Regula A icle
How does ha ed sp ead in I aly? Ma io D aghi’s dimision case
Da ia Mo a eale-Cal anese
*
, Se gio A ce-Ga cía , Elias Said-Hung
Uni e sidad In e nacional de La Rioja, Spain
1. In oduc ion
Social media pla o ms such as X ( o me ly Twi e ), TikTok, and
Facebook ha e become signi ican enues o he dissemina ion o
disin o ma ion and ha e speech (Sanchez e al., 2022; Ba ud & Olalekan,
2022). These pla o ms o en dissemina e umo s, s e eo ypes, and
p ejudices ha a ge speci ic social g oups. Al hough he cu en
discou se on ha e speech lacks consensus on i s de ini ion (Paz e al.,
2020; Rossini, 2022), he e seems o be a common poin o ag eemen
when es ablishing he ea u es ha cha ac e ize his ype o exp ession,
such as he public na u e and he in en ion o ha m an indi idual o a
speci ic g oup o e hnic, gende , eligious, o igin, o poli ical easons
(Sella s, 2016; S uck, 2019; Ziegele e al., 2018).
As Magall´
on (2019) igh ly poin s ou , ha e messages some imes
in ol e he p oduc ion and dissemina ion o in en ionally dis o ed in-
o ma ion. This misleads he audience in o posi ioning na a i es based
on ac ual inaccu acies abou issues o e en s, as well as s e eo ypes and
p ejudices owa ds ce ain g oups wi hin public opinion (Wood &
Po e , 2019).
Poli ical pa ies and leade s inc easingly le e age social media o
connec wi h po en ial suppo e s and o e s. Howe e , he sp ead o
ha e ul messages on hese pla o ms con ibu es o a g owing disconnec
be ween ins i u ions and public opinion (Thelwall & Cugelman, 2017),
ul ima ely weakening us in democ a ic ins i u ions, such as he p ess
and poli ical pa ies (Fa kas & Schou, 2018). The poli ical campaigns o
Ba ack Obama (Cha es-Mon e o, 2017) and Donald T ump (P´
e ez
Cu iel & Lim´
on-Naha o, 2019) exempli y his end, whe e social media
became e ile g ound o disin o ma ion and ha e speech. On hese
pla o ms, use s sha e and commen on con en p ima ily o exp ess
hei ideological posi ions on speci ic issues a he han o engage in
meaning ul dialogue. In his s udy, we analyze how ce ain g oups o
use s can in luence public opinion and os e pola iza ion h ough co-
o dina ed ac ions. This occu s h ough he dissemina ion o ha e mes-
sages by accoun s o ganized in in e ac i e ne wo ks, which ampli y
ex eme na a i es on he X pla o m, especially in poli ical con ex s in
a ious coun ies.
2. Resea ch con ex
The ole o social media pla o ms in c ea ing and ansmi ing ha e
speech is inc easingly signi ican , ye o en concealed behind poli ically
co ec discou ses and social desi abili y bias—ou endency o p esen
ou sel es posi i ely while dis o ing ou ac ual posi ions (Lanz, Thiel-
mann, & Ge po , 2022). Poli icians equen ly employ his phenomenon
o a ac use s’ a en ion, inad e en ly sp eading umo s, s e eo ypes,
and p ejudices agains speci ic g oups (Pole o e al., 2021).
Kalsnes and Ihlebæk (2021), Ghasiya and Sasaha a (2022), Roslan
e al. (2022), and Mülle and Schwa z (2023) con i m ha pla o ms like
X and Facebook ha e become b eeding g ounds o ha e speech and
an i-mino i y sen imen s despi e mode a ion e o s, p ima ily due o
inconsis en egula o y amewo ks and pla o m-speci ic s a egies
(S ockmann e al., 2023; Wang & Kim, 2023), d i ing many use s o-
wa d al e na i e pla o ms like Bluesky, which expe ienced subs an ial
g ow h a e he 2024 U.S. p esiden ial elec ion (Roge s, 2024).
In he domain o poli ical communica ion, dis inguishing be ween
ha e speech and nega i e poli ical campaigning is impo an . While bo h
employ emo ional and pola izing s a egies, hei unde lying p inciples
and objec i es a e dis inc om each o he . Ha e speech is speci ically
di ec ed agains g oups iden i ied by a ibu es such as e hnic o igin,
eligion, gende , o sexual o ien a ion, leading o s igma iza ion,
exclusion, and e en inci emen o iolence (Ru lo e al., 2024, pp.
352–369). I s aim is no o engage in discou se on poli ical p oposals bu
o delegi imize he social belonging o ce ain g oups, he eby con a-
ening he undamen al democ a ic p inciples o equali y and plu alism.
Nega i e poli ical campaigning, while legi ima e, emains con en-
ious in elec o al compe i ion. I p ima ily in ol es he c i ique o po-
li ical opponen s, pa ies, o p og ams, aiming o unde mine hei
c edibili y o ques ion hei go e ning capabili ies (Sch¨
a e e al., 2024).
Al hough i may employ emo ional appeals such as ea o indigna ion,
i s ocus is on he poli ical ad e sa y a he han a social communi y
This a icle is pa o a special issue en i led: AI in poli ical communica ion published in Social Sciences & Humani ies Open.
* Co esponding au ho . A enida de la Paz 137, 26006 Log o˜
no (La Rioja), Spain.
E-mail add esses: [email p o ec ed] (D. Mo a eale-Cal anese), [email p o ec ed] (S. A ce-Ga cía), [email p o ec ed] (E. Said-Hung).
Con en s lis s a ailable a ScienceDi ec
Social Sciences & Humani ies Open
jou nal homepage: www.sciencedi ec .com/jou nal/social-sciences-and-humani ies-open
h ps://doi.o g/10.1016/j.ssaho.2025.102148
Recei ed 23 May 2025; Recei ed in e ised o m 11 Sep embe 2025; Accep ed 24 Oc obe 2025
Social Sciences & Humani ies Open 12 (2025) 102148
A ailable online 6 No embe 2025
2590-2911/© 2025 The Au ho (s). Published by Else ie L d. This is an open access a icle unde he CC BY-NC-ND license (
h p://c ea i ecommons.o g/licenses/by-
nc-nd/4.0/ ).
de ined by sensi i e iden i ies. None heless, his o m o messaging o en
becomes en angled wi h o he exp essions o ha e, he eby blu ing
dis inc ions and ul ima ely con e ging he wo phenomena.
Abuín-Vences e al. (2022) s a e ha ha e speech on social media
pla o ms causes desensi iza ion, inc eases p ejudice, and os e s social
exclusion o he subjec s o ha e, e en igge ing iolen ac s. This
phenomenon, which seems o be on he ise due o he p oli e a ion o
social media pla o ms, has led o g ow h in esea ch in his ield,
p o iding new pe spec i es o unde s and and add ess his issue. This is
especially ue i we ake in o accoun he e ec s o he echo chambe
and he bubble il e ha occu s in use s o cu en digi al en i onmen s,
which end up encou aging he cons an ansmission and epe i ion o
messages, pe sis en censo ship o non-majo i y posi ions, and in o -
ma ion sa u a ion (A ce-Ga cía e al., 2023; Guess e al., 2018).
Despi e academic ad ances, he need o be e unde s and he ca-
paci y ha use s ha e in cu en digi al scena ios o ecognize and ca y
ou s a egies aimed a in ensi ying poli ical and social pola iza ion om
hese ypes o con ex s is s ill p esen (Samy-Tayie e al., 2023). The
na u e and beha io o online ha e s emain la gely unknown, pa icu-
la ly ega ding hei po en ial coo dina ion in sp eading “ha e on he
In e ne ,” “ olling,” o “ha ing” (Jęd yczka e al., 2022). Recen
esea ch by Ma ínez To alba e al. (2023) highligh s he use o
AI-designed p opaganda echniques and mic osegmen a ion o in luence
use a i udes and beha io s on key opics. These ac i i ies can be ca -
ied ou by bo s (au oma ed compu e algo i hms), oll a ms (g oups o
humans using ake accoun s anonymously), and “Eli e bodies” (ac o s
se ing speci ic poli ical o economic in e es s).
As o u ing is a no able s a egy ha uses social media pla o ms o
sp ead ca e ully c a ed in o ma ion, c ea ing he illusion o g ass oo s
exp ession while co e ly in luencing public opinion (Elmas e al.,
2021). This app oach employs mic o- and nano-in luence s o exploi
loose social connec ions—aligning wi h G ano e e ’s (1973) “weak
ies” heo y— o build ex ensi e ne wo ks a ound a ge ed messages
con aining ha e speech, nega i e emo ions, and disin o ma ion
(A ce-Ga cía e al., 2023; P´
e ez, 2020). These ac ics in ensi y cybe -
space pola iza ion by le e aging he “selec i e e en ion” mechanism o
he selec i e exposu e heo y (Tu k, 2021), whe ein use s p e e en ially
emembe nega i e in o ma ion abou opponen s (Ga cía-O osa, 2021;
T´
o h e al., 2023).
The pola iza ion a ising om hese discu si e p ac ices necessi a es
dis inguishing be ween ideological pola iza ion, which is oo ed in
p og amma ic and doc inal di e ences, and a ec i e pola iza ion,
which is cha ac e ized by hos ili y and he delegi imiza ion o oppo-
nen s. As no ed by Díez e al. (2023), he in e ac ion be ween hese di-
mensions demons a es how digi al en i onmen s no only ampli y he
di e gence be ween poli ical posi ions bu also exace ba e nega i e
emo ions owa ds hose who hold hem. The esul ing o ches a ed ex-
p essions, g ounded in sociological ounda ions and le e aging bo h
ideology and emo ions o pola ize poli ical discou se, cons i u e wha is
now e e ed o as “Cogni i e Wa a e.” This concep aims o coun e
disin o ma ion and de elop ools o comba manipula ion wi hin digi al
communica ion channels (Case o-Ripoll´
es e al., 2023; S a o Maggio e
della Di esa, 2023).
I is essen ial o add ess his p oblem h ough s a egies ha p omo e
in o med and espec ul public deba e, encou age ci ic pa icipa ion,
and s eng hen mechanisms o comba he sp ead o ha e speech online.
E ec i e egula ion o ha e speech on social media pla o ms and
educa ional and awa eness- aising ini ia i es can help mi iga e i s
nega i e impac on public opinion and democ acy in he long un. To
his end, his pape ocused, o he gene al objec i e se , on es ima ing
he in luence o ha e accoun s in shaping he deba e on social media
pla o ms such as X by p omo ing a pola ized scena io o public opinion.
The pe iod o Ma io D aghi’s esigna ion and he call o elec ions ha
p eceded he ic o y o Gio gia Meloni and he a - igh ideologically
o ien ed B o he s o I aly Pa y was aken as a case s udy.
This case s udy examines how poli ical ac o s le e age social media
pla o ms o ad ance hei poli ical-elec o al in e es s, as Cloudy e al.
(2023) documen ed. These s a egies p ima ily in ol e p omo ing
emo ionally cha ged con en wi h na ionalis and populis hemes
designed o p o oke ea and ange owa d speci ic social g oups
(Tu˜
n´
on-Na a o & Bouzas-Blanco, 2023). This app oach aims o a ac ,
mobilize, and gene a e signi ican in e ac ions wi h po en ial o e s
h ough i al messaging in digi al spaces.
Resea ch by Be i and Lone (2023) on Ma eo Sal ini, leade o he
a - igh “Lega” pa y, demons a es how hese ac ics unde mine op-
ponen s’ c edibili y and in luence public pe cep ion. Such s a egies
h ea en na ional iden i y by dissemina ing di icul - o- e i y con en
ha heigh ens pe cep ions o insecu i y and p omo es di isi e “us
e sus hem” na a i es, ul ima ely in ensi ying pola iza ion in I alian
socie y (Ce i, 2023; Den i & Faggian, 2021).
This s udy examined he I alian case, conside ing ha in 2021, he
P esiden o he I alian Republic, Se gio Ma a ella, appoin ed Ma io
D aghi, a echnoc a wi h ex ensi e expe ience in in e na ional inance,
as P esiden o he go e nmen . D aghi, a o me p esiden o he Bank
o I aly and he Eu opean Cen al Bank (ECB), o icially accep ed he
ask o o ming a new cabine in Feb ua y 2021. The new p esiden ’s
di ec ion seemed o lean, on he one hand, owa ds echnical expe s in
cha ge o he mos s a egic a eas o he go e nmen and, on he o he
hand, owa ds he majo i y o pa ies eage o be pa o a go e nmen
ha would be esponsible o dis ibu ing esou ces on a la ge scale in
he coun y’s economy (Ga zia & Ka emans, 2021).
Despi e a o able public c i icism o he new echnical go e nmen ,
on July 14, 2022, D aghi announced his esigna ion a e Giuseppe
Con i’s pa y, “Mo imen o 5 S elle” (M5S) o 5 S a Mo emen in En-
glish, an impo an pa o he go e nmen coali ion, would no suppo
a con idence mo ion in he Sena e (Sica, 2021). A he Eu opean le el,
he mos ele an aspec o he all o he D aghi go e nmen was ha
he alignmen wi h he Eu opean Union and he A lan ic Alliance was
comp omised (Cha e, 2022), e lec ing he di ision in I alian public
opinion and he e u n o powe o a - igh populis igu es such as
Sal ini and Be lusconi, he la e being he leade o he igh -wing
neolibe al pa y “Fo za I alia” (FI) (Ocche a, 2022; Zulianello, 2020).
In his con ex , and aking ad an age o social media pla o ms, populis
poli ical mo emen s began o use social media pla o ms o cap u e he
a en ion and ob ain he consen o use s (Ma ella & B acciale, 2022).
Valen i (2023) obse ed ha ha e speech has become no malized
wi hin I alian poli ical communica ion on social media pla o ms such as
X, whe e i is equen ly employed as a populis ins umen . This o m o
ha ed is exp essed h ough a combina ion o e bal and non e bal el-
emen s, including images and ex s, which p opaga e s e eo ypical and
disc imina o y na a i es. In his con ex , Ie aci ano, Balenzano, Gi -
a di, Gemmano, and Comunello (2023) elucida e ha mo al disen-
gagemen associa ed wi h ha e speech on digi al pla o ms like X is
linked o con o mi y wi h es ablished ules and au ho i y, which
pa ially accoun s o he accep ance and pe pe ua ion o hese messages
in digi al en i onmen s. Consequen ly, poli ical discou se on hese
pla o ms s a egically u ilizes mul imodal and implici linguis ic
s uc u es o sub ly ye e ec i ely con ey disc imina o y con en
(Alonso, 2015).
In he I alian con ex , poli ical discou se on social media has
inc easingly mani es ed as ha e speech di ec ed a mino i ies, immi-
g an s, and poli ical ad e sa ies (Pas a, 2019; Piange elli, 2020). The
I alian poli ical sys em is cha ac e ized by p onounced ins abili y,
ma ked by equen go e nmen al changes, ansien coali ions, and
pa isan agmen a ion, complica ing he o ma ion o s able majo i ies.
Consequen ly, he execu i e b anch and i s leade s ha e assumed a
cen al ole in decision-making p ocesses (Fi ipaldi & Musella, 2022).
This ins i u ional ola ili y, combined wi h escala ing pola iza ion, in-
luences he s a egic use o social media o mobilize suppo bases,
dissemina e emo ionally cha ged messages and cons uc “us e sus
hem” na a i es. In his manne , pla o ms such as X exace ba e exis -
ing poli ical ensions and ein o ce ex eme ideological posi ions,
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
2
he eby posing signi ican challenges o social cohesion and he quali y
o democ acy in I aly (Musella, 2024).
Du ing Ma io D aghi’s go e nmen , digi al communica ion channels
we e s a egically used o gene a e opposi ion o his adminis a ion.
This agg essi e app oach os e ed social and poli ical pola iza ion in
I aly (Capano & Sand i, 2022; Ie aci ano & Cen ola, 2025), c ea ing
condi ions a o able o Gio gia Meloni’s ise o powe wi h he a - igh
pa y, B o he s o I aly (FdI), which ep esen ed he opposi ion du ing
D aghi’s enu e (Imbe i, 2021). Meloni’s elec o al success may ha e
bene i ed om he i aliza ion o ha e ul con en by coo dina ed ac-
coun s on pla o ms like X, which deepened poli ical pola iza ion. This
s a egy, combined wi h suppo om Sal ini and Be lusconi and hei
espec i e pa ies, helped Meloni become P ime Minis e , ma king a
igh wa d shi in I alian poli ics (Me lo, 2022).
Based on he analysis conduc ed hus a , ou wo king hypo hesis
posi s ha he poli ical pola iza ion obse ed du ing he examined
pe iod on pla o ms such as X was in luenced by accoun s ha ope a ed
in a coo dina ed manne . These accoun s may ha e employed s a egies
such as as o u ing o shape I aly’s poli ical discou se.
The hypo hesis is based on wha has been s a ed by au ho s
conside ed in his wo k, such as P´
e ez (2020) o A ce-Ga cía e al.
(2023), who poin ou ha he dissemina ion o messages wi h ha e
speech is no spon aneous o andom, bu ha he e is a p e iously
designed s a egy behind i , ca ied ou by use s who y o go unno-
iced, o achie e he posi ioning o na a i es ha legi imize hei
ideological posi ions, in he case o I aly, in a o o he poli ical p ojec
led by Gio gia Meloni and he poli ical pa y, wi hin he public opinion
o his coun y.
3. Me hodology
This s udy aims o e alua e how ha e accoun s on pla o ms like X
shaped public deba e and p omo ed poli ical pola iza ion du ing Ma io
D aghi’s esigna ion and he subsequen elec ion campaign ha led o
Gio gia Meloni’s ic o y wi h he B o he s o I aly (July–Augus 2022).
Building on he con ex o I alian poli ical pola iza ion ou lined in he
in oduc ion, we examine how his en i onmen acili a ed he ise o
pa ies such as Meloni’s. Ou analysis ocuses on how hese poli ical
ac o s s a egically used social media pla o ms o gain signi ican ep-
esen a ion in I aly’s poli ical landscape, employing he ac ics discussed
ea lie in his s udy.
3.1. Obje i es o he s udy
To achie e he gene al objec i e o he p oposed wo k, he ollowing
is p oposed:
•SO1: De e mine he o m o g ouping o ha e accoun s.
•SO2: To es ima e he ha e load and in ensi y p esen a ound he case
s udy.
•SO3: Es ablish he gene al cha ac e is ics o ha e ul accoun s.
To achie e he p oposed gene al objec i e, all messages we e
collec ed om Decembe 7, 2022, o Augus 16, 2022, and associa ed
wi h he hash ags #D aghi, #c isidigo e no, #c isigo e no, and #D a-
ghiOu . This pe iod saw he announcemen o Ma io D aghi’s esigna-
ion o he P esiden o I aly, Se gio Ma a ella (BBC News Mundo,
2022); he accep ance o he esigna ion and call o ea ly elec ions by
he I alian Pa liamen (Hu Pos , 2022); and he deba e gene a ed as a
esul o hese e en s, which led o he elec ion esul s o Sep embe
2022.
3.2. Da a collec ion
Messages we e collec ed using he RTwee lib a y (Kea ney, 2019) in
R so wa e, connec ing o X’s API e sion 1.1, wi h bo h hash ag
selec ion and geoloca ion pa ame e s a ge ing coun ies iden i ied by
B adshaw e al. (2021) as disin o ma ion sou ces and Eu opean na ions.
While his me hod ga he ed a subs an ial sample o I alian-language
messages con aining speci ied hash ags (796,654 wee s o al), i
could no gua an ee comple e co e age.
API 1.1 only pe mi ed access o 7-day da a windows, and collec ion
occu ed weekly wi h duplica es emo ed. B adshaw classi ied coun ies
by online campaigning capaci y (17 high-capaci y including China and
Russia, 37 medium-capaci y, and 27 low-capaci y including I aly), wi h
he geoloca ion echnique o e ing 77.84 % wo ldwide eliabili y and
88.15 % in Eu ope (a e age e o : 256 km) (Lop ei e e al., 2021;
Van-de -Veen e al., 2015). No ably, he collec ion yielded unexpec ed
olumes om Thailand (22,085 messages) and Albania (17,525
messages).
The sample size p o ided su icien di e si y o add ess he s udy’s
gene al objec i e; howe e , i is essen ial o no e ha some ha e speech
messages may ha e been emo ed because o X’s con en mode a ion
policies (X Help Cen e , 2023). Resea ch indica es ha X’s mode a ion is
limi ed in scope, wi h o e 80 % o p oblema ic messages emaining on
he pla o m (CBS News Mundo, 2022; Cen e o Coun e ing Digi al
Ha e, 2023).
F om all he p ocedu es applied in his wo k, i was possible o
iden i y he g oups ha pa icipa ed in he social media pla o m X
a ound D aghi’s esigna ion and o see i he e was ha e speech, i s main
hemes, and he cha ac e is ics ha we e obse ed in he speeches o
de ec he possible p esence o as o u ing o condi ion he decisions
and pe cep ions o use s, as well as o domina e discou ses and elec o al
p ocesses on he In e ne .
3.3. Da a p ocessing echniques
To unde s and he ne wo k s uc u es, iden i y he g oups pa ici-
pa ing in he discussion, and s udy he ha e speech u e ed o help
achie e he gene al objec i e p oposed in his s udy, he ollowing was
conduc ed:
•To mee Speci ic Objec i e 2 (SO2), his s udy analyzed ha e ex-
p essions in he en i e sample using na u al language p ocessing wi h
he Syuzhe algo i hm in R (Jocke s, 2017). This echnique iden i ies
and quan i ies ha e speech using p e-es ablished lexicons. The s udy
used Hu lex in I alian (2018 e sion), which was de eloped in he
o iginal language o p ese e local meanings and nuances wi hou
ansla ion issues (Bassignana e al., 2018). Hu lex is a lexicon
ini ially de eloped in I alian ha encompasses o ensi e, agg essi e,
and ha e ul ocabula y. I was c ea ed by esea che s a he Uni-
e si y o Tu in, building on a lexicon o mula ed by I alian linguis s
(Bassignana e al., 2018). The applica ion o his lexicon in esea ch
in ol ing he Syuzhe algo i hm acili a es he de e mina ion o he
in ensi y le el o ha e based on he occu ence o speci ic wo ds,
hei combined usage, and he p esence o o he con ex ual wo ds
ha may ampli y o diminish hei in ensi y. To p e en he
misiden i ica ion o e e yday, i onic, o ambiguous exp essions as
ha e ul, only messages exhibi ing an in ensi y om he uppe hi d
qua ile we e examined. This app oach ensu ed ha only messages
clea ly p opaga ing ha e we e conside ed. Al hough he o iginal
lexicon comp ises a ious seman ic ca ego ies o ha e, o his s udy,
hey we e consolida ed in o a single gene al ca ego y.
A e iden i ica ion, he esea che s es ablished a ha e a e age index
by di iding he ha e alue by he o al wo d coun in each message. This
wo d-coun de ec ion app oach, p e iously employed in o he ha e
de ec ion algo i hms (Pe ei a-Koha su, 2019), enables a ai compa ison
be ween messages o a ying leng hs, which is pa icula ly impo an o
social media con en classi ica ion.
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
3
•To comply wi h SO3, we conduc ed a ne wo k analysis using g aph
heo y o examine use connec ions h ough e wee ed messages
(Ba abasi, 2016). Da a we e p ocessed in R and analyzed using Gephi
10.1. The Fo ceA las2 algo i hm (Jacomy e al., 2014) p o ided a
g aphical ep esen a ion, whe eas he Lou ain algo i hm (Blondel
e al., 2008) g ouped use s in o clus e s (Chen e al., 2020). Each
clus e was cha ac e ized by examining he ollowing key me ics:
modula i y, a e age edge leng h, a e age deg ee, la ice diame e ,
eigen ec o cen ali y, be weenness cen ali y, and numbe o edges.
•Fo SO3, a ex mining p ocess was conduc ed, elimina ing wo ds o
s opwo ds ha did no con ibu e sense o meaning o he message
(e.g., a icles and ad e bs). The iden i ied ha e messages we e
g aphically ep esen ed in wo dimensions using a mul idimensional
scaling (MDS) plo . The wo ds a e ep esen ed by colo s ha allow
o he di e en ia ion o he di e en clus e s, which a e mos
equen ly linked o one ano he . Clus e s we e iden i ied using a k-
means algo i hm wi h k =7 o ensu e ha a su icien numbe o
opics we e iden i ied. This echnique was used o de e mine he
p ima y issues o ha e speech wi hin each clus e iden i ied in he
ini ial ne wo k heo y analysis.
The R da a ile used o collec ing, p ocessing, and analyzing he
messages is a ailable in he Supplemen a y Ma e ial (see Supplemen a y
Ma e ial XXXXXX o he R da a ile).
All s a is ical analyses and igu e c ea ion we e pe o med using R
so wa e, which enables he analysis o la ge da ase s h ough
p og amming.
3.4. P ac ical di icul ies in he me hodology
Du ing he implemen a ion o he me hodological design, se e al
di icul ies we e iden i ied ha a ec ed bo h he collec ion and analysis
o da a:
•Limi a ions o he X API ( e sion 1.1): The limi a ion o da a access o
a se en-day window necessi a ed weekly sampling, he eby
inc easing he isk o empo al o e lap and he po en ial loss o
pe inen wee s du ing download in e up ions. Addi ionally, he
inabili y o e oac i ely access messages emo ed due o mode a-
ion policies cons ained he comp ehensi eness o he sample, as up
o 20 % o p oblema ic messages may ha e been il e ed be o e hey
could be ecollec ed.
•Geoloca ion accu acy: The geoloca ion echnique based on loca ion
me ada a demons a ed a eliabili y o 88.15 % wi hin Eu ope, albei
wi h an a e age e o ma gin o 256 km in he geoloca ion. This
imp ecision poses challenges o accu a e e i o ial alloca ion,
pa icula ly in c oss-bo de egions o o use s wi h ambiguous
geoloca ion da a, which can po en ially in oduce noise in o he
analysis o he o igin o ake accoun s, such as hose in Thailand and
Albania.
•Fil e ing non-explici con en : In applying he Hu lex lexicon
h ough he Syuzhe algo i hm, i was necessa y o es ablish a
h eshold wi hin he uppe hi d qua ile o in ensi y o mi iga e alse
posi i es a ising om i ony o colloquial exp essions. While his
app oach e ec i ely isola es o e ly ha e ul speech, i may inad-
e en ly exclude sub le o implici exp essions o ha e, he eby
limi ing he cap u e o discu si e nuances and po en ially dimin-
ishing he ep esen a i eness o ce ain message ypes.
•Ne wo k analysis wi h Gephi: Execu ing he Fo ceA las2 and Lou ain
algo i hms on da ase s comp ising o e hal a million edges esul ed
in p olonged p ocessing du a ions and occasional ins abili y in he
con igu a ion o he modula i y pa ame e s. The adjus men o hese
pa ame e s necessi a ed mul iple manual i e a ions, he eby signi i-
can ly ex ending he analysis du a ion and po en ially in oducing a
bias in he de ec ion o smalle cell clus e s.
•In eg a ion o quali a i e me hodologies: Al hough his s udy does
no p ima ily ocus on quali a i e me hods, he absence o manual
con en analysis o use in e iews p ecludes he abili y o compa e
he indings o his esea ch wi h he di ec pe cep ions o pa ici-
pan s o mode a o s o X. Consequen ly, his limi a ion cons ains
he unde s anding o he mo i a ions unde lying he obse ed co-
o dina ion and he alida ion o au oma ically iden i ied as o u ing
s a egies.
4. Resul s
4.1. Clus e ing o ha e accoun s
O he 796,654 wee s w i en in I alian con aining he selec ed
hash ags, 22,085 came om Thailand and 17,525 om Albania. Glob-
ally, he messages consis ed o 161,613 o iginal wee s and 634,951
e wee s (RT), whe eas he messages collec ed om Thailand and
Albania did no include any e wee s. The emaining coun ies we e
disca ded because hey did no each 1000 messages.
The empo al dis ibu ion o messages was p ima ily concen a ed
on July 13, 2022, and he week ha ollowed, wi h he highes olume
o wee s occu ing be ween July 20 and 21, d opping d as ically om
July 23. These da es coincide no wi h he ini ial announcemen bu
wi h he con i ma ion o he esigna ion in he I alian Sena e. The
sequence o messages was simila bo h globally and in he wo coun ies.
Howe e , in Thailand, he highes ac i i y was eco ded on he 21s ,
whe eas in he o he cases, i was on he 20 h.
The ne wo k s udy shown in Fig. 1 shows highly in e connec ed
g oups wi h no clea sepa a ion be ween g oups. The s uc u e p e-
sen ed a modula i y o 0.588, wi h 558 communi ies o clus e s de ec-
ed. Among hem, six main g oups we e de ec ed, ep esen ing 97.33 %
o he o al e wee a ic a ound he analyzed hash ags. Thei poli ical
leanings a e desc ibed by examining he main accoun s in each clus e
by eigen ec o :
•MEL G oup (in g een in Fig. 1): Re ol es a ound ollowe s o Gio gia
Meloni and Ma eo Sal ini, wi h 22.19 % o a ic.
•M5S G oup (o ange): Re ol es a ound he 5 S a Mo emen (M5S),
as well as an i- accine and conspi acy g oups, wi h 11.41 %.
•CD G oup (pink): I alian cen e - igh poli ical endency (23.48 %).
•US G oup (blue): Mainly ul a-le (17.4 %).
•DC G oup (b own): g oup o Ch is ian Democ a poli icians and
jou nalis s (16.87 %).
•SIN G oup (pink): I alian ade unions and he media (5.98 %).
The emaining clus e s we e disca ded owing o hei low ep e-
sen a i eness o he es o he analysis.
Table 1 analysis e eals ha MEL, M5S, and CD clus e s exhibi
signi ican ly highe a e age deg ees be ween nodes, la ge g oup di-
ame e s, and g ea e dis ances compa ed o US, DC, and SIN clus e s,
wi h sel -desc ip ions aligning wi h he algo i hm’s classi ica ions
h ough e minology equen ly associa ed wi h igh -wing ideology.
These i s h ee g oups demons a ed mo e ac i e accoun s and ex en-
si e pa icipan ne wo ks, cha ac e ized by e y low modula i y (min-
imal egocen ism), esul ing in high in e connec i i y among accoun s
despi e appa en ideological di e ences be ween he accoun s. Such
cha ac e is ics e lec poli ical s a egies designed o maximize ou each
h ough c oss-cu ing na a i es such as na ionalism o Eu oscep icism,
whe e use s wi h clea ideological o ien a ions domina ed nea ly all
deba es on he X pla o m du ing he s udy pe iod. By os e ing dynamic
in e ac ions and p esen ing hemsel es as open, ci il mo emen s a he
han eli is en i ies, hese pa ies acili a e he dissemina ion o mass
messages and apid esponse capabili ies in pola ized con ex s, he eby
s eng hening he cohesion o hei digi al campaigns.
Tex mining analysis allows us o see he mos used wo ds o sel -
de ini ion by he use s ha make up each de ec ed g oup:
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
4
•MEL G oup: always, only, be, eedom, ee, u h, ne e , I aly,
wo ld, igh , poli ics, do, God, e e y, man, I alian, no g een pass (No
Vaccina ion mo emen ).
•M5S G oup: always, wo ld, jus , poli ics, being, I aly, le , lo e, ime,
s a s, communis , an i- ascis , ne e , wa , music, do, people.
•US G oup: she/he , always, li e, hings, lo e, wo ld, ime, accoun ,
lo e, addic ed, poli ics, like, w i e, ans.
•DC G oup: poli ics, jou nalis , an i ascis , li e, always, wo ld, ime,
yea s, being, communica ion, lo e, passiona e, Rome, ne e , you,
only.
•CD G oup: always, poli ics, p o-Eu opean, i alia i a, I lo e, an i-
ascis , I aly, only, ne e jou nalis , libe al, wo ld, lo e , eedom,
being, his o y, a , e e y, enginee .
•SIN G oup: news, poli ics, news, social, jou nalis , and zazoom (So-
cial Media News’ accoun , which has been e y ac i e on Twi e o
many yea s.), I aly, web, so, ime, si e, wo ld, all, accoun .
4.2. P esence o ha e and in ensi y
Once he g oups we e de e mined, messages con aining ha e we e
iden i ied, o aling 136,887 messages, which ep esen ed 17.18 % o all
wee s in he da ase . Among he messages om ab oad, 5997 ha e
messages (1052 accoun s) we e de ec ed in Thailand and 4553 (814
Fig. 1. Ne wo k map o he g oups.
Sou ce. Own elabo a ion using Gephi 10.1.
Legend. MEL: G oup cen e ed on ollowe s o Gio gia Meloni and Ma eo Sal ini; M5S: G oup linked o he Fi e S a Mo emen ; CD: I alian cen e - igh poli ical
endency; US: Mainly ul a-le ; DC: G oup a ound Ch is ian Democ a poli icians and jou nalis s; SIN: I alian ade unions and media.
Table 1
Na u e o he clus e s de ec ed.
MEL (16) M5S (4) US (21) DC (48) CD (15) SIN (39)
Mean g ade 5,36 6,99 1,31 1,58 5,60 1,35
Diame e 14 13 6 10 21 5
Mean dis ance 5,32 4,25 2,43 3,11 6,55 1,68
Nodes (beads) 14.785 (22,19 %) 7.600 (11,41 %) 11.590 (17,4 %) 11.241 (16,87 %) 15.645 (23,48 %) 3.986 (5,98 %)
Edges (RT connec ions) 79.219 (24,5 %) 53.127 (16,43 %) 15.204 (4,7 %) 17.753 (5,49 %) 87.634 (27,1 %) 5.372 (1,66 %)
Modula i y 0,28 0,33 0,72 0,65 0,31 0,77
Sou ce. Own elabo a ion using R so wa e.
No e. The numbe o he clus e s de ec ed is shown in b acke s.
Table 2
Numbe o ha e messages de ec ed pe clus e .
MEL (16) M5S (4) US (21) DC (48) CD (15) SIN (39)
To al 36.669 30.663 4.721 13.525 47.501 3.808
Albania 134 125 72 136 276 71
Thailand 171 184 101 173 344 173
Sou ce. Own elabo a ion using R so wa e.
No e. The numbe o he clus e s de ec ed is shown in b acke s.
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
5
accoun s) in Albania. The dis ibu ion by clus e s is p esen ed in Table 2,
highligh ing ha he highes numbe o ha e messages was concen a ed
in he MEL, M5S, and CD clus e s. Ha e messages om Albania and
Thailand ep esen 3.3 % and 4.4 %, espec i ely, o he o al, which,
al hough no a majo i y, is signi ican since almos 8 % o ha e messages
in I alian come om hese wo non-I alian-speaking coun ies.
The in ensi y o ha e speech a ied signi ican ly among he di e en
g oups analyzed. The MEL (0.467), CD (0.355), and M5S (0.324) g oups
showed he highes a e age ha e load pe wo d in hei messages, ol-
lowed by DC (0.306), US (0.297), and SIN (0.167). Howe e , he
a e age alues we e e y simila and low in all g oups, anging om
0.0311 o 0.0344.
Fig. 2. Mul idimensional ep esen a ion o he messages o each clus e .
Sou ce. Own elabo a ion, using R so wa e.
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
6
Fo he messages om Thailand and Albania, hei maximum (0.194
and 0.192, espec i ely) and mean alues (0.0314 and 0.0316, espec-
i ely) we e wi hin he o e all a e age, wi h no no able ex emes
compa ed o he o e all maximum o 0.467 and he mean o 0.0316.
No ably, 81.34 % o he e wee s (RT) a e on messages classi ied as ha e
speech. These messages had a median o 69 o wa ds and a mean o
234.4, compa ed o 47 and 194.7, espec i ely, o all messages. The ex
mining analysis, depic ed in Fig. 2, e eals he p edominan hemes in
he ha e speech o each g oup:
•MEL and M5S: They associa ed D aghi’s go e nmen c isis wi h he
people, in oducing hemes such as he g een pass and appeals o he
coun y’s his o y.
•DC: Focuses on he c isis o Sal ini and Con e.
•CD: Rela es he c isis o Be lusconi, Sal ini and Con e.
•US: C i icizes Con e and Sal ini, using de oga o y language owa ds
poli icians.
•DC: Accuses Sal ini o a lack o esponsibili y and con as s de-
moc acy wi h he “obscu an ism” o Meloni and Sal ini.
•CD: C i icizes Meloni and Sal ini o hei alleged p o-Pu in posi ions
and lack o A lan icism.
•SIN: Focuses on his o ical esponsibili y a e esigna ion and calls
o elec ions.
Table 2, he e o e, does no show insul s among he wo ds mos
equen ly used in ha e speech, bu i does es ablish he hemes and
objec i es, which a e mainly ocused on emo ional discou se (making
his o y) and a acks on well-known poli ical igu es.
Table 3 illus a es, by way o example, how ha ed is p ima ily
concen a ed on he opics indica ed: D aghi’s esigna ion, o he poli -
ical ac o s such as Sal ini and Meloni, and ancilla y issues including
an i- accine and Pu in- ela ed wee s. Ha ed is mainly mani es ed
h ough accusa ions o incapaci y, c iminali y, and dishones y. I was
obse ed ha some insul s con ained double le e s o a oid being easily
iden i ied by algo i hms.
A clea le el o ha e is he e o e de ec ed, al hough no o g ea in-
ensi y and wi hou c ossing legal bounda ies, as o he esea ch has
highligh ed (A ce-Ga cía e al, 2023).
4.3. Cha ac e is ics o he accoun s dissemina ing ha e
An analysis o he accoun s pa icipa ing in he deba e o ha e mes-
sages, acco ding o he clus e s iden i ied, shows (Table 4) ha hese a e
use s wi h a nanoin luence - ype p o ile, ypical o p epa ed
As o u ing- ype campaigns. Se e al aspec s indica e his end: a small
numbe o ollowe s and accoun s ollowed, low p oximi y o he leade s
o in luence s o each clus e , almos null impo ance in he ne wo k
(eigen ec o ), and nea ly non-exis en in e media ion be ween use s.
Howe e , di e ences we e obse ed be ween he clus e s. The MEL
and M5S clus e s showed e y high ac i i y in e ms o message emission
Table 3
Messages wi h highe ha e load de ec ed by clus e .
MEL
(16)
● The eading o he #c isiso go e nmen o he decla ed and non-decla ed PD o e s is as much as he mos idio ic subno mal bebe oid idio dishones disloyal insince e
hypoc i e and ake I ha e e e seen.
● #inondala7 his bad people un il yes e day wi h he MAFIOUS LEADER o COSA NOSTRA #Be lusconi now s ay wi h he Cen e Le !!! EVERYTHING in o de o
REMAIN HANGED TO THE ARMCHAIRS AND KEEP THEIR FATHER DELINQUENT CRIMINAL FARABUTTO AGUZZINO PERSECUTOR OF THE ITALIANS #D aghi
● #co eeb eakla7 #la iache i a #inondala7 Ci izens unde s and he game o hese MISERABLES? Now hey p e end Di isions … bu hen hey eplay o you again he
#Duce DELINQUENT FARABUTTO AGUZZINO CRIMINAL OF THE ITALIANS #D aghi o Head o Go e nmen (“Eu ope asks us”)
M5S
(4)
● 30 yea s o Go e nmen s o who emonge s, ma iosi, midge s and dance s ha e b ough us shi up o ou necks. Now hey all y o sa e hemsel es by blaming #M5S
and #Con e o e e y ne a iousness aking ad an age o he usual Ox People wi h gold ish memo ies. We dese e i . #PD #D aghi #Renzi
● 37 mise able despised Di Maio, 79 dead d agon D aghi, 43 illage idio Renzi. Te no on Rome and all. #D aghi #bu oon #disappea
● only he #PD de mme da s ays in he cou oom o suppo his mise able c iminal go e nmen . Laughing. #Le a puppe ucked up ac oss he boa d. Idio . Go o bed
peace ully. #ma a honmen ana #c isisgo e nmen #d aghi #D aghiGe OUT NOW
US (21) ● “The pa ies ha suppo ed D aghi o he end showed sense o s a e and esponsibili y” his ime I hal ag ee! In democ acy he sense o s a e is also o hose who and in
disag eemen wi h #D aghi Maybe hey a e s upid, dumb, igno an , bu no i esponsible
● Bu please. Con e ne e did any hing w ong o ally himsel wi h Pd and 60/70 in a go e nmen led by a pe son who conside ed himsel incapable, igno an and
concei ed. #c isisgo e nmen #CalendAI #Calenda #15July #unaFollia #ins ag amdown #ma a honmen ana
● a ool is alking abou # di and #d aghi is looking a him like he’s a mo on … ac ually his # di one is
DC (48) ● Poli ics, in I aly, is in a s a e ha o call pi i ul is an unde s a emen , bu he esponsibili y also lies wi h an indolen , boo ish and igno an elec o a e. Ma io D aghi you
do no dese e. #c isidigo e no
● #c isidigo e no anyone who b ings down a go e nmen in he cu en in e na ional con ex is simply an i esponsible mo onic imbecile useless good o no hing !
Thank you on behal o all ci izens.
● Tha his was he mos boo ish and clumsy Pa liamen in epublican his o y was al eady known. Tha i was also he dumbes was con i med yes e day. I aly hos age o
incompe en wa le s, his is. #c isisgo e nmen
CD (15) ● I’m old ha igno an men eca o Gianluigi #Pa agone ga e he usual b aying igno an bu oon show @gpa agone he one who blocks hose who c i icize him poli ely
#c isidigo e no
● @InOndaLa7 @co ado o migli he usual useless asshole. A mo on who ga e uel o he i e o populism. Now he would like o know abou #D aghi’s p i a e. You
mo on, wha do you ca e?
● Poli ics, in I aly, is in a s a e ha o call pi i ul is an unde s a emen , bu he esponsibili y also lies wi h an indolen , boo ish and igno an elec o a e. Ma io D aghi you
do no dese e. #c isidigo e no
SIN (39) ● Rega dless o who i will be and when hey will be held, he nex elec ions will ha e a winne . This one would be a mad, c azy, das a dly c e in i he decides o dispense
wi h he au ho i y, compe ence and se iousness o Ma io #D aghi. Pe iod. #p esscon e ence
● @jacopo_iacoboni I hink D aghi is no in i , I see i as bad. On Facebook, ou newspape ’s colleague F.M. w i es, “I hope his #c isidigo e no is no he e odi ec ed by
o eign o ces.” The e, one eally has o hink so. Ei he he is a mo on o a scound el, I don’ know
● The bes o D aghi’s quali ies? No being a poli ician. Bu we a e a guil ily igno an people and ha e a disa ec ed poli ical class composed o emba assing indi iduals.
And now? The beginning o a ho o mo ie #c isidiggo e nmen #D aghi #Ma ioD aghi #Sal ini #Meloni #m5s #PD
Sou ce. Own elabo a ion.
No e. The numbe o he clus e s de ec ed is shown in b acke s.
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
7
and bookma king, signi ican ly su passing he o he clus e s. The
numbe o daily messages and a o i es o hese accoun s was h ee o
ou imes highe han ha o he di e en clus e s du ing hei en i e
exis ence.
The MEL, M5S, and CD clus e s show high connec ion deg ees and
suppo o he accoun s wi h minimal in e media ion, sugges ing ha
hese accoun s con inuously ollow main accoun s and ac i ely e wee
con en . No ably, accoun s egis e ed in Albania and Thailand exclu-
si ely sen di ec messages wi hou e wee ing, indica ing a po en ially
s a egic app oach ha a oids public ampli ica ion o messages. This
dis inc i e beha io pa e n sugges s ei he a di e en communica ion
s a egy o accoun s being used o speci ic pu poses.
This s udy iden i ies a highly concen a ed communica i e dynamic
wi hin six mac o-communi ies, which collec i ely accoun o 97 % o
he ne wo k a ic. This dynamic is p edominan ly cha ac e ized by he
MEL, M5S, and CD clus e s, which a e no ed o hei high in e -
connec i i y, low modula i y, and employmen o ans e sal na a i es
associa ed wi h he poli ical igh . The empo al pa e n o digi al ac-
i i y co esponds o pa liamen a y miles ones, unde sco ing he in e -
dependence be ween poli ical e en s and discu si e p oduc ion.
The p esence o ha e messages, cons i u ing 17 % o he o al, is
no able o i s signi ican i al po en ial, as e idenced by an 81 %
e wee a e, despi e i s mode a e ocabula y and absence o explici
insul s. The in ol emen o accoun s om Thailand and Albania, cha -
ac e ized by a lack o e wee s and di ec b oadcas ing pa e ns, sugges s
as o u ing p ac ices aimed a s a egically in luencing I alian
discou se.
P o ile analysis indica es a p edominance o nano-in luence s cha -
ac e ized by low ne wo k cen ali y bu exhibi ing ele a ed le els o
hype ac i i y. This sugges s a dissemina ion s a egy elian on swa ms
o mic o-accoun s a he han on indi idual leade ship. The indings
depic a pola ized digi al ecosys em in which ha ed se es as a ec o
o communi y cohesion and a mechanism o poli ical in ensi ica ion.
5. Discussion
The analysis o 796,654 I alian-language messages conce ning Ma io
D aghi’s esigna ion iden i ied coo dina ion pa e ns ha subs an ia e
he ini ial hypo hesis ega ding as o u ing s a egies in I alian poli ical
discou se. The six p incipal g oups iden i ied (MEL, M5S, CD, US, DC,
and SIN) cons i u ed 97.33 % o he o al e wee ac i i y, wi h a
modula i y sco e o 0.588, indica ing signi ican in e connec edness
among communi ies ha appea o be ideologically di e se. This
con igu a ion implies s a egic coo dina ion a he han o ganic pola -
iza ion, pa icula ly gi en ha he MEL, M5S, and CD g oups exhibi ed
dis inc cha ac e is ics o low modula i y (minimal egocen ism) and
ele a ed message dissemina ion le els.
The iden i ica ion o 136,887 ha e speech messages, cons i u ing
17.18 % o he o al, p edominan ly concen a ed wi hin hese h ee
g oups, alongside he obse a ion ha 81.34 % o e wee s pe ained o
messages classi ied as ha e speech, unde sco es he coo dina ed na u e
o hese campaigns. The in ensi y o ha e speech a ied signi ican ly
among he g oups, wi h MEL (0.467), CD (0.355), and M5S (0.324)
demons a ing he highes alues. This pa e n aligns wi h he populis
s a egies documen ed by Ma ella and B acciale (2022), who used
emo ions as ins umen s o ga ne consen om in luen ial use s.
Ha e na a i es we e ope a ionalized using he Syuzhe algo i hm in
conjunc ion wi h he I alian Hu lex lexicon (2018 e sion), which
classi ies exp essions in o 17 ca ego ies o po en ially ha m ul con en .
To educe he incidence o alse posi i es esul ing om i onic o
colloquial exp essions, a il e based on he uppe hi d qua ile o in-
ensi y (≥0.324) was employed, ensu ing ha only exp essions wi h a
subs an ial emo ional cha ge we e iden i ied as ha e speech. The ha e-
a e age index, de e mined by di iding he ha e alue by he o al
wo d coun in each message, acili a es equi able compa isons be ween
messages o a ying leng hs —a c i ical conside a ion o con en
analysis on social media.
This me hodology demons a es ha ha e speech is no p edomi-
nan ly exp essed h ough explici insul s. Ins ead, i is con eyed h ough
highly emo ional and pe sonalized discou se, which, as no ed by
Abuín-Vences e al. (2022), os e s a pola ized and nega i ely cha ged
pe spec i e. The p ima y hemes cen e ed on he pe sonal a ibu es o
poli icians, emo ional appeals o na ional esponsibili y, and associa-
ions wi h o eign leade s such as Pu in—s a egies ha a e cha ac e -
is ic o a - igh populism, as documen ed in he li e a u e.
The indings con i m G ano e e ’s (1973) “weak ies” heo y, as
iden i ied nano-in luence accoun s (wi h ew ollowe s and posi ioned
on he ne wo k pe iphe y) p o ed e ec i e in in oducing pola izing
con en . These accoun s, wi h cha ac e is ics ypical o p epa ed
As o u ing campaigns, le e aged loose social connec ions o build
ex ensi e ne wo ks a ound a ge ed messages con aining ha e speech
and conspi acy na a i es.
Tu k’s (2021) selec i e exposu e heo y is e iden in he da a,
pa icula ly h ough he mechanism o “selec i e e en ion,” whe ein
use s p e e en ially ecall nega i e in o ma ion abou opponen s. This
cha ac e is ic is displayed by he MEL, M5S, and CD g oups, as e i-
denced by hei e wee s, which p edominan ly ocus on ha e ul mes-
sages ha encou age epe i ion, censo mino i y posi ions, and c ea e
in o ma ion sa u a ion. These indings align wi h Bobba and Ronca -
olo’s (2018) obse a ions ega ding heigh ened engagemen wi h
populis messages.
The disco e y o 22,085 messages o igina ing om Thailand and
17,525 om Albania, cons i u ing nea ly 8 % o he iden i ied ha e
Table 4
Na u e o pa icipa ing accoun s.
MEL (16) M5S (4) US(21) DC (48) CD (15) SIN(39)
Median Media Median Media Median Media Median Media Median Media Median Media
Followe s 567 2051 653 2142 318 2667 233 2767 530 2316 736 36523
F iends 617 1468 726 1456 559 1507 460 1132 720 1462 671,5 1400
B oadcas ed
messages his o y
24667 66022 29867 69688 8789 34136 8056 35171 17581 52572 20786 93027
Fa o i es 27298 65650 33960 68746 13322 48576 13035 42028 25896 69991 1407 7494
Accoun c ea ion 17/07/
2016
15/09/
2016
31/03/
2014
21/06/
2015
08/04/
2014
18/04/
2015
10/09/
2013
13/12/
2014
25/02/
2014
16/04/
2015
15/08/
2013
07/09/
2014
Deg ee 23 54,89 50 90,74 3 16,13 6 23,87 31 79,33 7 37,98
Closeness 0,208 0,506 0,227 0,558 0,170 0,727 0,156 0,964 0,199 0,284 0,174 2189
Be weenness 0 272003 0 354064 0 68490 0 46078 0 315525 0 52557
Eigen 0 0,005 0 0,097 0 0,002 0 0,002 0 0,007 0 0,006
Pos s/day 13,26 36,68 11,07 30,94 3,27 14,74 2,78 14,40 6,43 22,68 7,10 36,64
Fa o i es/day 14,68 36,47 12,58 30,53 4,95 20,97 4,49 17,21 9,48 30,19 0,48 2,95
Sou ce. Own elabo a ion.
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
8
speech, indica es a ansna ional aspec o his phenomenon. These ac-
coun s exhibi ed a unique pa e n: hey exclusi ely ansmi ed di ec
messages wi hou engaging in e wee s, sugges ing a po en ially coo -
dina ed s a egy aimed a a oiding public ampli ica ion. This beha io
implies ha accoun s a e u ilized o speci ic objec i es, possibly linked
o o eign in luence ope a ions documen ed by B adshaw e al. (2021) in
hei comp ehensi e global in en o y o o ganized media manipula ion.
This esea ch ad ances he heo e ical amewo k o As o u ing by
elucida ing how hese s a egies unc ion wi hin speci ic poli ical con-
ex s, he eby c ea ing “cogni i e wa a e” (S a o Maggio e della Di esa,
2023). The da a indica e ha As o u ing does no in a iably manipu-
la e public opinion di ec ly; a he , i s in en , coupled wi h indi iduals’
selec i e biases, in luences he selec ion and consump ion o news, ul-
ima ely shaping collec i e pe cep ions.
The in e ac ion be ween emo ional con en , in o ma ion dissemina-
ion, and opinion o ma ion highligh s he complexi y o he in o ma ion
ecosys em on social media and i s po en ial o in luence public pe cep-
ion. These manipula i e e o s can sub ly a ec he selec ion o use s o
ollow and how hese choices in luence hei wo ld iews, he eby
playing a signi ican ole in shaping he media landscape and, conse-
quen ly, in shaping opinions and social a i udes.
In addi ion o con i ming he speci ic objec i es, hese indings
ex end he weak- ie heo y amewo k by demons a ing ha pe iphe al
nano-in luence s can exe a signi ican in luence h ough coo dina ed
e wee s. This sugges s he need o e ise in o ma ion di usion models
ha cu en ly ocus solely on cen al ac o s. Mo eo e , he high i ali y
o messages classi ied as ha e speech, despi e hei low lexical in ensi y,
unde sco es he e ec i eness o sub le emo ional s a egies and he in-
adequacy o exis ing mode a ion sys ems. This has impo an implica-
ions o bo h egula o s and digi al pla o ms. This phenomenon
pe pe ua es a cycle o selec i e ein o cemen and in o ma ion sa u a-
ion, which, acco ding o selec i e exposu e heo y, exace ba es a ec-
i e pola iza ion by p io i izing ha m ul con en o e a ional
delibe a ion.
5.1. Limi a ions o he s udy and u u e lines o esea ch
Se e al me hodological limi a ions cons ained his s udy’s scope.
Fi s , he analysis pe iod was con ined o July–Augus 2022, which
p ecludes he assessmen o As o u ing dynamics beyond an immedi-
a e elec o al con ex . Second, eliance on he X API 1.1 necessi a ed da a
e ie al windows o only se en days, equi ing weekly sampling ha
could esul in empo al o e laps and he omission o pe inen wee s
be ween download in e als. Addi ionally, up o 20 % o po en ially
ha e ul messages may ha e been au oma ically dele ed be o e collec ion
because o he pla o m’s mode a ion policies, limi ing he comp e-
hensi eness o he sample. Las ly, he geoloca ion echnique based on
loca ion me ada a p o ided a eliabili y o 88.15 % in Eu ope, bu wi h
an a e age ma gin o e o o 256 km, which in oduced noise in o he
e i o ial a ibu ion o accoun s, pa icula ly in bo de egions o wi h
ambiguous me ada a.
To add ess hese limi a ions and enhance he comp ehension o
as o u ing and pola iza ion s a egies, i is ecommended ha u u e
esea ch be conduc ed as ollows:
•Ex ending he empo al scope o he s udy beyond elec o al pe iods
o enable di e en ia ion be ween disc e e elec o al campaigns and
con inuous in luence ope a ions.
•Employ mul imodal me hodologies ha in eg a e ex analysis wi h
au oma ed image and ideo de ec ion o iden i y manipula i e ele-
men s ha a e no solely lexical bu also isual o audio isual.
•Conduc compa a i e analyses ac oss mul iple Eu opean na ions o
disce n ansna ional pa e ns o media manipula ion and he in lu-
ence o o eign en i ies on na ional democ a ic p ocesses.
•In eg a e quali a i e me hodologies, such as conduc ing in e iews
wi h pla o m mode a o s o pe o ming manual con en analysis o
ep esen a i e cases, o complemen and enhance quan i a i e ind-
ings and subs an ia e au oma ically iden i ied coo dina ion
s a egies.
•Assess he implemen a ion o upda ed API e sions and al e na i e
da a sou ces, such as academic APIs o wee a chi ing ools, o
add ess he pla o m’s empo al and mode a ion cons ain s.
6. Conclusions
This s udy empi ically alida es he hypo hesis ega ding he p es-
ence o coo dina ed poli ical pola iza ion s a egies in X du ing he c isis
o D aghi’s go e nmen . An analysis o 796,654 messages e ealed ha
17.18 % con ained ha e speech, p edominan ly concen a ed wi hin
h ee g oups (MEL: 22.19 %, M5S: 11.41 %, and CD: 23.48 %),
demons a ing cha ac e is ics consis en wi h as o u ing ac ics. The
subs an ial p opo ion o e wee s in messages classi ied as ha e (81.34
%) and he coo dina ion pa e ns iden i ied h ough ne wo k analysis
co obo a e he s a egic a he han o ganic na u e o he obse ed
pola iza ion.
The key heo e ical con ibu ions o his s udy encompass he
empi ical alida ion o G ano e e ’s weak- ie heo y wi hin he con ex
o digi al manipula ion, he elucida ion o as o u ing mechanisms
h ough sub le emo ional na a i es ha ci cum en algo i hmic
de ec ion, and he iden i ica ion o ansna ional dimensions in in lu-
ence ope a ions, no ably in ol ing signi ican pa icipa ion om
Albania and Thailand. Fu he mo e, his s udy ad ances he unde -
s anding o ha e speech as a s a egic ool ha , despi e i s low in ensi y,
main ains sus ained e ec i eness by e ading con en mode a ion.
These indings unde sco e he necessi y o enhanced egula o y
amewo ks ha acknowledge he nuanced na u e o s a egic ha e
speech. Addi ionally, he e is a need o de elop de ec ion ools ha
accoun o pa e ns o empo al and geog aphic coo dina ion, as well as
o implemen media li e acy p og ams aimed a inc easing awa eness o
emo ional manipula ion echniques. Fo digi al pla o ms, hese esul s
highligh he impo ance o moni o ing no only explici con en bu also
he pa e ns o coo dina ed beha io and suspicious c oss-bo de
ac i i ies.
This esea ch elucida es how speci ic poli ical e en s ac as ca alys s
o he p oli e a ion o ha e speech, he eby con ibu ing o social and
poli ical pola iza ion h ough he p omo ion o a - igh na a i es. The
da a e eal a de ini i e co ela ion be ween nega i ely cha ged mes-
sages and poli ical pola iza ion, pa icula ly when hese messages a ge
poli ical igu es and sensi i e opics such as he Russia–Uk aine wa o
an i- accine mo emen s. This dynamic po en ially acili a es he esu -
gence o a - igh populis leade s in posi ions o powe .
The p esen indings illus a e ha As o u ing ini ia i es, e en
hose cha ac e ized by minimal lexical in ensi y, sus ain hei e ec-
i eness by ci cum en ing con en mode a ion and coo dina ing on a
ansna ional scale. The capaci y o hese campaigns o dissemina e
pola izing na a i es highligh s he c i ical need o enhance he de ec-
ion o beha io al pa e ns and o i y public media li e acy policies.
CRediT au ho ship con ibu ion s a emen
Da ia Mo a eale-Cal anese: W i ing – o iginal d a , In es iga-
ion, Fo mal analysis, Concep ualiza ion. Se gio A ce-Ga cía: W i ing –
o iginal d a , Valida ion, Me hodology, Da a cu a ion. Elias Said-
Hung: W i ing – e iew & edi ing, P ojec adminis a ion, In es iga ion,
Concep ualiza ion.
E hical s a emen
E hical app o al is no applicable o his manusc ip .
D. Mo a eale-Cal anese e al.
Social Sciences & Humani ies Open 12 (2025) 102148
9
