Dynamics of Personal Responses to Terror Attacks: A Temporal Network
Analysis Perspective
Ema Ku
ˇ
sen
1
and Mark Strembeck
1,2,3
1
Vienna University of Economics and Business, Vienna, Austria
2
Secure Business Austria (SBA), Vienna, Austria
3
Complexity Science Hub (CSH), Vienna, Austria
Keywords:
Death Anxiety, Emotional Pain, Network Dynamics, Temporal Network Analysis, Terror Attack, Twitter.
Abstract:
In this paper, we analyze responses to terror attacks through the lens of the Terror Management Theory. We
focus on the temporal evolution of Twitter messages that convey death anxiety, emotional pain, as well as
positivity. We model the responses to terror attacks as personal reactions that include the use of a first person
singular pronoun along with cues of affect and personal concerns. In order to detect these textual features,
we used the Linguistic Inquiry and Word Count (LIWC) tool. Our data-set includes tweets related to three
terror attacks: the 2017 Manchester terror attack, the 2019 Christchurch terror attack, and the 2020 Vienna
terror attack. Our analysis is based on 3.8 million tweets that have been sent by 1.6 million users. The results
indicate that positive messages associated with the use of religious words (e.g., messages of prayers and hope)
dominate over those that convey emotional pain and fear of death. This points to a tendency to spread hope
and empathy in the aftermath of a terror attack. We found that the acute phase of a terror attack exhibits a high
volume of messages that sharply decline in the immediate aftermath. In contrast, positive messages exhibit
smaller peaks even one week after a terror attack happened.
1 INTRODUCTION
People who face immediate threats to their own life,
their family members, or friends, tend to respond
with elevated emotions of fear, anxiety, anger, and
grief (Hardy and Miller, 2022; Ku
ˇ
sen and Strembeck,
2021). Such reactions to crises have long been in
focus of behavioral scientists who found that people
universally respond with an initial disbelief, followed
by a wide spread of information and rumors, venting
of intense emotions, as well as various collective reac-
tions (incl. collective panic, organized help and rescue
efforts). Other studies also showed that expressions of
solidarity, togetherness, as well as blame commonly
occur before people can return to their normal routine
(Flynn, 1997; Smelser and Mitchell, 2002).
Terror attacks are a specific type of crisis that hap-
pen unexpectedly and without a prior warning. More-
over, terror attacks pose an immediate threat to the
well-being and life of those directly affected, aim
to destabilize our world-view (Miller and Landau,
2005), and give rise to an intense anxiety. The latter
is strongly related to the reminder of our own mortal-
ity which is a focal point of Terror Management The-
ory (TMT) (Greenberg et al., 1986; Rosenblatt et al.,
1989). In general human beings are not only aware
of their own mortality but also have a desire for self-
preservation. In this light, a positive worldview serves
as an effective barrier to death anxiety (Kastenbaum,
2007).
Death anxiety related to terror attacks is univer-
sal (Greenberg et al., 1986) and not only concerns
those who are directly affected, but also those who
are reminded of past experiences of terror. (Rosen-
blatt et al., 1989) further showed that people are of-
ten reminded of their own mortality simply by read-
ing newspapers or watching the news on TV. Thereby,
such studies confirm that the fear of death caused by
terror attacks is one of the strongest psychological
threats (Becker, 1997).
As in any type of crisis, Online Social Networks
play an important role during terror attacks. They
serve an informational purpose during the acute phase
of a terror attack (Steensen, 2018; Simon et al., 2014)
and facilitate the organization of crisis relief efforts as
well as expressions of togetherness in the aftermath of
a terror attack (Merrill et al., 2020). While multiple
studies focused on the types of messages and the vol-
36
Kušen, E. and Strembeck, M.
Dynamics of Personal Responses to Terror Attacks: A Temporal Network Analysis Perspective.
DOI: 10.5220/0011078100003197
In Proceedings of the 7th International Conference on Complexity, Future Information Systems and Risk (COMPLEXIS 2022), pages 36-46
ISBN: 978-989-758-565-4; ISSN: 2184-5034
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All r ights reserved
ume of messages sent in each phase of a terror attack,
it has rarely been studied how specific types of social
media messages are exchanged and which temporal
communication patterns emerge during terror attacks.
In this paper, we investigate the temporal pat-
terns resulting from messages that convey death anx-
iety, emotional pain, as well as positive perspectives
about an otherwise bleak future (associated with one’s
faith). We model these responses as personal stories
told from a person’s individual perspective (indicated
by the use of the first singular personal pronoun) in or-
der to make sense out of a terror attack. According to
(Hermans and Hermans-Jansen, 1995), self-narratives
serve as a good indicator of situations that have a high
personal meaning during an episode in the speaker’s
life.
Our findings bring forth a couple of valuable in-
sights into the psychological underpinnings of terror
attacks. We extend the existing literature on human
responses to terror attacks by showing that positive
messages associated with the use of religious words
(e.g., messages of prayers and hope) play a special
role in the acute phase of the terror attack as well as
in its aftermath.
The remainder of this paper is organized as fol-
lows. In Section 2 we provide a summary of related
work. Section 3 describes our research method. Re-
sults are presented in Section 4 and discussed in Sec-
tion 5. We conclude the paper in Section 6 and pro-
vide directions for future research.
2 RELATED WORK
Informational Role of Twitter. (Simon et al., 2014)
studied the role of Twitter during the 2013 Westgate
mall terror attack in Kenya and found that initially
an uncontrolled surge of messages occurred. Though
there are obvious positive effects of such a high rate
of message sharing (e.g., social bonding with the
community, stress relief), the authors found that the
terrorists also monitored Twitter to acquire informa-
tion about the actions of the police and armed forces.
Along these lines, a study by (Oh et al., 2011) found
the same phenomenon during the 2008 Mumbai ter-
rorist attack. As a result of the uncontrolled surge of
messages, sensitive situational information about the
officials’ operational activities were revealed to the
terrorists.
Another study on the 2013 Westgate mall in
Kenya analyzed narratives formed by a terrorist group
(Mair, 2017). This study found that the terrorists
took control of their own narrative by not providing
any links to external websites or engaging with other
Twitter users. Their sole purpose was to spread their
ideology and justify their actions.
Twitter as a Channel for Collective Sharing of
Emotions. (Bruns and Hanusch, 2017) focused on the
dissemination of audiovisual content during the 2015
Paris terror attack and the 2016 Brussels terror attack.
This study found that the intent of audiovisual content
was predominantly of an affective nature (e.g., photos
of landmarks lit up in French national colors, images
with text “Pray for Paris”) and informational nature
(eyewitness photos). The study pointed to the im-
portance of affective intent in the audiovisual content
as it reveals how a collective responds to an attack.
Another study on collective emotions is presented in
(Steensen, 2018), who found that the narratives of em-
pathy associated with the #PrayForOslo and #Pray-
ForNorway hashtags emerged about two and a half
hours since the 2011 attack in Oslo and on the island
of Utøya in Norway. One week after the terror attack,
tweets of grief and love (#oslove, #showyourhearts)
dominated the Norwegian Twitter discourse.
The 2013 Woolwich/London terror attack was
studied by (Burnap et al., 2014). They found that af-
fects conveyed in tweets are directly associated with
message diffusion rates. In particular, the authors
found that tweets that convey a high level of tension
and antagonism circulated for a shorter amount of
time compared to those with a low level of these two
sentiments. Several important implications were re-
ported in this study, one of them being the role of pos-
itive messages which were shown to propagate longer
than the negative messages.
Twitter as a Channel for Sense-making. In terms
of message creation, (Kwon et al., 2017) found that
tweets sent during terror attacks predominantly fo-
cus on present (rather than past or future) and in-
cluded detailed and concrete descriptions of the event
rather than an abstract account of the event. (Eriks-
son, 2016) analyzed post-attack responses to the 2011
terror attacks in Oslo and on the island of Utøya, Nor-
way. The results revealed that people dealt with the
collective trauma by expressing solidarity and col-
lectively tried to make sense of what had happened
by searching for the underlying causes for the attack.
The latter discussion especially focused on the polit-
ical ideology of the perpetrator and those sharing a
similar ideological viewpoint. Interestingly, the study
found that people try to compare and reference dif-
ferent violent incidents to construct a narrative of a
growing trend of violence.
Twitter as a Channel for Pro-social Behavior. Dur-
ing the acute phase of a terror attack, a collective sol-
idarity towards those who seek shelter may emerge.
One such example was studied in (Reilly and Vicari,
Dynamics of Personal Responses to Terror Attacks: A Temporal Network Analysis Perspective
37
2021) during the 2015 Paris terror attack where the
hashtag #PorteOuverte (open door) was highly shared
on Twitter as means of offering a safe place to hide
from the terrorists roaming the streets of Paris. This
inspired a series of similar initiatives (Tikka, 2019),
such as #openhouse during the 2016 attack in Brus-
sels, #offenet
¨
ur during the 2016 attack in Munich,
#opendoors during the 2016 attack in Nice, #room-
formanchester during the 2017 attack in Manchester,
#openstockholm during the 2017 attack in Stock-
holm, or #bedinbarcelona during the 2017 attack in
Barcelona.
In addition to offering shelter, another form of
pro-social behavior during terror attacks is the expres-
sion of togetherness and unity. While studying the
responses to two terror attacks (Ezzati, 2021) found
that people are united through common values, such
as tolerance, co-existence, love, justice, peace, and
fraternity. The author pointed out that the purpose of
such value-conveying messages is to dissociate reli-
gious groups and members of a certain ethnicity from
the perpetrator(s).
3 METHOD
In this paper we analyze temporal characteristics of
emotional human reactions to terror attacks through
the lens of TMT. Our paper is guided by the following
main research question:
How do responses to terror attacks evolve over
time?
To further elaborate on this question, we explore the
following types of human responses to terror attacks:
1) emotional pain, 2) death anxiety, and 3) messages
of positivity associated with a use of religious words,
and define the following sub-questions:
How does the expression of emotional pain by
those affected evolve over time?
How does the expression of anxiety related to own
mortality by those affected evolve over time?
How does the expression of positivity related to
the use of religious words evolve over time?
In order to answer to these questions, our research
method follows three phases, as outlined below.
Data Extraction. We extracted the data by using
Twitter’s Search API and a list of predefined hashtags
and key terms for each terror attack
1
. The data extrac-
1
Christchurch terror attack: #NewZealandTerroris-
tAttack, #Christchurch, #ChristchurchMosqueShootings,
#ChristchurchMosqueAttack, #ChristchurchTerrorAttack;
tion started on the day of the terror attack and contin-
ued for another two weeks. After removing duplicate
entries, our data-set comprised 3,854,670 tweets au-
thored by 1,643,939 Twitter users. Details are pre-
sented in Table 1.
Data Pre-processing. In our analysis, we used
the Linguistic Inquiry and Word Count (LIWC) tool
(Pennebaker et al., 2015) to identify psycholinguistic
features in Twitter messages. LIWC’s internal lexicon
is organized into a set of categories, each designating
a specific psycholinguistic phenomenon. For exam-
ple, a category called “Personal Concerns” includes
variables “religion”, “death”, “money”, and “home”
along with their corresponding internal lexica, while
the category Affective Processes” includes anxiety,
fear, anger, positive emotions, and negative emotions,
(see (Pennebaker et al., 2015) for details).
We first defined a rule to capture a narrative (main
theme) conveyed in each tweet. The rule was as fol-
lows:
f (x,y) =
(
x + y, if x > 0 y > 0
0, otherwise
where x and y represent a variable in the category
“Personal Concerns” and Affective Processes” re-
spectively. If both values were positive, we multiplied
their sum with the LIWC score of the first person sin-
gular personal pronoun. The LIWC variables used in
this paper are outlined in Table 2 along with a note
on the rationale behind the choice and application of
LIWC variables.
Network Construction. Upon inferring the scores
for emotional pain, death anxiety, and religion (in this
context associated with positivity), we measured the
daily intensities of the three personal responses. For
each of the three types of personal responses, we con-
structed temporal direct messaging (DM) networks
for each type of a narrative by tracing the @-mentions
in the text
2
of a tweet. Users mentioned via the @-
mechanism are considered target nodes in our net-
Manchester terror attack: #ManchesterArena, #manch-
esterstrong, “terror manchester”, “manchester ariana
grande”; Vienna terror attack: “vienna terror”, “terrorist
attack #Vienna”, “#ViennaAttack””, “#Viennashooting”,
“#prayforvienna”, “#wienATTACK”, “angriff vienna””,
“#terrorwien”, “#PrayforWien”, “#viennaattacks”, “#Vien-
naTerrorAttack”, “#austriaAttack”, “sorgen Wien”, “#Vi-
ennaterror”, “#ViennaTerroristAttack”, “#viennapolice”,
“#austriashooting”, “terror wien”, “wien Hintergrund”, “vi-
enna background”, “#Schwedenplatz”, “wien #staysafe”,
“vienna #staysafe”, “#StayStrongAustria”, “#zibspezial”,
“#Nehammer”, “#0211w”, “@ORFBreakingNews”,
“#Synagoge”, “#Schießerei”, “#terroranschlag wien”,
“#terroranschlag vienna”, “#schleichdiduoaschloch”.
2
Note that we removed the retweets prior to a network
construction.
COMPLEXIS 2022 - 7th International Conference on Complexity, Future Information Systems and Risk
38
Table 1: Basic information about the data-set.
Event Time period Tweets Retweets Usernames Language
Christchurch 15 – 28 March, 2019 1,775,901 1,567,946 (88%) 846,858 en
Manchester 22 May – 4 June, 2017 1,582,586 1,232,085 (77.8%) 675,183 en
Vienna 2 – 15 Nov, 2020 496,183 378,171 (76%) 202,719 en, de
Table 2: The choice of LIWC variables.
LIWC variables Explanation
Emotional pain
“I” × (sadness +
anger)
(Junghaenel et al., 2017; Rude et al., 2004; Tausczik and Pennebaker, 2010) indicated that
those who experience emotional and physical pain tend to draw attention to themselves (via
a first person singular pronoun) and use words that indicate their emotions, especially sad-
ness and anger.
Example:
“I’m sorry I’m sorry I’m sorry I send my love if I could come hold all the families #Manch-
esterArena”,
“I didn’t sleep a second. I am tired. My head hurts. My heart hurts. #PrayforVienna”,
“I f**ing hate humanity man #ChristchurchMosqueAttack”
Death anxiety
“I” × (death +
anxiety)
(Greenberg et al., 1986) indicated that threat to one’s mortality is reflected through anxiety
over death, either an own death or death of someone else.
Example:
“Every time I see the confirmation picture of another death in Manchester attack I feel so
much pain... #manchester #terror”,
“#PrayforVienna I swear I’ve never been so scared of death”,
“I just learned about #Christchurch terrorist massacre... I’m just...appalled. Speechless. I
just can’t...
Religion
“I” × (religion +
positive affect)
(Flynn, 1997) found that religious beliefs provide hope for the otherwise bleak future.
Example:
“I pray for the victims. I pray for the families of the victims. I pray for #Manchester. I pray
for Ariana Grande. I pray for the world.,
“My heart is sore for my other hometown tonight. God bless Vienna #viennaterrorattack”,
“I hope the victims find paradise #NewZealandTerroristAttack”
work, while the author of a corresponding tweet is
considered a source node. Basic information about
our networks is presented in Table 3.
Data Analysis. We analyzed our networks with
respect to their underlying temporal dynamics that
emerge due to direct message exchanges among the
Twitter users. Subsequently, we analyzed the tempo-
ral networks with respect to their momentary struc-
ture (edge formation patterns and dyad density). To
construct and analyze the temporal networks, we used
tsna (Bender-deMoll and Morris, 2021) R package.
Our analysis was conducted on a quad-core machine
with Intel(R) Core(TM) i7-5600U CPU @2.60GHz.
4 RESULTS
The results revealed distinctive temporal patterns
among the three personal responses analyzed in this
paper. As shown in Figure 1, all three emotional re-
Table 3: Basic information about each network.
Event Nodes Edges
EMOTIONAL PAIN
Christchurch 241 133
Manchester 324 264
Vienna 2344 2675
DEATH ANXIETY
Christchurch 140 78
Manchester 282 246
Vienna 1555 1728
RELIGION
Christchurch 1101 690
Manchester 648 540
Vienna 3498 4106
sponses are present throughout the 14-day extraction
period in all data-sets. We observed a higher level
of death anxiety during the acute phase of each ter-
Dynamics of Personal Responses to Terror Attacks: A Temporal Network Analysis Perspective
39
ror attack and its gradual decrease over time. Positive
messages (associated with the use of religious words)
are dominant throughout the entire extraction period,
while emotional pain peaks in the immediate after-
math and afterwards decreases over time.
DATE
INTENSITY
18 March 25 March
DATE
22 May
29 May
2 Nov
9 Nov
DATE
INTENSITYINTENSITY
Death anxietyEmotional pain
Religion
CHRISTCHURCH
MANCHESTER
VIENNA
Figure 1: Intensities of emotional pain, death anxiety, and
religion in each data-set.
Table 4 shows reactions to the tweets that
convey death anxiety, emotional pain, and re-
ligious words. The results point to a tendency
to highly retweet messages that convey emo-
tional pain (µ(Christchurch) = 873.72 ± 788.76,
e
x(Christchurch) = 571; µ(Manchester) =
27303.87 ± 11746.99,
e
x(Manchester) = 32373;
µ(Vienna) = 7560.01 ± 6182.08,
e
x(Vienna) = 4620)
while those that convey death anxiety re-
ceived substantially smaller number of retweets
(µ(Manchester) = 28.13 ± 39.54,
e
x(Manchester) =
5; µ(Vienna) = 561.15 ± 773.70,
e
x(Vienna) = 186;
with an exception of the Christchurch terror at-
tack data-set). In terms of “likes”, the results are
inconclusive.
Next, we turn to the analysis of temporal commu-
nication networks. Three examples of daily message-
exchange networks shown in Figure 2 reveal that
the networks substantially change their structure over
time (note that, in our analysis, once an edge is
formed, it does not cease to exist but a new edge is
added in each time step).
3
In order to describe the
evolution of each network, we first report on the mea-
sures of momentary structure.
Figure 3 illustrates daily edge formations for each
network and shows that a great majority of edges
are formed in the first two days of the terror attacks
we analyzed. The edge formation count sharply de-
creases during the immediate aftermath and only oc-
casionally new edges are formed. This is illustrated
via small spikes in Figure 3 (we point the attention
to “religion” for the Christchurch and Vienna terror
attacks). We also observed a substantially higher vol-
ume of edges formed in the religion-related networks
throughout the data-extraction period for all terror at-
tacks in our study. For example, the Manchester terror
attack counts between 130 and 160 edges for the ex-
change of emotional pain and death anxiety and over
340 edges for messages that convey positivity and re-
ligious words (which corresponds to an increase of
approx. 100%).
We also measured dyad density. The results indi-
cate that for all possible dyads, the fraction of time
they are tied is relatively low in all networks (see Fig-
ure 4). For example, if we randomly select a pair of
nodes from the network representing emotional pain
(Vienna), there is only about 5 in 10,000 chance that
the pair will be connected by an active edge. How-
ever, one pattern still emerges across all terror at-
tacks we analyzed the dyad duration of all net-
works representing religion (positivity) is higher than
the dyad duration in emotional pain and death anxiety
networks.
5 DISCUSSION
Our study on the evolution of messages expressing
emotional pain, death anxiety, and positivity brings
forth some interesting insights into human (online)
behavior during terror attacks. While our focus was
on the personal reactions to the terror attacks, we ob-
served a high dominance of positive messages related
to the use of religious words throughout the data ex-
traction period. We found that all personal narratives
initially show a sudden surge in a message volume,
which is in line with the findings presented in (Simon
3
Our animated networks are available at
https://nm.wu.ac.at/nm/Complexis22-supplemental.
COMPLEXIS 2022 - 7th International Conference on Complexity, Future Information Systems and Risk
40
Table 4: User reactions to tweets conveying emotional pain, death anxiety, and religion.
RETWEETS LIKES
Event
Mean Median Max Mean Median Max
EMOTIONAL PAIN
Christchurch 873.72±788.76 571 1918 3.35±119.2 0 9590
Manchester 27303.87±11746.99 32373 32373 1.22±140.55 0 23247
Vienna 7560.01±6182.08 4620 13752 7.50087±745.73 0 114013
DEATH ANXIETY
Christchurch 1839.74±1804.45 1770 4654 2.28±83.91 0 6842
Manchester 28.13±39.54 5 127 2.14±20.84 0 614
Vienna 561.15±773.70 186 2008 4.67±81.20 0 3918
RELIGION
Christchurch 1123.84±1617.37 314 4731 3.57±183.00 0 31906
Manchester 24729.12±13621.65 32373 32373 1.17±133.70 0 23247
Vienna 1092.27±1836.44 120 4636 5.97±312.28 0 40771
t=1
t=2
t=3
c) Religion (Vienna)
b) Death anxiety (Manchester)
a) Emotional pain (Christchurch)
Figure 2: Examples of daily message exchange networks that convey emotional pain, death anxiety, and religion sent during
the first three days of the Christchurch terror attack, Manchester terror attack, and Vienna terror attack, respectively.
et al., 2014). While the majority of previous stud-
ies focused on the informative role of Twitter during
terror attacks, we focused on the personal emotional
responses to the terror attack. In contrast to our prior
assumption that death anxiety and emotional pain will
be dominant in the acute phase of the terror attack, our
data points to an immediate surge in messages that
convey positivity. To explain this observation, we re-
fer to (Steensen, 2018) who showed that the messages
of prayer emerge already a few hours after the news of
a terror attack have reached the audience. These mes-
sages often contain a #prayfor hashtag (followed by a
name of a country or a city affected by a terror attack,
e.g., #prayforparis). This clearly points to a human
tendency to show resilience against the otherwise dis-
ruptive effects of a terror attack and engage in a pro-
social behavior while expressing empathy, hope, and
well-wishes.
Dynamics of Personal Responses to Terror Attacks: A Temporal Network Analysis Perspective
41
Time Time Time
Time Time Time
Time Time Time
Edge formation countEdge formation countEdge formation count
Edge formation countEdge formation countEdge formation count
Edge formation countEdge formation countEdge formation count
EMOTIONAL PAIN DEATH ANXIETY RELIGION
Christchurch
Christchurch
Christchurch
EMOTIONAL PAIN DEATH ANXIETY RELIGION
Manchester
Manchester
Manchester
EMOTIONAL PAIN DEATH ANXIETY RELIGION
Vienna
Vienna
Vienna
Figure 3: Edge formation.
Dyad duration density
Networks
Emotional pain (Christchurch)
Figure 4: Dyad duration. The networks are grouped according to the event of study (Vienna terror attack in orange, Manchester
terror attack in blue, Christchurch terror attack in green alongside x-axis).
These findings are in line with the Terror Man-
agement Theory’s basic premise, which points to a
human tendency to focus on a positive worldview as
a shield against the fear of death (Greenberg et al.,
1986). We examined a 3% random sample of tweets
(excluding retweets) for each personal response and
each terror attack
4
to get further insights into the
4
In total, we manually inspected 261 messages related
COMPLEXIS 2022 - 7th International Conference on Complexity, Future Information Systems and Risk
42
content associated with emotional pain, death anxi-
ety, and positivity associated with the use of religious
words.
The emotional pain category predominantly in-
cluded messages of condolences (92%). This cat-
egory also conveyed a smaller amount of messages
of intense anger (2.5%), e.g., “I’m BREWING OVER
with anger at THIS evil loser. #ManchesterArena”,
“What a cruel and evil cu*t this Brenton Tarrant
is hope you rot in hell you vile evil cunt”, “Jesus
fu**ing christ, I’m so unbearably fu**ing sad and
angry.”, I hope the perpetrators are caught alive,
these deranged scum disgust me. #PrayforVienna
#fu*kterrorism” and annoyance (1%), e.g., “I think by
equating #Brexit and #Christchurch attack you have
reached a new low and “Anyone who sends private
videos of terror, suffering and death to [magazine
name] to grab a few euros should be struck by light-
ning”.
To a lesser amount (1.5%), values such as the role
of “love” to “silence hate” or describing terror as a
“cowardly act” that destroys lives and “violates our
human values”, were also a theme of the messages in
the emotional pain category. Such messages that con-
vey values indirectly also communicate a messages of
togetherness. (Ezzati, 2021) showed that values serve
as means to unite those affected by the terror attack.
While the purpose of such messages is to dissociate
religious groups from the perpetrator, we found that
our examples do not explicitly refer to any specific
religious group or ethnicity but rather generally refer
to human values as means for hoping for a better fu-
ture “War may be loud and strong - I hope we can
be louder and stronger. For the peace. Together - no
matter who and from where.
However, distancing oneself from the perpetrators
was conveyed in responses to islamophobic tweets.
We label such tweets as messages of counter-bigotry
(1%), e.g., “As a Muslim I feel double pain, (1) for the
lose of innocent people, (2) coz terrorists kill using the
name of Islam. #ManchesterArena”.
In total, 2% of the messages in the emotional pain
category were misclassified, for example: “#Manch-
esterStrong I would hope that politicians have re-
spect enough to not use the tragic events to launch
scatching attacks at each other.”, “The obnoxious
[anonymized] has made a glorifying video about the
Islamist attack in Vienna. I ask everyone to report
the video on YouTube and rate it negatively! #vienna
#terror #islam #attack”.
Our second type of a personal response to ter-
ror attacks, death anxiety, was largely conveyed in
to religion, 101 related to death anxiety, and 299 related to
emotional pain, equally distributed over each terror attack.
messages of condemnation and detestation against the
perpetrators followed by well-wishing to those af-
fected (78%), e.g.:
“Unhuman act of terror(IS)m hits Manchester, my
thoughts and feelings go out to the ones who died
#manchester”
“Another senseless tragedy for a pointless ’war’.
My heart hurts too much to sleep.
This is one of the most disgustingly disturbing
mass shooting footage I’ve ever seen. Sick. Burn-
ing him alive would do him no justice. #pray-
forchristchurch
However, we also found messages that express con-
cern for the lives of the ones directly affected by a
terror attack (9%), e.g.:
I can only think of #Paris #Nice #Kabul #Vienna
all the victims of the #terrorist attack.
I hope that no one else has to die out there
from these terrorists who are still walking around
freely.
Death anxiety was also associated with a reminder of
previous terror attacks (8%), e.g., “I am reminded to-
day of the awful feeling when I learned that a terrorist
had killed dozens of my people at night”, and “Horri-
fying visuals and chilling reminder of Mumbai”. Such
a reference to other acts of violence is commonly
found in various reports on human responses to terror
attacks and, as pointed in (Eriksson, 2016), is used to
put the current terror attack in a context of a grow-
ing trend of violence. Moreover, we also found that
the three terror attacks studied in this paper brought
memories of other terror attacks which further points
to the universal fear of death regardless of one’s tem-
poral, social, or geographical proximity to the event.
In the category “death anxiety”, we also found
instances of islamophobic tweets (3.5%) that convey
anger and convey a lack of tolerance towards specific
religious groups. For example, “When I think of Vi-
enna I see architecture and culture. When Muslims
think of Vienna they see murder and terrorism. Is-
lam is a cancer on Europe. Outlaw it now. Banish
them all. Their teachings tell them to kill us. Their
books guide them into terror #islamiscancer”, “One
after another terror attacks r happening in world I
don’t know when will dis sought of mentality will end
these radicalised islamics r still living in previous
centuries they r not getting that dis is 21st century
u can’t get power by killing people #ViennaTerrorAt-
tack”. Islamophobic tweets were shown to be a com-
mon narrative in the aftermath of terror attacks, espe-
cially when the perpetrator(s) were identified as mem-
bers of the Islamic religion (see, e.g., (Awan, 2014;
Dynamics of Personal Responses to Terror Attacks: A Temporal Network Analysis Perspective
43
Hardy and Miller, 2022)). Islamophobia is seen as a
prejudice against Muslims with a goal to cause hostil-
ity and provoke feelings of intolerance (Awan, 2014).
As pointed out in (Hardy and Miller, 2022), islamo-
phobic tweets can be regarded as a form of blaming,
which commonly emerges in the aftermath of terror
attacks and serves as a mechanism to regain a psy-
chological control over the highly emotional situation
(Flynn, 1997).
The remaining messages (1.5%) conveyed disbe-
lief and a shift in one’s worldview. Messages such
as “I never thought that this would happen. Vienna
is one of the best places in the world. I always felt
so good there - far away from all the war and ter-
ror. Now it’s over. Wtf is wrong with this world.
are in line with the goal of terrorists to destabilize our
worldview (see (Miller and Landau, 2005)) and an ex-
pected reaction to the terror attack by those affected.
Other tweets in the sub-category of “death anxi-
ety” included feelings of pain as a result of hearing
about people dying (“each massacre and terrorist at-
tack kills a tiny piece of me... maybe all of us?... no
words are adequate., “I cannot begin to image how
one deals with ones child being killed by a terrorist.
#ManchesterBombing #ManchesterArena”), as well
as glorification of human values (“Whoever kills an
innocent life it is as if he has killed all of human-
ity..”), and a critique to human behavior during the
acute phase of the terror attack (“With all the videos
of the #terrorist attack in #Vienna, I wonder how peo-
ple manage to keep their smartphones steady while
people are being shot in front of their eyes.).
We found that a great majority of tweets re-
lated to positivity were messages that convey prayers
that express well-wishing and sympathies (86.59%),
e.g, “I’m praying for everyone in Manchester!
#staystrongmanchester”, “All my prayers and sym-
pathies are with those who have lost their precious
lives and are in the hospitals. #newzealandterroris-
tattack”.
Some of the prayers conveyed calls for peace, to-
getherness, and unity, e.g.:
I pray for the victims and their families. Enough
violence! Let us together strengthen peace and
fraternity. Only love can silence hate. #viennater-
rorattack”,
“Whether you agree with people religious beliefs
or not, we should love them equally. My heart
goes out to the victims’ families. #christchurch”,
“Murdering people - and supposedly in the name
of God: unbearable. Terrorism knows no reli-
gion. Together we are stronger than hate and ter-
ror. #viennaATTACK”.
These findings are in line with (Steensen, 2018), who
found that the messages of empathy and love domi-
nated a discourse even two weeks since the 2011 Nor-
way terror attack occurred.
Among the remaining 13.41% tweets in this cat-
egory, some instances called for (symbolic) action,
e.g., “Let our memory of the victims remind us of
the dignity of every human being, which we will de-
fend together. I invite you to attend a nationwide
minute of silence at 12 p.m. #viennaterrorattack”,
while some tweets glorified diversity and acceptance
towards those of a different culture or religion, e.g.,
Edinburgh Central Mosque. Crowds gathering in
solidarity with #christchurch. So glad I came. Such
diversity here.”, “I hugged my neighbors this morn-
ing! They are Muslim, and the best family I’ve ever
met!”.
Other messages expressed counter-bigotry, e.g.:
“I’m beyond proud of my lil sis and her friends
at their first protest against islamophobia! The
youth are so powerful.
I will forever fight against antisemitism, any
kinda of segregation isn’t okay, we all should be
against hate crime. violence is never the answer,
today i sadly came across some ppl being ex-
tremely racist and islamophobic under the #Vien-
naAttack hashtag
Approximately 3% of tweets were misclassified. For
example, messages such as I hope you rot in hell”,
No innocent person deserves to die from a flawed
ideology.”, and You going to the gig to enjoy yourself
and create beautiful memories. Not to die. were clas-
sified as messages that predominantly convey positiv-
ity, but instead they are of a predominantly negative
affect.
Limitations. Even though we conducted a focused
data extraction and monitored Twitter to select appro-
priate key terms and hashtags to fetch tweets, we can-
not exclude the possibility that we missed some rele-
vant tweets. Moreover, social bots and their role in so-
cial networks is still of a great concern among the re-
searchers. While our data-set on the Vienna terror at-
tack excludes bots (based on Botometer score 0.8),
we did not process the remaining two terror attacks
data-sets to identify and exclude social bots due to the
large number of screennames. We acknowledge this
as a limitation as our results might be affected by the
influence of bots to a certain degree. Finally, our com-
putational method to detect death anxiety, emotional
pain, and positive messages associated with the use of
religious words is solely based on LIWC and its inter-
nal lexicon. Though we manually inspected a ran-
dom sample of tweets and their corresponding scores
COMPLEXIS 2022 - 7th International Conference on Complexity, Future Information Systems and Risk
44
assigned by LIWC during the pre-processing phase
of our research method, we still identified tweets that
were misclassified.
6 CONCLUSION
In this paper, we analyzed a data-set consisting over
3.8 million tweets related to three terror attacks (the
2017 Manchester terror attack, the 2019 Christchurch
terror attack, and the 2020 Vienna terror attack). Our
goal was to examine the temporal evolution of three
types of personal responses to the terror attacks
death anxiety, emotional pain, and words of positivity
associated with the use of religious words. To ana-
lyze our tweets and detect the level of emotional pain,
fear of death, and positivity, we used the Linguistic
Inquiry and Word Count (LIWC) tool for English and
German language.
Our findings clearly indicate the prominent role
of positive religion-related messages that convey em-
pathy and well-wishing. We showed that not only
do such messages dominate over the messages of
death anxiety and emotional pain, but they are dis-
seminated with a higher frequency over Twitter (via
retweeting), and discussed in a substantially higher
rate than their counterparts. Given such a differ-
ence in the underlying communication networks, we
subsequently examined the measures of momentary
structure and showed that a duration of dyads in the
religion/positivity-related networks is comparatively
higher than the remaining networks in all terror at-
tacks that we analyzed.
In our future work we plan to further investi-
gate the temporal dynamics of specific narratives that
emerge during different types of crisis events. More-
over, in our previous studies we showed that the net-
works resulting from the exchange of messages that
convey basic emotions, exhibited substantial differ-
ences in their underlying structures. Along these
lines, it would be interesting to expand our work to
the underlying structure and a dynamic evolution of
narratives.
REFERENCES
Awan, I. (2014). Islamophobia and Twitter: A Typology of
Online Hate Against Muslims on Social Media. Policy
& Internet, 6(2):133–150.
Becker, E. (1997). The denial of death. Simon and Schuster.
Bender-deMoll, S. and Morris, M. (2021). tsna: Tools for
Temporal Social Network Analysis. R package version
0.3.5.
Bruns, A. and Hanusch, F. (2017). Conflict imagery in a
connective environment: audiovisual content on twit-
ter following the 2015/2016 terror attacks in paris and
brussels. Media, Culture & Society, 39(8):1122–1141.
Burnap, P., Williams, M., Sloan, L., Rana, O., Housley,
W., Edwards, A., Knight, V., Procter, R., and Voss,
A. (2014). Tweeting the terror: modelling the social
media reaction to the Woolwich terrorist attack. Social
Networks Analysis and Mining, 4(206).
Eriksson, M. (2016). Managing collective trauma on so-
cial media: the role of twitter after the 2011 norway
attacks. Media, Culture & Society, 38(3):365–380.
Ezzati, R. T. (2021). ‘United through our values’? Express-
ing unity through value-talk after terrorism in France
and Norway. Migration Studies, 9(3):852–871.
Flynn, B. (1997). Psyhological aspects of disasters. Renal
failure, 19(5):611–620.
Greenberg, J., Pyszczynski, T., and Solomon, S. (1986).
The causes and consequences of a need for self-
esteem: A terror management theory. In Baumeis-
ter, R., editor, Public and private self, pages 189–212.
Springer-Verlag, New York.
Hardy, M. M. and Miller, B. M. (2022). Memorializ-
ing Tragedy on Twitter: Analyzing #PRAYFOROR-
LANDO Following the 2016 Pulse Night Club Shoot-
ing. Communication Studies, 0(0):1–15.
Hermans, H. J. M. and Hermans-Jansen, E. (1995). Self-
narratives: The construction of meaning in psy-
chotherapy. Guilford Press.
Junghaenel, D. U., Schneider, S., and Broderick, J. E.
(2017). Linguistic indicators of pain catastrophizing
in patients with chronic musculoskeletal pain. The
journal of pain, (5):597–604.
Kastenbaum, R. (2007). Death anxiety*. In Fink, G., ed-
itor, Encyclopedia of Stress (Second Edition), pages
717–722. Academic Press, New York, second edition
edition.
Ku
ˇ
sen, E. and Strembeck, M. (2021). Building blocks of
communication networks in times of crises: Emotion-
exchange motifs. Computers in Human Behavior,
123:106883.
Kwon, K. H., Chadha, M., and Pellizzaro, K. (2017).
Proximity and Terrorism News in Social Media: A
Construal-Level Theoretical Approach to Networked
Framing of Terrorism in Twitter. Mass Communica-
tion and Society, 20(6):869–894.
Mair, D. (2017). #Westgate: A Case Study: How al-
Shabaab used Twitter during an Ongoing Attack.
Studies in Conflict & Terrorism, 40(1):24–43.
Merrill, S., Sumartojo, S., Closs Stephens, A., and Cow-
ard, M. (2020). Togetherness after terror: The more
or less digital commemorative public atmospheres
of the manchester arena bombing’s first anniversary.
Environment and Planning D: Society and Space,
38(3):546–566.
Miller, C. and Landau, M. (2005). Communication and
Terrorism: A Terror Management Theory Perspective.
Communication Research Reports, 22(1):79–88.
Oh, O., Agrawal, M., and Rao, H. (2011). Information con-
trol and terrorism: Tracking the Mumbai terrorist at-
Dynamics of Personal Responses to Terror Attacks: A Temporal Network Analysis Perspective
45
tack through Twitter. Information Systems Frontiers,
13:33–43.
Pennebaker, J., Booth, R., Boyd, R., and Francis, M.
(2015). Linguistic Inquiry and Word Count: LIWC-
2015. Austin, TX: Pennebaker Conglomerates.
Reilly, P. and Vicari, S. (2021). Organizational Hashtags
During Times of Crisis: Analyzing the Broadcasting
and Gatekeeping Dynamics of #PorteOuverte During
the November 2015 Paris Terror Attacks. Social Me-
dia + Society, 7(1):2056305121995788.
Rosenblatt, A., Greenberg, J., Solomon, S., Pyszczynski, T.,
and Lyon, D. (1989). Evidence for terror management
theory: I. the effects of mortality salience on reactions
to those who violate or uphold cultural values. Journal
of Personality and Social Psychology, (4):681––690.
Rude, S., Gortner, E.-M., and Pennebaker, J. (2004). Lan-
guage use of depressed and depression-vulnerable col-
lege students. Cognition and Emotion, 18(8):1121–
1133.
Simon, T., Goldberg, A., Aharonson-Daniel, L., Leykin, D.,
and Adini, B. (2014). Twitter in the cross fire—the
use of social media in the westgate mall terror attack
in kenya. PLOS ONE, 9(8):1–11.
Smelser, N. and Mitchell, F., editors (2002). Terrorism:
Perspectives form the Behavioral and Social Sciences.
Panel on Behavioral, Social, and Institutional Issues,
Committee on Science and Technology for Countering
Terrorism. The national academies Press, Washington
D.C.
Steensen, S. (2018). Tweeting Terror: An Analysis of the
Norwegian Twitter-sphere during and in the Aftermath
of the 22 July 2011 Terrorist Attack. Emerald Publish-
ing Limited, Bingley.
Tausczik, Y. and Pennebaker, J. (2010). The psychological
meaning of words: Liwc and computerized text anal-
ysis methods. Journal of Language and Social Psy-
chology, 29(1):24–54.
Tikka, M. M. T. (2019). Ritualisation of Crisis Communica-
tion : Crowd-enabled responses to the Stockholm ter-
ror attack on Twitter. NORDICOM Review, 40:105–
120.
COMPLEXIS 2022 - 7th International Conference on Complexity, Future Information Systems and Risk
46