 4 Internet newspapers (named „Facty“, „Segodnya“,
„Komsomolska Pravda“ and „Dzerkalo tyzhnia“) and 3 manipulation methods select-
ed by experts in accordance to 7 Stage of proposed approach.
The influence of selected mass media presented in Tabs. 1, 3, 5, 7; the effectiveness
of influence showed in Tabs. 2, 4, 6, 8 in accordance to (18). The value of operational
influence will be determined by the number of articles (using competent expert esti-
mation in this sphere).
Table 1. Table captions should be placed above the tables.
x10 x11 x12 x13
1 3 4 5
Di n/a 2 3 4
Table 2. Effectiveness of influence of newspaper „Facty“
w10 w11 w12 w13
The value of influence effectiveness presented in % 18 28 34 33
E1Pi n/a 14 11,3 9
Table 3. Manipulative influence of newspaper „Segodnya“
x 20 x12 x 22 x23
1 3 4 5
Di n/a 2 3 4
Table 4. Effectiveness of influence of newspaper „Segodnya“
w20 w12 w22 w23
The value of influence effectiveness presented in % 11 14 18 20
E2Pi n/a 7 6 5
Table 5. Manipulative influence of newspaper „Komsomolska Pravda“
x30 x31 x32 x33
1 3 4 5
Di n/a 2 3 4
Table 6. Effectiveness of influence of newspaper „Komsomolska Pravda”
w30 w31 w32 w33
The value of influence effectiveness presented in % 9 24 28 34
E3Pi n/a 12 9,3 8,5
Table 7. Influence of newspaper „Dzerkalo Tyzhnia“
x 40 x14 x 42 x43
1 3 4 5
Di n/a 2 3 4
Table 8. Effectiveness of influence of newspaper „Dzerkalo Tyzhnia”
w40 w14 w42 w43
The value of influence effectiveness presented in % 10 20 30 40
E4Pi n/a 10 10 10
These tables present the influence of different Internet mass media and the value of
manipulative influence on public opinion.
In Fig. 1 graphically depicts the ratio of the value of operational influence, that is the
number of articles, to the value of the effectiveness of manipulative influence.
Fig. 1. Effectiveness of mass media influence on public opinion
In accordance to Fig.1 can be declared that the biggest value of destructive manipula-
tive influence (with smallest operational influence) shows Internet mass media
“Facty” – when the number of articles is 2, the effectiveness of manipulative influ-
ence is equal to 28%. For comparison another Internet mass media “Segodnya” shows
the effectiveness of manipulative influence equal to 14% with the number of articles
equal to 2. Experimental study has confirmed and demonstrated the developed net-
work-centric approach ability in practical cases for effective destructive manipulative
influence evaluation in social media.
5 Conclusions
In this paper the network-centric approach to destructive manipulative influence eval-
uation in social media was developed. This approach by means of evaluating financial
expenditures, defining goals, objectives and strategies for manipulating, selecting
Internet mass media and classified methods of manipulation, based on the generated
databases of causes, goals of the criteria, focus groups and mass media, allows calcu-
lating the quantitative parameters that characterize the level of destructive manipula-
tive media influences on public opinion. Experimental study confirmed this approach
ability for quantity manipulative influence evaluation which realized by modern mass
media and using manipulation influence methods on personal or public opinion. Giv-
en results will be useful in information security for evaluation harmful mass media
influence on public opinion or on consciousness individuals. In the next research, it is
going to conduct an experimental study of the developed method for its verification,
correctness confirmation, as well as the establishment of the practical application
possibility for raising the information and psychological security level of citizens,
society and the state in cyberspace (for example it will be very actual and important in
critical information infrastructure of the state [13]). Also using this approach special-
ized software tools can be developed for manipulative influence automatized detec-
tion and quantitative evaluation.
References
1. Peleschyshyn, A., Holub, Z., Holub, I.:The preliminary stage of the algorithm for detecting
information and psychological manipulation in online communities. In 2018 IEEE Pro-
ceedings of 13th International Scientific and Technical Conference on Computer Sciences
and Information Technologies, pp. 24-32 (2018).
2. Fedushko, S., Syerov, Yu., Korzh, R.: Validation of the user accounts personal data of
online academic community. In: 13th International Conference “Modern Problems of Ra-
dio Engineering, Telecommunications and Computer Science”, TCSET 2016, pp. 863–
866. Lviv-Slavske (2016). DOI: 10.1109/TCSET.2016.7452207.
3. Shiyan, A.: Method of protecting a person from negative informational and psychological
influence on the basis of activity typology. Information security, №3 (15), pp. 92-99 (2014).
4. Gubanov, D., Novikov, D., Chkhartishvili, A.: Models of influence in social networks.
Large-Scale Systems Control, vol. 27, pp. 205-281 (2009).
5. Ryabyy, M., Hatyan, O., Bagatskуy, S.: The model of PR-impact detection by means of Internet
mass-media. Ukrainian Scientific Journal of Information Security, vol. 21, issue 2,
pp. 131-139 (2015).
6. Danik Yu., Hryschuk R., Gnatyuk S.: Synergistic effects of information and cybernetic in-
teraction in civil aviation, Aviation, vol. 20, №3, рр. 137-144 (2016).
7. Tikhomirov, A., Kinash, N., Gnatyuk, S., Trufanov, A. et al: Network Society: Aggregate
Topological Models, Communications in Computer and Information Science. Verlag:
Springer International Publ, vol. 487, рр. 415-421 (2014).
8. Trahtengerts, E., Pashchenko, A.: Principles of Implementation and Estimation of Influence
Factors in Network Centric Systems. In Proceedings of 2017 IEEE 11th International Con-
ference on Application of Information and Communication Technologies (2017).
9. Trach, O., Peleshchyshyn, A.: Development of directions tasks indicators of virtual commu-
nity life cycle organization. In Proceedings of 2017 12th International Scientific and Tech-
nical Conference on Computer Sciences and Information Technologies (2017).
10. Fedushko, S.: Development of a software for computer-linguistic verification of socio-
demographic profile of web-community member. Webology 11 (2), article 126 (2014).
11. Syerov, Yu., Fedushko, S., Loboda, Z.: Determination of Development Scenarios of the
Educational Web Forum. Proceedings of the XIth International Scientific and Technical
Conference (CSIT 2016). pp. 73-76. Lviv (2016). DOI: 10.1109/STC-CSIT.2016.7589872
12. Polishchuk, Yu., Zhmurko, T.: Information-psychological security of society in the context
of information warfare: monograph. Akademia Techniczno-Humanistyczna, Bielsko-Biała,
Poland. pp. 132-161 (2016).
13. Gnatyuk, S., Sydorenko, V., Aleksander, M.: Unified data model for defining state critical
information infrastructure in civil aviation, In Proceedings of the 2018 IEEE 9th Interna-
tional Conference on Dependable Systems, Services and Technologies, Kyiv, Ukraine,
May 24-27, pp. 37-42 (2018).