You are on page 1of 21

Journal of Contemporary Religion

ISSN: 1353-7903 (Print) 1469-9419 (Online) Journal homepage: https://www.tandfonline.com/loi/cjcr20

The sacred and the profane: social media and


temporal patterns of religiosity in the United Arab
Emirates

Justin Thomas, Aamna Al Shehhi & Ian Grey

To cite this article: Justin Thomas, Aamna Al Shehhi & Ian Grey (2019) The sacred and the
profane: social media and temporal patterns of religiosity in the United Arab Emirates, Journal of
Contemporary Religion, 34:3, 489-508, DOI: 10.1080/13537903.2019.1658937

To link to this article: https://doi.org/10.1080/13537903.2019.1658937

Published online: 02 Oct 2019.

Submit your article to this journal

Article views: 298

View related articles

View Crossmark data

Citing articles: 4 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=cjcr20
JOURNAL OF CONTEMPORARY RELIGION
2019, VOL. 34, NO. 3, 489–508
https://doi.org/10.1080/13537903.2019.1658937

The sacred and the profane: social media and temporal


patterns of religiosity in the United Arab Emirates
Justin Thomas, Aamna Al Shehhi, and Ian Grey

ABSTRACT ARTICLE HISTORY


Large datasets associated with internet search engines and Received 25 May 2017
social media platforms are increasingly used to study Accepted 10 June 2018
psychological variables. Over the past decade, ‘big data’, as KEYWORDS
they have become known, have become central to the Religion; social media;
exploration of a diverse range of topics. Few studies, however, Twitter; Islam; profanity
have examined religiosity (religious belief, commitment, and
devotion), particularly Islamic religiosity in the Arab world. This
study looked at religiosity in the United Arab Emirates through
data extracted from Twitter, a popular social media platform.
The data comprised 152 million Twitter messages, spanning the
period 1 April–30 September 2016. Bilingual search algorithms
were employed to investigate the temporal patterns of
religiosity expressed within the dataset. The study also
explored patterns in the expression of obscenity (offensive
language), hypothesising a negative relationship with religious
sentiment. Religiosity followed hypothesised temporal patterns
and was also inversely correlated with obscenity. There were
differences observed between languages (Arabic vs English)
and gender, with males, surprisingly, expressing greater
religiosity than females. This research contributes to the
nascent study of religiosity through social media.

Introduction
The latter half of the twentieth century saw the emergence of a new
techno-economic paradigm and the proliferation of digital devices (e.g.
personal computers, smart phones) across all walks of life (Freeman and
Louçã 2002). One consequence of these developments is the rise of
increasingly large datasets, characterised by exponential growth. These
datasets have been referred to as ‘big data’, a term commonly elaborated
with reference to the ‘three Vs’: volume, velocity, and variety (Eaton et al.
2012). ‘Volume’ refers to the extraordinary size of these datasets, while
‘velocity’ refers to their rapid exponential growth, with ‘variety’ connoting
the diversity of data types being stored (e.g. images, audio, video). More
recent commentators have also appended a fourth ‘V’, referring to value
(Cehn et al. 2014)—the idea that commercially and societally valuable

CONTACT Justin Thomas justin.thomas@zu.ac.ae


© 2019 Informa UK Limited, trading as Taylor & Francis Group
490 J. THOMAS ET AL.

information can potentially be extracted from these datasets (Fuller


2015, 569).
One facet of the value offered by big data is the opportunity for
researchers to explore a wide variety of phenomena at the macro-level.
The emerging discipline of data science has enabled and supported this
endeavour, offering new tools and techniques to interrogate and extract
meaning from big datasets. The idea that we can derive actionable
intelligence and find valid and reliable answers to research questions
from these big datasets has, over the past decade, received fairly
unequivocal empirical support. For instance, data relating to the
frequency and geography of internet search terms have been used to
improve and support traditional methods of syndromic surveillance.
Using Google web search logs, researchers were able to develop models
that could rapidly and accurately predict the percentage of influenza-like
illness within a given locality, providing the ability to detect influenza
epidemics weeks in advance of traditional surveillance methods
(Ginsberg et al. 2009). Other studies have explored psychological
constructs, such as happiness or subjective well-being (Dodds et al. 2011;
Yang and Srinivasan 2016). Studies of social media, most notably Twitter,
have focused on attempts to quantify the affective aspects of subjective
well-being (positive and negative sentiment) at a macro-level. Notably,
Peter Dodds et al. (2011) developed a big data search algorithm known
as hedonometer. This algorithm, subsequently used by other researchers,
aimed to quantify societal sentiment by calculating the relative frequency
with which 10,000 keywords, previously ranked (1 to 9) for positive and
negative affective valence, were used across time. Shifts in patterns of
positive and negative word usage are viewed as being indicative of
societal sentiment or public mood. Applied to big datasets, such as
Twitter, hedonometer-derived analysis is fairly consistent with intuition.
For example, weekends are happier than weekdays and shifts in societal
sentiment predictably correspond with widely reported positive and
negative events, for example, royal weddings or the deaths of popular
entertainers.
Beyond transient states such as happiness and influenza, Chao Yang and
Padmini Srinivasan (2016) used an algorithmic analysis of Twitter data to
explore life satisfaction, a trait-like component of subjective well-being. As
hypothesised, estimates of life satisfaction were relatively stable across time
and appeared to be uninfluenced by external events, such as seasonal
changes, the deaths of celebrities or political crises. Furthermore, Twitter
users who had been algorithmically categorised as ‘satisfied’, as opposed to
‘dissatisfied’, demonstrated patterns in their messages (tweets) that were
consistent with intuition and previous findings in the subjective well-being
literature. For example, the ‘satisfied’ expressed significantly more positive
JOURNAL OF CONTEMPORARY RELIGION 491

and less negative emotion. They also used less obscenity and were
significantly more positive about religion than their ‘dissatisfied’
counterparts. Another methodologically similar study—similar in that it
aimed to categorise and compare two groups—used a socio-linguistic
analysis of tweets posted by individuals with ostensibly differing political
affiliations (Sylwester and Purver 2015). Again, intuitively consistent
patterns emerged such as Republicans (conservatives) mentioning
religion more frequently than Democrats (liberals) and the reverse
pattern for the frequency of the use of swear words (obscenity/profanity).
Beyond individual states, traits, and preferences, cultural values at the
national level have also been explored using Twitter data. Ruth Garcia-
Gavilanes, Daniele Quercia, and Alejandro Jaimes (2013) found that the
act of mentioning other people in one’s tweets was negatively correlated
with a nation’s individualism score (nations had previously been assigned
scores for various cultural dimensions). Similarly, Twitter users from
individualistic countries socialised with strangers more frequently, while
users from collectivist countries demonstrated a preference for
strengthening existing social ties. These findings converge with previous
research and observations, suggesting that collectivistic societies tend to
place greater value on interdependence and strong social connections
among in-group members (Hofstede 2001).
Another aspect of behaviour that has been viewed through the lens of
big data is the use of profanity (swear words). This is particularly relevant
to the present study, as swear words, especially obscene language, might be
considered antithetical to religious sentiment. Exploring swear words in
English, using a lexicon containing 788 English expletives (e.g. f***, sh**,
ass), Wenbo Wang et al. (2014) found at least one swear word in 7.73% of
the 15 million tweets within their dataset. Inexplicably, swear-word usage
appeared to peak on Wednesday, decline on the weekend, and began its
ascent again on Sunday. Exploring gender, the authors found that males
used swear words more than females, especially in male-to-male Twitter
interactions. All these patterns closely echo the findings of studies
exploring the use of swear words in offline contexts (Mehl and
Pennebaker 2013).
Despite the breadth of focus, very few big data studies have directly
explored religiosity. We can find no studies that have examined Islamic
religiosity within the Arabian Gulf context. Religiosity or religiousness has
been defined as “an individual’s conviction, devotion, and veneration
towards a divinity” (Gallagher and Tierney 2013, 1653). One big data
study that did attempt to examine an aspect of Christian religiosity,
however, quantified the number of biblical passages posted on Twitter in
2016 (40 million). The same study also ranked the top five most frequently
posted biblical passages (Lee 2016). Another study, loosely focused on
492 J. THOMAS ET AL.

Christianity, used sentiment analysis (the hedonometer) to compare the


subjective well-being of Christians and Atheists among 16,000 Twitter
users across 2 million tweets (Ritter, Preston, and Hernandez 2013).
Belief status, Christian or Atheist, was assigned by followership. On
Twitter, one can follow the accounts of other Twitter users and thereby
receive each message/comment that these individuals post. Twitter users
following the accounts of leading public figures espousing either
Christianity (e.g. Pope Benedict XVI) or Atheism (e.g. Richard Dawkins)
were assigned to the respective belief group. This type of heuristic-based
assignment is deemed acceptable, as the sample size in such studies
protects genuine effects from being easily obscured by sampling error
and other minor extraneous factors (‘noise’). For example, not every
person who follows the Pope on Twitter will be religiously inclined, but
it is safe to assume that the majority will be. Concordant with previous
research exploring subjective well-being and religiosity (Abdel-Khalek
2008; Abdel-Khalek and Eid 2011), the religious were ‘happier’ than the
non-religious.
The relative scarcity of big data studies examining religiosity, however,
suggests a gap in the literature and a need for more studies focused on this
aspect of online human behaviour. Additionally, there is a growing interest
in the impact of the emergence of social media on religion and vice versa.
This interest has been reflected in the launch of new journals such the
Journal of Religions on the Internet as well as more broadly focused
journals publishing special issues about religion in the information age
(Campbell and Connelly 2015). Furthermore, in the context of the Arabian
peninsula, the birthplace of Islam, religious practice remains a widespread
aspect of daily life (Thomas 2014, 5). Exploring religiosity in this context
might be particularly informative, given the idea that Islamic religiosity
remains tightly intertwined with national and cultural identity (Abd-Allah
2004). Over time, increases or decreases in the online expression of
religious sentiment are likely to be important indicators of socio-cultural
change.
The present study uses Twitter data to explore religiosity in the context
of the United Arab Emirates (UAE), a country of approximately 9 million
people, located on the east coast of the Arabian Peninsula. The official
religion of the UAE is Islam and the official language is Arabic (Thomas
2014). The UAE was founded in 1971, shortly after the commercial
exploitation of the country’s large oil and gas reserves had begun. Since
then, according to the World Health Organization (WHO 2006, 289), few
nations on earth have witnessed such rapid socio-economic and
technological development. Before the federation of the UAE formed in
the late 1960s, it was difficult to find clean drinking water and electricity;
today, the UAE boasts the world’s tallest building, largest shopping mall,
JOURNAL OF CONTEMPORARY RELIGION 493

and state of the art museums, universities, and hospitals (Thomas 2014,
27). This rapid development has necessitated a large expatriate workforce.
Data from UAE’s National Statistics Bureau suggest that Emiratis (UAE
citizens) comprise 11.38% of the population, resulting in the UAE having
one of the highest expatriates to citizen ratios in the world (NBS 2009).
The UAE has modernised its educational system, with most tertiary
educational institutions now using English as the language of tuition. In
the early days of the UAE, only a handful of citizens received college
education; today the vast majority engage in tertiary education either in
the UAE or abroad (Davidson 2008). Such factors—population
demography and education—are viewed as having an acculturating
influence on the Emirati populace and there are concerns about the loss
of language (Arabic), traditional values, and the disintegration of cultural
identity (Fox, Mourtada-Sabbah, and Al-Mutawa 2006). Nada Mourtada-
Sabbah et al. (2008) argue that, in the context of the UAE, factors such as
demographic imbalance, education, and new media represent a broad
socialising matrix, instilling the values of consumerism and individualism
at the expense of the traditional collectivist values rooted in the Islamic
heritage. Similarly, several studies exploring Western acculturation and
mental health among young Emiratis suggest that this process is
associated with an elevated risk of mental health problems (Thomas
et al. 2016; Thomas, Quadflieg, and O’Hara 2016).
In the future, using a wider time frame, the quantitative examination of
religiosity might provide a useful indicator for the speed and direction of
the posited social changes. Being aware of such changes and able to
quantify them opens up the possibility of being able to enact and
evaluate remedial initiatives. Similarly, exploring levels of obscenity,
specifically, instances of hate speech, can also provide government
agencies with an indicator of how (in)tolerant society is or is becoming.
Over narrower time frames, looking at Islamic religiosity can provide
insights into the temporal patterns of religiosity, identifying the types of
events that strengthen or attenuate religiosity’s online expression.
Big data, particularly Twitter, have proven to be useful sources for
examining online behaviour and making inferences about offline
attitudes and activities. As indicated, the patterns identified, and their
subsequent interpretation, are frequently consistent with intuition and
regularly converge with findings obtained using more traditional research
methods (Ginsberg et al. 2009; Sylwester and Purver 2015; Wang et al.
2014; Yang and Srinivasan 2016). The present study explores Twitter data
for the UAE, spanning six months and comprising 152 million tweets.
These data were algorithmically searched to examine the relative frequency
of use of concrete nouns that are unambiguously related to devotional
aspects of everyday Islamic practice. The output of this search is taken as
494 J. THOMAS ET AL.

an imprecise indicator of praxis-orientated religiosity or religious


sentiment. Relatedly, the study also looks at a brief lexicon of obscenity.
Obscenity is typically defined as offensive or indecent language “abhorrent
to morality or virtue: designed to incite lust or depravity” (Jay 1992, 5). In
Arabic, as in English, obscenities tend to be of a sexual nature, frequently
associated with sex acts and genitalia. In the context of Islam, as in other
religious traditions, the use of such offensive language is discouraged.

Hypotheses
The present study hypothesises a negative relationship between obscenity
and religiosity. Furthermore, as both religiosity and obscenity are observed
across a six-month time frame, it is hypothesised that occasions of
religious significance, such as Fridays and Ramadan (Islam’s holiest day
and month), will be associated with increases and decreases in religiosity
and obscenity, respectively. All data are examined using English and
Arabic and comparisons between the patterns observed in both
languages are reported. Language comparisons very roughly reflect two
of the main demographic groupings in the UAE: indigenous UAE citizens
and the diverse expatriate workforce.

Method
Dataset description
Twitter Streaming API was used to collect tweets from 1 April to
30 September 2016. The Streaming API provides a real-time random
sample from the complete Twitter corpus. However, the present dataset
was intentionally limited to tweets sent by users located in the UAE, that is,
users with an account summary explicitly mentioning the UAE or any of
its cities, which typically excludes tourists. The additional metadata
associated with each tweet also included, for example, geolocation (GPRS
co-ordinates), language, Twitter name, unique ID. In total, the dataset
comprised 152,835,727 million tweets sent by 531,413 unique users (see
Table 1).
It is noteworthy that Arabic users accounted for more tweets than
English users, despite Arabic users making up just 16% of the sample.
This pattern might be related to the demography of the UAE, where the
Arabic-speaking citizens are a minority, but a relatively young minority.
UAE citizens comprise around 12% of the population, with a younger
median age than the expatriate population (NBS 2009). The discrepancy in
tweet frequency by language might also reflect cultural differences
(individualism/collectivism) in internet usage (see Würtz 2005) or
JOURNAL OF CONTEMPORARY RELIGION 495

Table 1. Breakdown of tweets by language and unique users.


Tweets Tweets (%) Unique Users Users %
Arabic 82,065,108 53.69 88,280 16.61
English 42,235,907 27.63 404,689 76.15
Other* 28,534,712 18.67 38,444 7.23
*There were 62 other languages recorded. After Arabic and English, Urdu, Tagalog, and Hindi were
the most commonly used.

cultural differences in communication style, high vs low context


communication (Hall 2000). Exploring these differences, however, is
beyond the scope of the present study.

Dataset validity checks


Validity checks based on common-sense expectations were undertaken
(Dodds et al. 2011, 3). This involved plotting the hourly frequency (three-
hour blocks) of the English and Arabic terms for ‘morning’ and ‘evening’.
Peak occurrences of these words were, as expected, noted in the morning
and evening, respectively (see Figure 1).
The peak frequency of the use of the word ‘morning’ in English and
Arabic is between 6 am and 9 am, while the peak time for ‘evening’ is
between 6 pm and 9 pm. It is worth noting that in Arabic, ‫( مساء‬evening)
refers to any time after 12 pm.

Lexicon of religiosity
As indicated, the term ‘religiosity’ captures the idea of an individual’s belief in,
and devotion and veneration of, a divinity (Gallagher and Tierney 2013). For
ease of implementation, the present lexicon emphasises Islamic devotional
praxis as opposed to belief. For this reason, words that are not directly related

Figure 1. Graphs showing the mean hourly frequencies for the terms ‘morning’ and ‘evening’
in Arabic and English.
496 J. THOMAS ET AL.

to observable routine religious practices were excluded. Suitable items for the
lexicon were derived from the Islamic glossary provided by the Ahlul Bayt
Digital Islamic Library Project (Kassamali 2016). This work contains an
extensive glossary of terms directly relevant to the practice of Islam. To
improve the lexicon’s sensitivity (increase the number of true positives), all
polysemic words were omitted (for example, the word for ‘Imam/religious
leader’, in written form, is indistinguishable from the locative adverb ‘in front
of’). It was also decided to exclude all proper nouns (Mohammed, Mecca, Ali).
These names all remain in common usage and would very often fail to convey
any religious sentiment (e.g. Fatima Mohammed flew to Mecca to see her Uncle
Ali). Similarly, the word ‘Allah’ was left out of the lexicon, as this term is
frequently uttered without religious sentiment in the context of formulaic
greetings and curses; it is also frequently employed within proper names, e.g.
Abd Allah (‫)عبد الله‬. In addition to excluding proper nouns, it was also decided
to omit abstract nouns related to spiritual/unseen or legalistic dimensions of
Islam (fiqh—jurisprudence, aqeeda—theology, barakaat—divine blessings).
Focusing on concrete nouns associated with observable devotional practices
ensured that the lexicon was praxis orientated. Finally, to avoid obvious
seasonal effects, the lexicon focused on perennial, rather than seasonal acts of
devotion, with fasting, Ramadan, and haj (the greater pilgrimage), being mostly
seasonal acts, being excluded (see Table 2).
Common transliterations of these words made up the English version of
the religiosity lexicon (e.g. Mosque/masjid, Quran/Koran, wudo/woodu).

Lexicon of obscenity
Obscenities are words and phrases that, by social consensus, are considered
extremely indecent and offensive. Obscenity is described, as mentioned earlier,
as language that is against morality and virtue, insightful of lust or depravity (Jay
1992, 1). In English and Arabic, obscene words are typically of a sexual nature,
frequently referring to genitalia and sex acts. In the present study, Arabic
obscenities were collected by reviewing several blogs dedicated to swear words
in Arabic (Jesa 2015; El-Ghanim 2015). A group of five Emirati Arabic-speaking
colleagues were asked to rate these candidate or selected words independently of

Table 2. Words included in the lexicon of Islamic religiosity.


Arabic Religious Translation/Interpretation
‫ﻣﺴﺠﺪﻣﺴﻴﺪ‬ Mosque: a place where prayer takes place. The Emirati dialect of Arabic has
a unique spelling for this word, which was also included.
‫ﺍﻟﻘﺮﺁﻥ‬ The Qur’an: the holy book recited during prayer
‫ﺩﻋﺎﺀ‬ Supplication: a prayer directed straight to God
‫ﺻﺪﻗﺔ‬ Charity: typically alms given to the deserving poor
‫ﻭﺿﻮﺀ‬ Ablution: a prerequisite state of ritual purity prior to formal prayer
‫ﺻﻼﺓ‬ Formal prayers: involving prescribed movements and words
‫ﺗﺴﺒﻴﺢ‬ Prayer beads: these are often used in a form of devotional remembrance
JOURNAL OF CONTEMPORARY RELIGION 497

Table 3. Words included in the lexicon of Arabic obscenity.


Arabic Obscenity Translation/Interpretation
‫ﻛﺲ‬ This is a vulgar expression for the word ‘vagina’, ‘pussy’. However, it is typically
accompanied with the phrase ‘your mother’s’ or ‘your sister’s’ and contextually
it equates to ‘f*** you!’ or ‘f*** your mother/sister’.
‫ﺯﺏ‬ This is a vulgar expression for the penis: ‘cock’. Again, this word is frequently
combined with other words to form additional obscene phrases.
‫ﻗﺤﺒﺔ‬ A sexually promiscuous woman: ‘slut’ or ‘whore’
‫ﺷﺮﻣﺘﻪ‬ A sexually promiscuous woman: ‘slut’ or ‘whore’; this is often prefixed with ‘son
of’ to direct the obscene insult at a male.

each other from 1 (not offensive) to 5 (extremely offensive), based on the


perceived magnitude of offensiveness. Words that received a unanimous score
of 5 were included in the lexicon. The final words included in the Twitter dataset
in order of frequency of occurrence are shown in Table 3.
Examples of excluded words include the Arabic equivalents of ‘dog’,
‘donkey’, and s***. These words failed to be unanimously deemed ‘extremely
offensive’. For the English version of the lexicon, we used close cognates of the
Arabic terms (f***, cock, slut, whore). These four words are also among the top
ten of the most frequently used swear words on Twitter in English (Wang et al.
2014). None of the Arabic obscenities was polysemic. The English word ‘cock’ is
polysemic; however, exploring a random sample of 100 tweets revealed that the
obscene meaning is by far the most frequently intended meaning. There was no
desire to lengthen this lexicon of obscenity, as the aim of the study was not to
quantify the absolute frequency of profanity use, but rather to observe diurnal
and temporal patterning based on a sample of frequently used and
unambiguously obscene words.

Results
Descriptive analysis
The Arabic lexicon of religiosity produced more ‘hits’ than the English
equivalent. Conversely, English produced more hits for the lexicon of
obscenity, for example, the word ‘f***’ was recorded in 222,772 tweets, while
the most frequent example of the Arabic profanity ‫ كس‬occurred in 42,195,
a ratio of around 37:7. Table 4 provides details of the relative percentages for
both lexicons in both languages.

Table 4. Counts and percentages of Arabic/English tweets containing religiosity/obscenity.


Language Tweets Religiosity Rel% Obscenity Obs%
Arabic 82,065,108 2,753,593 3.35 111,424 0.13
English 42,235,907 314,861 0.74 249,745 0.59
498 J. THOMAS ET AL.

Table 5. Counts and percentages of Arabic/English tweets by gender.


Gender Tweets Arabic English % Arabic Unique Users
Female (Fatima) 346,600 213,875 87,137 61.70 995
Male (Mohammed) 647,102 528,168 76,176 81.62 3,350

Gender differences
Gender is not a given field in the Twitter dataset; however, based on Arabic
naming conventions, a sub-sample of males and females was extracted. This
was achieved, as indicated, by filtering Twitter users with the display name
Fatima or Mariam (using multiple English and Arabic spellings, e.g. ‘Fa6ma’,
‘Maryooma’). Similarly, the name ‘Mohammed’ was used to provide an
ostensibly male sample. The details of these samples are given in Table 5.
To validate this gender division heuristic, an intuitive hypothesis was
tested, comparing both groups on the frequency with which they used
gender-typed terms. The selected terms were ‘football’/‫ كرة القدم‬and ‘make-
up’/‫تجميل‬. There was a clear and powerful interaction with gender, where
putative females mentioned ‘make-up’ more than males and ‘football’ less
than males. The inverse of this pattern was clearly observed for our putative
males (more mention of ‘football’, less of ‘make-up’).
Upon exploring gender differences with our derived sub-samples (e.g.
Fatima vs Mohammed), several statistically significant differences emerged.
Firstly, males tweeted more than females using Arabic (χ2 [1, N =993702] =
3629.88, p < .001). Similarly, based on Arabic tweets, males also expressed more
religious sentiment than females, with 50.08 and 22.89 instances per 1,000
tweets, respectively (χ2 [1, N =742043] = 278.69, p <.001). The same is true
for males more frequently expressing religious sentiment in English (60.34 per
1,000) than females (5.36 per 1,000). Again, these differences were statistically
significant (χ2 [1, N =162429] = 3699.50, p <.001). This finding is contrary to
expectations. Much previous research suggests that across faiths—Islam
included—religiosity tends to be greater among women (Stark 2002).
Concerning Arabic obscenity, males expressed this at a rate of 0.28 per
1,000 tweets compared to 0.05 for women (χ2 [1, N =742043] = 17.13,
p <.001). The gender pattern for obscenity was reversed when English
tweets were explored. In this condition, females used English obscenities at
a rate of 5.5 per 1,000 tweets, while males produced 3.78 (χ2 [1,
N =163313] = 30.12, p <.001).

Daily patterns
In both languages, religiosity peaked around midday (12–2 pm.). There
was also a relative peak in the early morning (3–5 am.). In both languages,
obscenity appeared to be more frequent in the early hours: 12–6 am, for
JOURNAL OF CONTEMPORARY RELIGION 499

Figure 2. Graphs showing the proportion (per 1,000) of tweets expressing religiosity/obscenity
in Arabic/English throughout the day.

English, and 12–3 am for Arabic. In English, obscenity seemed to be


inversely correlated with religiosity, peaking when religiosity declined
and vice versa (see Figure 2). A correlational analysis confirmed this
relationship, with r (247) = -.60, p <.001. Obscenity use in Arabic was
relatively infrequent and stable across the day.

Weekly patterns

There was a clear increase in religiosity on Friday. In Islam, Friday (yaum al


Jumma) is considered particularly holy, the day of the Friday sermon (khutba)
during the congregational prayer (Jumma). Concerning Arabic obscenity,
there were no obvious peaks or troughs. Similarly, there were no distinctive
weekly patterns for obscenity or religiosity in English (see Figure 3).

Figure 3. Graphs showing the proportion (per 1,000) of tweets expressing religiosity/obscenity
in Arabic/English throughout the week.
500 J. THOMAS ET AL.

Figure 4. Graph showing the proportion (per 1,000) of tweets expressing religiosity/obscenity
in Arabic across a six-month period including Ramadan.

Seasonal patterns
Islam follows a lunar calendar and, for this reason, there is not a fixed
correspondence with the Gregorian calendar. In the graphs we include the
Islamic months with their Gregorian equivalents for 2016. After scrutiny of
these data across the entire six-month period, it was apparent that
Ramadan (Islam’s most holy month) had the expected impact on
religiosity and obscenity (see Figure 4).
Religiosity peaked early in the holy month (Ramadan 1/6 June); the use of
obscenity also decreased and remained relatively low throughout the whole
month. There was also a second religiosity spike occasioned by an obscenity
trough at the beginning of the month of Dhul Hijjah (2 September). Dhul Hijjah
is Islam’s twelfth month and marks the start of the hajj (pilgrimage) season. The
Dhul Hijjah spike in religiosity arguably corresponds to the hajj season, another
significant holy occasion in Islam.
Regarding obscenity in English, the six-month patterns are different.
Concerning Ramadan, the impact of the holy month appears to be far
more short-lived, in that the spike in religiosity and obscenity is limited to
week one of Ramadan (see Figure 5).
In both languages, the transition into the first week of Ramadan is particularly
pronounced, with obscenity decreasing in both English and Arabic and religious
terms increasing in both languages (see Figure 5). Figure 6 shows the mean
occurrence of religiosity/obscenity per 1,000 tweets during the last week of
Shaban—the run-up to the holy month—and the first week of Ramadan.
JOURNAL OF CONTEMPORARY RELIGION 501

Figure 5. Graph showing the proportion (per 1,000) of tweets expressing religiosity/obscenity
in English across a six-month period including Ramadan.

Correlational analysis

Pearson’s correlational coefficient was used to explore the relationship


between obscenity and religiosity in both languages. Across the six-
month dataset, Arabic religiosity was negatively correlated with
obscenity. In general, increases in the expression of Arabic religiosity
were associated with decreases in the use of Arabic obscenity. The
correlations between all variables are reported in Table 6.

Discussion
In a cross-linguistic analysis of 152 million tweets, the proposed indicators of
religiosity and obscenity produced temporal patterns that were mostly

Figure 6. Graphs showing the proportion (per 1,000) of tweets expressing religiosity/obscenity
in English/Arabic for the last week of Shaban (the month preceding Ramadan) and the first
week of Ramadan.
502 J. THOMAS ET AL.

Table 6. Correlations between the proportion of tweets (per 1,000) containing religiosity
and obscenity in English and Arabic.
English Religious Arabic Obscenity English Obscenity
Arabic Religious .29* -.23* -.06
English Religious 21* .22*
Arabic Obscenity .37*
Note: N = 183, representing the number of days spanned by the dataset, * p <. 001

consistent with expectations. Friday (Islam’s holy day) was associated with an
increase in the overall proportion of religiosity-related vocabulary used in
Arabic. Similarly, Ramadan (Islam’s holiest month) was associated with
a substantial increase in religious sentiment (English and Arabic) and
a month-long decrease in the frequency of Arabic obscenity. It is notable
that religiosity peaked very early in Ramadan (during the first week), then
quickly returned to pre-Ramadan levels for much of the month, before rising
again during the last week. In part, this pattern seems analogous to the idea of
the ‘hedonic treadmill’, according to which an event gives rise to positive affect
before a rapid return to a hypothesised hedonic set point (Diener, Lucas, and
Scollon 2006). The dawning of the event—in this case Ramadan—gives rise to
heightened religiosity, which quickly returns to baseline levels (religiosity set
point). The second rise in religiosity, during the last week of Ramadan,
arguably reflects the religious importance of Laylat Al-Qadr (the night of
power). This is believed to be the night when the first verses of the Quran were
revealed to the Prophet Muhammad. Muslims are encouraged to perform
a special night vigil prayer (Qiam ul Lail) at this time, an activity associated
with an abundance of blessings. According to the Quran, “The night of Qadr
is better than a thousand months” (97:3). Nobody knows when Laylat Al-
Qadr is; however, it is agreed that it falls on one of the last ten nights of
Ramadan. Attempting to benefit from Laylat Al-Qadr would explain the
resurgence of religiosity in the Twitter data corresponding with the last
week of Ramadan.
Earlier studies of Twitter datasets have demonstrated similar, although
less pronounced, temporal patterns for happiness, i.e. increased
expressions of positive sentiment at weekends, during public holidays,
and around the time of celebrity weddings (Dodds et al. 2011). The
present study demonstrates that the temporal patterns of religiosity and
obscenity can also be studied using similar big data analytics.
Regarding the relationship between religiosity and obscenity in the
present study, Arabic religiosity was significantly correlated with less use
of obscenity. Similarly, English religiosity (transliterations of the Arabic
terms) was negatively correlated with the use of obscenity on an hourly
basis throughout the day—in other words, peak times for religiosity (12–3
pm) tended to be associated with relatively lower levels of obscenity in
English and vice versa.
JOURNAL OF CONTEMPORARY RELIGION 503

Based on a relatively small sub-set of the data, the study also found
expected and unexpected gender differences. As mentioned, gender was
assumed by Twitter display names: Fatima/Mariam vs Mohammed.
Proportionally, females (Fatimas and Mariams) tweeted in English more
than males (Mohammeds). This might be explained by females
outperforming males in tertiary education in the UAE. Almost all
colleges and universities in the UAE use English as the language of
tuition; previous research has documented greater female participation in
tertiary education as well as higher academic performance (Thomas, Al-
Marzooqi, and Raynor 2012; Davidson 2008). It might be that female
Twitter users in the UAE have greater English language proficiency and
therefore feel more comfortable tweeting in English. However, this
linguistic gender difference could equally be framed as males being more
proficient in Arabic and perhaps less acculturated to the West. At least one
previous study in the UAE has suggested that acculturation to the West is
greater among UAE females (Ghubash et al. 2001, 568). In addition to
tweeting more in Arabic, males also expressed more religiosity, both in
English and Arabic. Their greater public expressivity of religiosity may
reflect traditional gender roles in Islam, where males are encouraged to
pray in the mosque and women are encouraged to pray at home. Similarly,
Islamic public service roles, such as that of community Imam (prayer
leader at a large mosque), have traditionally been restricted to men.
However, past research supports the idea of elevated religiosity among
women (Schnabel 2015; Miller and Stark 2002). Although most of this
research has focused on Christian denominations, there are also a few
studies (the World Values Survey 1995–1997, as quoted in Stark 2002) that
show a similar, although less pronounced, pattern among Muslims (Stark
2002, 499). The present a-typical findings (females expressing less
religiosity than males) can be interpreted in terms of the socialisation
hypothesis of female religiosity, which suggests that social expectations
contribute to females reporting greater religiosity than males (Schnabel
2015). Social media, however, afford people a degree of anonymity and this
might lessen the pressure to conform to social expectations. Yet there are
numerous alternative interpretations. For example, among Muslims,
female social media users might be highly a-typical of the general
Muslim female population of the UAE, while their male counterparts are
more generally representative. Such a situation might arise due to
pervasive gender-role expectations about social interaction. For example,
within some households, it might be argued that a ‘good Muslim woman’
does not speak to, or share information with, strangers on the internet.
Whatever the final explanation for this finding, it certainly merits further
exploration. It would also be worth examining gender differences in
religiosity among other religious groups on social media across different
504 J. THOMAS ET AL.

nations. This finding, in particular, demonstrates that using Twitter data to


explore questions pertinent to contemporary religion is feasible and can
shed new light on old debates. Furthermore, Twitter (social media) data
might also be a useful tool for observing changes in religiosity, gender-
related or otherwise, across time (trends) and in real time (surveillance).
The exploration of swear words across genders, in Arabic, is consistent with
previous research exploring English swear words (Wang et al. 2014): males used
more obscenities than females. However, this pattern was reversed in English
and, proportionally, females used more obscenity than males. The more
frequent use of English obscenities by females might be related to their greater
use of English. In cases where English is the second language, swearing in
English is likely to be far less emotive than swearing in Arabic. Previous
research among bilingual (Turkish/English) speakers found that using
obscenity (e.g. ‘whore’, ‘asshole’, oral sex) in the second language evoked
significantly less autonomic reactivity than it did in the first language (Harris,
Aycicegi, and Gleason 2003). The affective attenuation associated with second
language use might also explain the higher rate of English obscenity across the
whole dataset (bilinguals choosing to swear in English).
The UAE is a highly bilingual nation, with English commonly spoken in
education and at the workplace, while Arabic remains widely spoken at
home by UAE citizens and Arab expatriates (Mourtada-Sabbah et al.
2008). The idea of social change and the loss of cultural identity has
been extensivley discussed in the UAE and broader Arabian Gulf.
Alanoud Alsharekh and Robert Sprinborg argue that “no region of the
globe is so simultaneously localised and globalised, so rooted in unique
traditions and yet so interactive with the world” (Alsharekh and Sprinborg
2008, 13). They further suggest that such a dynamic could result in cultural
disintegration or, as they describe it, “cultural schizophrenia” (ibid).
Social media analysis, such as that undertaken here, might help further
explore the rate and direction of social change in rapidly developing Arabian
Gulf states such as the UAE. These surveillance methods offer obvious
advantages over self-report survey methodologies, in that the data are less
influenced by socially desirable responding and other forms of reactivity
(Ritter, Preston, and Hernandez 2013, 243). Social media use frequently
involves individuals simply conversing with one another on wide-ranging
topics, freely expressing attitudes, fears, and aspirations. In addition to this
naturalistic setting, the sample size that can be obtained through social media
is unparalleled in direct survey work. It is this sample size that silences many
potential criticisms, such as the fact that some tweets are posted by news agencies
or commercial entities. Similarly, some tweets are not a reflection of a person’s
thoughts or emotional state as they might simply be quoting a celebrity or the
lyrics of a song. Big data, however, will typically ensure that such ‘noise’ is
overwhelmingly drowned out by the masses of data that reflect people’s genuine
JOURNAL OF CONTEMPORARY RELIGION 505

concerns and sentiment (Dodds et al. 2011, 6). The patterns observed in the
present study extend this idea to observing religious sentiment among the
population of the UAE.
This study has, however, some important limitations. Its exploratory nature
necessitated the development of a lexicon of Islamic religiosity. This was
relatively brief and focused solely on devotion-related concrete nouns.
A future lexicon might take a broader focus and include more words. The
lexicon used in the present study was able to detect anticipated patterns.
Another limitation was the relatively small gender-separated sub-samples
generated by using common male (Mohammed) and female (Fatima/
Mariam) names. Future work could greatly increase this sample size by
including additional names. The three names selected (the most common
names for UAE citizens) are associated with important Islamic personages,
which greatly reduces sub-sample representativeness of UAE twitter users.
However, our aim was to identify expected patterns, which we were able to do.
The present study was able to detect anticipated patterns of religiosity in the
UAE twitter dataset. This work has implications for longitudinal monitoring
of social change in the UAE. Future studies may explore other events, such as
global media stories, and their impact on religious sentiment.

Acknowledgments
We would like to thank Pegasus FZC for technical support.

Disclosure statement
No potential conflict of interest was reported by the authors.

Notes on contributors
Justin Thomas holds a PhD in experimental psychology from the University of Manchester,
UK. He is professor of psychology at Zayed University in the United Arab Emirates (UAE),
where he co-directs the research cluster for culture, cognition, and well-being.
Aamna Al Shehhi holds a PhD in data science from Masdar Institute of Science and
Technology in the United Arab Emirates (UAE). She works on projects exploring ‘big data’
and is based at Khalifa University in the UAE.
Ian Grey holds a PhD in psychology and a doctorate in clinical psychology. He is an
associate professor of clinical psychology at Lebanese American University, Lebanon.

CORRESPONDENCE: Justin Thomas, Zayed University, Abu Dhabi, PO Box 144534,


United Arab Emirates (UAE)
506 J. THOMAS ET AL.

References
Abd-Allah, Umar Faruq. 2004. “Islam and the Cultural Imperative.” Nawawi Foundation
Papers 1 (1): 1–14.
Abdel-Khalek, Ahmed M. 2008. “Religiosity, Health and Well-being among Kuwaiti
Personnel.” Psychological Reports 102 (1): 181–184.
Abdel-Khalek, Ahmed M, and Ghada K. Eid. 2011. “Religiosity and its Association with
Subjective Well-being and Depression among Kuwaiti and Palestinian Muslim Children
and Adolescents.” Mental Health, Religion & Culture 14 (2): 117–127.
Alsharekh, Alanoud, and Robert Sprinborg, eds. 2008. Popular Culture and Identity in the
Arab Gulf States. London: Saqi Books in association with the London Middle East
Institute at School of African Studies (SOAS).
Campbell, Heidi A., and Louise Connelly. 2015. “Religion and New Media.” International
Encyclopedia of the Social & Behavioral Sciences 20 (2): 273–278.
Cehn, Min, Shiwen Mao, Yin Zhang, and Victor C. M. Leung. 2014. Big Data: Related
Technologies, Challenges and Future Prospects. Heidelberg: Springer.
Davidson, Christian. M. 2008. “Higher Education in the Gulf States: A Historical
Background.” In Higher Education in the Gulf States, edited by
Christian M. Davidson and Peter Mackenzie Smith, 59–75. London: Saqi Books in
association with the London Middle East Institute at School of African Studies
(SOAS).
Diener, Ed, Richard E. Lucas, and Christie N. Scollon. 2006. “Beyond the Hedonic
Treadmill: Revising the Adaptation Theory of Well-being.” American Psychologist 61
(4): 305–314.
Dodds, Peter Sheridan, Kameron Decker Harris, Isabel M. Kloumann, Catherine A. Bliss,
and Christopher M. Danforth. 2011. “Temporal Patterns of Happiness and Information
in a Global Social Network: Hedonometrics and Twitter.” PLoS ONE 6 (12): e26752.
Eaton, Chris, Dirk DeRoos, Thomas Deutsch, George Lapis, and Paul. C Zikopoulos. 2012.
Understanding Big Data: Analytics for Enterprise Class Hadoop and Streaming Data.
New York: McGraw Hill.
El-Ghanim, Joseph. 2015. “YouSwear.com: Arabic Language.” Accessed
20 December 2016. http://www.youswear.com/index.asp?language=Arabic+kuwait
Fox, John W., Nada Mourtada-Sabbah, and Mohammed Al-Mutawa, eds. 2006.
Globalization and the Gulf. London: Routledge.
Freeman, Chris, and Francisco Louçã. 2002. As Time Goes By: From the Industrial
Revolutions to the Information Revolution. Oxford: Oxford University Press.
Fuller, Michael 2015. “Big Data: New Science, New Challenges, New Dialogical
Opportunities.” Zygon 50 (3): 569–582.
Gallagher, Stephen, and Warren Tierney. 2013. “Religiousness/Religiosity.” In Encyclopedia
of Behavioral Medicine, edited by Marc D. Gellman and J. Rick Turner, 1653–1654.
New York: Springer.
Garcia-Gavilanes, Ruth, Daniele Quercia, and Alejandro Jaimes. 2013. “Cultural
Dimensions in Twitter: Time, Individualism and Power.” Paper presented at the
Seventh Conference of the International Association for the Advancement of Artificial
Intelligence (AAAI) on “Weblogs and Social Media”, Boston, MA, 8–11 July.
Ghubash, Rafia, Tewfik K. Daradkeh, Sheikha M. A. Al-Muzafari, Muna E. El-Manssori,
and Mohammed T. Abou-Saleh. 2001. “Al Ain Community Psychiatric Survey IV:
Socio-cultural Changes (Traditionality-Liberalism) and Prevalence of Psychiatric
Disorders.” Social Psychiatry and Psychiatric Epidemiology 36 (11): 565–570.
JOURNAL OF CONTEMPORARY RELIGION 507

Ginsberg, Jeremy, Matthew H. Mohebbi, Rajan S. Patel, Lynnette Brammer, Mark


S. Smolinski, and Larry Brilliant. 2009. “Detecting Influenza Epidemics Using Search
Engine Query Data.” Nature 457 (19): 1–5.
Hall, Edward. T. 2000. “Context and Meaning.” In Intercultural Communication: A Reader,
edited by Larry Samovar and Richard E. Porter, 34–43. Belmont, CA: Wadsworth.
Harris, Catherine L., Ayse Aycicegi, and Jean Berko Gleason. 2003. “Taboo Words and
Reprimands Elicit Greater Autonomic Reactivity in a First Language than in a Second
Language.” Applied Psycholinguistics 24 (4): 561–579.
Hofstede, Geert. 2001. Culture’s Consequences: Comparing Values, Behaviors, Institutions
and Organizations across Nations. 2nd ed. London: Sage.
Jay, Timothy. 1992. Cursing in America: A Psycholinguistic Study of Dirty Language in the
Courts, in the Movies, in the Schoolyards and on the Streets. Philadelphia, PA: John
Benjamins.
Jesa. 2015. “10 Most Common Swear Words and Expressions in Arabic.” Transparent
Language, 20 December 2016. Accessed 21 December 2017. http://blogs.transparent.
com/arabic/10-most-common-swear-words-and-expressions-in-arabic/
Kassamali, Hasnain. 2016. “The Islamic Glossary: An Explanation of Names, Terms and
Symbols.” Ahlul Bayt Digital Islamic Library Project. The Islamic Glossary: An
Explanation of Names, Terms and Symbols. Accessed 12 December 2017. https://
www.al-islam.org/
Lee, Morgan. 2016. “Spamming the Good News.” Christianity Today 60 (5): 23.
Mehl, Matthias R., and James W. Pennebaker. 2013. “The Sounds of Social Life:
A Psychometric Analysis of Students’ Daily Social Environments and Natural
Conversations.” Journal of Personality and Social Psychology Bulletin 84 (4): 857–870.
Miller, Alan, and Rodney Stark. 2002. “Gender and Religiousness: Can Socialization
Explanations be Saved?” American Journal of Sociology 107 (6): 1399–1423.
Mourtada-Sabbah, Nada, Mohammed Al-Mutawa, John W. Fox, and Tim Walters. 2008.
“Media as Social Matrix in the United Arab Emirates.” In Popular Culture and Political
Identity in the Arab Gulf States, edited by Alanoud Alsharekh and Robert Sprinborg,
121–142. London: Saqi books in association with the London Middle East Institute at
the School of African Studies (SOAS).
NBS (National Statistics Bureau). 2009. “Population Estimates (2006–2009).” Abu Dhabi,
United Arab Emirates: UAE National Bureau of Statistics. Accessed 14 December 2017.
http://ghdx.healthdata.org/organizations/national-bureau-statistics-united-arab-
emirates
Ritter, Ryan S., Jesse Lee Preston, and Ivan Hernandez. 2013. “Happy Tweets: Christians
are Happier, More Socially Connected, and Less Analytical than Atheists on Twitter.”
Social Psychological and Personality Science 5 (2): 243–249.
Schnabel, Landon. 2015. “Review of Why are Women More Religious than Men?”
Secularism and Nonreligion 4 (1): Art. 7.
Stark, Rodney. 2002. “Physiology and Faith: Addressing the ‘Universal’ Gender Difference
in Religious Commitment.” Journal for the Scientific Study of Religion 41 (3): 495–507.
Sylwester, Karolina, and Matthew Purver. 2015. “Twitter Language Use Reflects
Psychological Differences between Democrats and Republicans.” PLoS ONE 10 (9):
e0137422.
Justin, Thomas. 2014. Psychological Well-being in the Gulf States: The New Arabia Felix.
London: Palgrave Macmillan.
Justin, Thomas, Richard P. Bentall, Lowri Hadden, and Lily O’Hara. 2016. “Ethnic Identity
and Paranoid Thinking: Implicit Out-group Preference and Language Dominance
508 J. THOMAS ET AL.

Predict Paranoia in Emirati Women.” Journal of Behavior Therapy and Experimental


Psychiatry 56: 122–128.
Justin, Thomas, Amal Al-Marzooqi, and Monique Raynor. 2012. “Marital Status and
Gender as Predictors of Undergraduate Academic Performance: A United Arab
Emirates Context.” Learning and Teaching in the Gulf 9 (2): 1–9.
Thomas, Justin, Susanne Quadflieg, and Lily O’Hara. 2016. “Implicit Out-group Preference
is Associated with Eating Disorders Symptoms amongst Emirati Females.” Eating
Behaviors 21: 48–53.
Wang, Wenbo, Lu Chen, Krishnaprasad Thirunarayan, and Amit P. Sheth. 2014. “Cursing
in English on Twitter.” Paper presented at the 17th conference of the Association for
Computing Machinery (ACM) on “Computer Supported Cooperative Work & Social
Computing”, Baltimore, MA, 15–19 February.
WHO (World Health Organization). 2006. “Country Cooperation Strategy for WHO and
the United Arab Emirates 2005–2009.” Cairo: WHO.
Würtz, Elizabeth. 2005. “A Cross-cultural Analysis of Websites from High-context
Cultures and Low-context Cultures.” Journal of Computer-Mediated Communication
11 (1): article 13.
Yang, Chao, and Padmini Srinivasan. 2016. “Life Satisfaction and Pursuit of Happiness on
Twitter.” PLoS ONE 11 (3): 1–30.

You might also like