ArticlePDF Available

Abstract and Figures

Flags are important national symbols that have transcended into the digital world with inclusion in the Unicode character set. Despite their significance there is little information about their role in online communication. This paper examines the role of flag emoji in political communication online by analyzing 640,676 tweets by the most important political parties and Members of Parliament in Germany and the USA. We find that national flags are frequently used in political communication, and are mostly used in-line with political ideology. As offline, flag emoji usage in online communication is associated with external events of national importance. This association is stronger in the USA than in Germany. The results also reveal that the presence of the national flag emoji is associated with significantly higher engagement in Germany irrespective of party, whereas it is associated with slightly higher engagement for politicians of the Republican party and slightly lower engagement for Democrats in the USA. Implications of the results and future research directions are discussed.
Content may be subject to copyright.
The Role of Flag Emoji in Online
Political Communication
Ankit Kariryaa
, Simon Rund´
, Hendrik Heuer
Andreas Jungherr
, and Johannes Scho
Flags are important national symbols that have transcended into the digital world with inclusion in
the Unicode character set. Despite their significance, there is little information about their role in
online communication. This article examines the role of flag emoji in political communication online
by analyzing 640,676 tweets by the most important political parties and Members of Parliament in
Germany and the United States. We find that national flags are frequently used in political com-
munication and are mostly used in-line with political ideology. As off-line, flag emoji usage in online
communication is associated with external events of national importance. This association is
stronger in the United States than in Germany. The results also reveal that the presence of the
national flag emoji is associated with significantly higher engagement in Germany irrespective of
party, whereas it is associated with slightly higher engagement for politicians of the Republican party
and slightly lower engagement for Democrats in the United States. Implications of the results and
future research directions are discussed.
political communication, political parties, social media, emoji, flag, Twitter
Flags are important national symbols that provide a tangible representation of the shared national
and political identity (Harr´e, 2002; Jaskulowski, 2016; Schatz & Lavine, 2007; W. Smith, 1975).
They are a common element in political communication (Schill, 2012), and they significantly impact
political cognitions and feelings of political affiliation (Carter et al., 2011; Chan, 2017; Hassin et al.,
2007; Kalmoe & Gross, 2016). Political actors use flags to establish an association between them
and symbols of common identity to send subliminal cues triggering subconscious prejudices or to
signal alignment with a shared set of values or beliefs (e.g., Ehrlinger et al., 2011; Kalmoe & Gross,
2016). Correspondingly, studies show that exposure to flags can increase feelings of attachment to
the nation as an abstract concept or feeling of superiority toward other nations (Kemmelmeier &
Winter, 2008; Schatz & Lavine, 2007).
University of Bremen, Germany
University of Konstanz, Germany
Corresponding Author:
Ankit Kariryaa, University of Bremen, Bremen 28359, Germany.
Social Science Computer Review
ªThe Author(s) 2020
Article reuse guidelines:
DOI: 10.1177/0894439320909085
Flags thus serve as powerful symbols of belonging and specific interpretations of a nation. Yet, as
classic work on symbolic interactionism argues, symbols can vary in their interpretation across con-
texts, groups, or time (H. S. Becker, 1988; Blumer, 1966; Snow, 2001). Forexample, we find the use of
flags to vary greatly across countries due to history and cultural norms. In Germany, national flags are
rarely displayed in public and private spheres after World War II (Elias, 1992), while in the United
States, public display of flags is a common phenomenon (e.g., Billig, 1995; Butz et al., 2007; Ferguson
& Hassin, 2007). Yet as recent controversies in the Colin Kaepernick/National Football League (NFL)
kneeling incident (Boykoff & Carrington, 2019) show, this unpartisan way of interacting with the flag
as a national symbol in the United States might be contingent on political actors not choosing to use it
as a symbol of political division. Finally, studies show that exposure to flags tends to increase support
of political parties strongly emphasizing national identity and belonging and not others (Carter et al.,
2011; Chan, 2017). As symbolic interactionists would expect, we find the meaning and power of flags
as a symbol to shift between cultural, national, political, and cultural contexts. We examine conse-
quences of these diverging meanings in the use of flags in online communication by politicians in
Germany and the United States on the microblogging service Twitter.
Recently, flags were introduced to online communication through the integration in Unicode
6.0.0, which was released in 2010 and which included 256 national flags and along with seven other
flags ( , , , , , , ; Unicode 6.0.0, 2010). Other than for the uses and effects of exposure
to actual flags, little is known about the uses and effects of their digital representations in form of
emoji. To fill this gap, we examine the use of flag emoji in 640,676 tweets by the most important
political parties and members of parliament (MP) in Germany and the United States. This allows us
to examine whether the use of emoji varies between parties and MP depending on their political
leaning, depends on events of shared national importance, varies in semantic context, and has
divergent effects for different parties depending on their emphasis of national identity. The
between-country comparison allows us to identify whether the effects of flag use are universal or
contingent on cultural and political contexts.
We find that in both Germany and the United States the national flag emoji are frequently used in
online political communication, with parties and politicians on the political right using the flag emoji
more frequently than those on the left. Among these users, national flag emoji are very prominent.
They are even more frequent than facial emoji, such as [winking face] or [face with tears joy],
which generally are the most widely used emoji in online communication (Hu et al., 2017). We
observe a high association of national flag emoji with national events in United States, whereas this
association is somewhat weaker in the case of Germany. Political actors use flag emoji in the context
of their political ideology, and the meaning of flag emoji changes notably for different parties. This
is especially true in Germany, where we see more variance between parties. In the United States,
both parties use the national flag in context of elections and to represent the country; however, we do
not observe a clear divide along ideologies of the political parties. Interestingly, the national flag is
strongly associated with higher engagement for all political parties in Germany irrespective of their
ideology. This is most pronounced for Alternative for Germany (AfD), Gru
¨ne, and Free Democratic
Party (FDP), where we see a more than 5-fold increase in engagement for tweets with a national flag.
In the United States, presence of flag emoji is associated with increase in engagement for Repub-
licans, whereas it is associated with decrease in engagement for Democrats, though only marginally
for both parties.
These findings demonstrate that as symbolic interactionists would expect, the uses and meaning
of national flags as a symbol vary across cultural and political contexts and have different conse-
quences as expressed in stimulating audience reactions. Here the divergent contextual uses of
national flags in tweets of political actors depending on their ideological leaning are especially
interesting and offer promising perspectives for future research. More generally, our findings
2Social Science Computer Review XX(X)
illustrate the research potential examining varying uses and effects of symbols in digital commu-
nication in understanding their power with audiences (Jungherr, 2015).
Related Work
Our article draws on multiple lines of research on understanding the function and importance of
national symbols and role of emoji in online communication.
Role of National Flags
Schatz and Lavine (2007) found that national symbols promote national identification by signifying
the group and providing a tangible representation of the group which communicates groupness,
promotes the in-group identification and out-group distinction. Additionally, they suggest that
symbols supplement an individual’s identity with a timeless group entity. In his seminal work,
Mueller (1970) studied the increased support of the president of the United States during time of
crisis and formulated the theory of “Rally around the national flag.” It suggests that during crisis the
president is seen as an embodiment of national unity, in a similar way as a national symbol. Among
the national symbols, national flag is arguably the most important one. On the value of national flag,
Durkheim (1911/1953) wrote:
A flag, as such, is only a piece of cloth from which no emotional meaning can be derived. However, the
emotional meaning of the flag can become so dramatic that people are willing to sacrifice their life for it.
The flag is the bearer of the notion of collectivity; it represents the soul of society and, as such, the flag is
sacred. (p. 87)
The role of national flags extends far beyond their symbolic function. Flags are also known to
significantly alter people’s opinion and behavior. In one such study, Hassin et al. (2007) analyzed
the effect of exposure to national flag in Israel on political thought and behavior. By examining
participant’s stance on issues of national importance, they found out that national flags significantly
influence political thoughts and pulled people toward the center from both left and right side of
political extremes. On the same lines, J. C. Becker et al. (2011) analyzed the effect of exposure to
national flag in Germany and found that it increased out-group prejudice among highly nationalistic
participants. In a similar study in the United States, Kemmelmeier and Winter (2008) found that
exposure to national flag increased nationalism, which they defined as a sense of superiority over
others. A study conducted later on by Butz et al. (2007) in the United States contradicts the result of
Kemmelmeier and Winter. Butz et al. found that exposure to national flag reduces out-group
hostility among highly nationalistic participants. They argue that exposure to national flag increases
egalitarianism and humanitarian values, core characteristics of American identity. On the contrast
with earlier studies, Butz et al. reasoned that reaction to the flag exposure could depend upon the
social context. Even though the direction of the effect is sometimes debated, there is a consensus that
exposure to national flag significantly affects the audience.
In the political context, Kalmoe and Gross (2016) found that American flag has a pro-Republican
effect on the people identifying as Republicans but offers no advantage to the Democrats. In a
similar study, Carter et al. (2011) found limited evidence that exposure to national flag can shift
support toward Republicans for months and in general alter people’s behavior and political judg-
ment. Flag emoji have increased the accessibility of the national flag in online communication and
since flags have varying effects on people with different political views, some political parties may
benefit more from increased accessibility than others.
Kariryaa et al. 3
We compare the intensity of engagement toward posts from various political parties to check
whether flag emoji disproportionately affects their audience. On engagement, Khan (2017) wrote
“Engagement may be viewed as an individual’s interaction with the media. (p. 237)” Online inter-
actions such as likes, retweets, and replies can have various meaning and purposes, and they do not
imply support for the content (Dhir et al., 2019; Hayes et al., 2016; Lee et al., 2016; Scissors et al.,
2016; Wohn et al., 2016). In this study, we do not distinguish the purpose of an interaction and only
consider the total amount of interaction in form of favorites and retweets. To limit confounding
factors, we measure the relative change in engagement within a political party and only compare
these relative changes among the parties. Here we assume that the confounding factors would
symmetrically affect the tweets with national flag and tweets without national flag, within a party.
Role of Emoji
We investigate the prevailing role of emoji in communication to better understand the role of flag
emoji. The recent progress in emoji-related research postulates the varying roles emoji serve in a
text. Na’aman et al. (2017) identified three main functions of an emoji. Firstly, as a replacement by
a similar sounding word, for example, as a replacement for “do not,” secondly as stand-in for a
lexical word or phrase, such as, emoji for the word “tomato,” and lastly as a multimodal affective
marker such as to show happiness. They also found that emoji are most commonly used as a
multimodal affective marker.
Their findings are supported by that of Hu et al. (2017) which further indicate that facial emoji are
mostly likely used as a multimodal marker. They studied the intentions and sentiments of an emoji
by dividing them into positive, negative, and neutral facial emoji and nonfacial emoji. Their results
indicate that emoji are mostly used to express sentiment, strengthen expression, and adjust tone.
And, they found that the sentiment effect of the nonfacial emoji is relatively small as compared to
other categories. Nonetheless, flags in off-line communication are very powerful and influential
symbols (Butz et al., 2007; Carter et al., 2011; Chan, 2017; Ehrlinger et al., 2011; Hassin et al., 2007;
Kemmelmeier & Winter, 2008), this motivated us to investigate whether flag emoji would behave
more like their off-line equivalents or would they behave more like their categorical emoji counter-
parts. We test this by studying the engagement with national flag emoji and comparing them with
other emoji. Since related work suggests that anthropomorphic emoji are the most commonly used
emoji, we check if this also holds true for specific contexts such as political communication.
Meaning of an Emoji
Another goal of the study is to understand the differences in the meaning and use of the national flag
differs across countries as well as across parties. This question is directly motivated by the theory of
symbolic interactionism (H. S. Becker, 1988; Blumer, 1966; Janssen & Verheggen, 1997; Snow,
2001). Symbolic interactionism argues that, through social interaction, symbols come to develop
seemingly stable meanings, but those meanings are contingent upon person and context. Different
groups of people could develop different interpretations of a symbol, and these interpretations may
change based on the specific context.
Similar studies have been conducted in the context of emoji, for example, Barbieri et al. (Barbieri,
Kruszewski, et al., 2016) studied the semantics of emoji in varying sociogeographical situations.
They found that meaning of some emoji changes depending upon the context. They also suggest that
meaning ascribed to certain emoji is not universal but something that needs to be investigated for
groups of users. However, there has been little discussion about whether this variance of meanings is
also applicable with regard to emoji representing strong national symbols like national flags. This
motivated us to study if the national flag is used in the same context by various parties.
4Social Science Computer Review XX(X)
To study the meaning of the flag emoji, we use the distribution hypothesis. In linguistics, the
distributional hypothesis states that words with similar meaning occur in similar contexts. This
implies that the meaning of a word can be inferred from its distribution across contexts (Sahlgren,
2005). In natural language processing, the distributional hypothesis was implemented in word
representations, sometimes also called word embeddings. Such representations are increasingly used
to represent the meaning of words and sentences (Mikolov, Sutskever, et al., 2013). Barbieri et al.
used a vector space skip-gram model on Twitter emoji and 10 million tweets posted by users from
the United States (Barbieri et al., 2016). In their qualitative evaluation, they found that semantically
similar emoji are represented similarly as emoji and words with similar meanings.
Motivated by the lack of emoji in resources like the Google News-based word representations,
Eisner et al. (2016) released pretrained vector representations of all Unicode emoji learned from
their description in the Unicode emoji standard. They show that such representations are useful for
the downstream task like sentiment analysis. They also found that emoji representations learned
from short descriptions outperform word representations trained on a large collection of tweets. This
motivated us to train our own vectors since the investigation shows that the meaning captured by the
representations differs from those trained-on tweets, that is, that the emoji have a specific meaning
on such platforms.
Case Selection and Research Questions
We investigate the role of national flag emoji in online political communication in international
comparison. In this, we explore if the use of national flag emoji varies between political parties, over
time, contexts, and whether it affects engagement with political content. This helps us to understand
varying meanings and effects of flag emoji in political communication. Given the varying emphasis
on national identify by political parties, we are especially interested if use of the flag emoji benefited
some parties more than others in increasing engagement with content posted by them or their MP.
This leads us to the following research questions:
Research Question 1: How does the national flag emoji usage differ across political parties?
Research Question 2: How do the external events impact the flag emoji usage?
Research Question 3: How does the meaning of national flag emoji differ across parties?
Additionally, as noted above, national flags act as symbols of common identity and can send cues
to audience members. Accordingly, we test if the use of flag emoji translates into higher engagement
metrics, potentially driven by the heightened symbolic appeal of flag emoji:
Research Question 4: Is the national flag emoji associated with higher engagement?
To answer these questions, we study the flag emoji usage on Twitter by political parties and MP
in Germany and United States. Social media is integral to contemporary political communication.
Politicians, journalists, and public alike use social media to publicly comment on the news of the
day, strategically post information, or research information (Evans et al., 2018; Jungherr, 2016;
Jungherr et al., 2020; McLaughlin & Velez, 2019; Schroeder, 2018; Spierings et al., 2019). We focus
on the use of Unicode characters representing national flags in Twitter posts by parties and MP in
Germany and the United States.
Germany and United States are promising contrasting cases. Both countries have different rela-
tionships to national identity and national symbols. The use of the national flag in Germany as a
national symbol is highly contested as could be seen by an ongoing public discussion about the
Kariryaa et al. 5
appropriateness of the ubiquitous use of German flags during soccer world cup games (Dohmen
et al., 2006; Heinrich, 2003; Stehle & Weber, 2013). Before the FIFA World Cup in 2006, the
German newspaper Spiegel (2006) published an article with the title: “Germany’s Patriotism Prob-
lem; Just Don’t Fly the Flag.” The article highlighted the very sensitive topic of nationalism in
Germany. And the national flag, a symbol of national identity, was at the core of the debate. After
World War II, the national flag had become a taboo, and it became rare to see the flag on buildings
and houses, unlike in many other countries (Elias, 1992). Starting from 2006, Germany’s association
with their national flag has changed a lot. A major turning point was the FIFA World Cup itself.
Shortly after the World Cup, New York Times published the article “In World Cup Surprise, Flags
Fly With German Pride” (Bernstein, 2006), emphasizing that at least for the period of the sporting
event, Germans embraced the national flag. However, even today, the relationship with nationhood
and national symbols in Germany remains politically conflicted. This behavior is not limited to the
common public, and politicians also avoid from waving the national flag during party or national
events (Roßmann, 2016). This more conflicted relationship with the flag might translate into Ger-
man politicians using the flag more sparingly in their posts and if they do so less as an overt
reference to national identity but instead as a playful replacement for the word “Germany” in their
posts. Further, we should expect differences in the use of the national flag emoji by party affiliation,
given a party’s relationship to nationhood and national symbols.
In the United States too, liberals and conservatives associate different meaning to national flag as
recently seen in the Colin Kaepernick/NFL kneeling controversy (Boykoff & Carrington, 2019).
However, differences are smaller as compared to Germany, and we find general political consensus
across the political spectrum concerning the importance of national identity and national symbols,
especially the American flag (Huntington, 2004; A. D. Smith, 1991). We can thus expect represen-
tatives of both major parties, Democrats and Republicans, to use flag emoji with the same propensity
and potentially more in line with national identity.
For Germany, we considered the six main political parties represented in the Bundestag (Lower
House of the Parliament). Sorted from left to right of the political spectrum, these parties are Linke
(The Left), Gru
¨ne (Alliance 90/The Greens), Social Democratic Party of Germany (SPD), FDP,
coalition of Christian Democratic Union of Germany and Christian Social Union of Germany (CDU/
CSU), and AfD. Similarly, for the United States, we focused our analysis on the two main parties in
the House of Representatives of the U.S. Congress: the Democratic Party and the Republican Party.
Study Platform: Twitter
Twitter was our choice of social media platform for this study for two main reasons. Firstly, due to
Twitter’s growing importance as a political communication channel (Jungherr, 2016). And secondly,
we observed that posts on Twitter often contained more emoji than similar posts on other social
media platforms. Figure 1 (bottom) shows one such example. Michael Roth, a member of German
Bundestag from SPD, uses no emoji in a Facebook post as seen on the left, while similar post on
Twitter, as seen on the right, contains and emoji in the text and in the screenname. This
difference between the text across two platforms could be due to redundancy in the Facebook’s input
interface where you can add an emoji either as or directly though the icon. Or it could
also be due to the strict character limit on Twitter where replacing words with emoji can save space.
Data Set
For each political party, we looked at the party’s official Twitter accounts at the national and state
level and the Twitter accounts of the members of the 18th and 19th Bundestag in Germany and
members of the House of Representatives of 115th and 116th U.S. Congress. We included members
6Social Science Computer Review XX(X)
from the current and the previous sessions of parliament of the two countries to minimize biases in
comparison since the two countries held elections at different point of time. Elections for the 19th
Bundestag were held of September 24, 2017, and elections for 116th U.S. Congress were held on
November 6, 2018.
Using the described method, we collected 642 accounts from Germany and 634 Twitter accounts
from the United States. In there, we had 547 accounts from members of German Bundestag (MdB)
and the rest from the official party outlets at national and state level. For the United States, we
collected 531 accounts from the Republican and Democrat members of the House of Representatives
of the U.S. Congress, while the rest of the accounts were official accounts of the parties at national
and state levels. In rare occasions, we found multiple Twitter accounts for a politician; in such cases,
we choose the account that was designated as official or the account that was oriented toward politics
instead of personal use. We focused on the members of Bundestag and Congress plus the state and
national accounts to give us a comprehensive set of active and official political accounts on Twitter.
We started our data collection in June 2018. Using the Twitter application programming interface
(API; User timelines, 2016), we downloaded up to 3,200 tweets for each account. This data set was
periodically updated for new tweets for each account posted after June 2018 until March 2019. In
this way, we collected a total of 3.04 million tweets.
Filtering Tweets
We removed tweets outside the period of March 1, 2017, to Feb 28, 2019, which left us with a total
of 1,605,370 tweets. March 1, 2017, was chosen as the starting date since by then about 75%of the
Twitter accounts had at least one tweet in our corpus, and it provided us good coverage for both
countries. About quarter of accounts did not have tweets covering the entire period as some accounts
were only opened in this period, some were deleted in this period and for rest the initially down-
loaded tweets did not extend as far back as March 1, 2017.
We also removed retweets, quoted status (retweets with additional text), and replies from this
corpus. This left us with a total of 640,676 tweets, out of which 218,077 were from Germany and
422,599 from the United States. There were two main reasons for focusing on original tweets for our
Figure 1. The differences in emoji use in a post on Facebook (on the left) and Twitter (on the right) by
Alternative for Germany and Michael Roth, members of German Bundestag Social Democratic Party of
Kariryaa et al. 7
analysis. Firstly, to eliminate the problem of double counting of tweets, which were retweeted by
another person in our data set, and secondly, to limit the analysis to the flag usage in the writing style
of national representatives.
Tables 1 and 2 show the number of accounts, tweets, tweets with an emoji, and tweets with a
national flag for each party in Germany and United States, respectively. For Germany, European
Union is the macroregion; hence, the number of tweets with the emoji are also included.
Similarity of the Emoji
To investigate the similarity of different words and emoji, we operationalized similarity as co-
occurrence, that is, how often a specific emoji or a specific word occurs together with another emoji
or word. This connects to the distributional hypothesis, which states that words with similar meaning
occur in similar contexts (Sahlgren, 2005). There are a variety of computational models that imple-
ment the distributional hypothesis, including word2vec (Mikolov, Sutskever, et al., 2013), GloVe
(Pennington et al., 2014), and Random Indexing (Sahlgren, 2005). We used the word2vec imple-
mentation in Gensim (Mikolov, Chen, et al., 2013; Rehurek & Sojka, 2010). word2vec takes a text
corpus as input and produces vector representations as output. We trained word2vec using the
continuous-bag-of-words approach for all words that appear at least 3 times. We trained the neural
network to predict a certain word using a context window of 3, that is, we used the three words
preceding the current word and the three words succeeding the current word to predict it. For
example, in the sentence “Angela Merkel is the chancellor of ,” the term “chancellor” is predicted
using the words “Merkel,” “is,” “the,” “of,” and “ .” To illustrate what this means semantically,
LeCun provides the example of a news story (LeCun et al., 2015). In a news story, words like
Tuesday and Wednesday can be easily replaced, that is, they are used very similarly in any given
sentence. The word representations also have other properties: In the vector space of word2vec,
similar words are encoded by similar vectors. This means that the angle between the vectors can be
used to quantify the similarity of the words. In this investigation, we use the cosine between two
vectors to investigate how similar two words are. To train the word2vec model, we removed all
Table 1. For Each Party in Germany; Number of Accounts, Tweets, Tweets With an Emoji, Tweets With ,
and Tweets With .
Party Accounts Tweets Tweets With Emoji Tweets With Tweets With
Linke 78 36,129 1,442 31 12
Gru¨ne 87 37,637 3,663 102 147
SPD 146 47,766 5,469 563 813
FDP 80 24,773 3,257 167 288
CDU/CSU 156 38,189 2,955 233 90
AfD 96 33,583 7,375 1,633 7
Note. SPD ¼Social Democratic Party of Germany; FDP ¼Free Democratic Party; CDU/CSU ¼coalition of Christian
Democratic Union of Germany and Christian Social Union of Germany; AfD ¼Alternative for Germany
Table 2. For Both Parties in United States; Number of Accounts, Tweets, Tweets With an Emoji, and Tweets
With .
Party Accounts Tweets Tweets With Emoji Tweets With
Democrats 327 281,883 10,613 718
Republicans 307 199,624 9,356 1,624
8Social Science Computer Review XX(X)
special characters except for the #, -, and _. The first was kept due to its semantic meaning as a
hashtag, the last two were important for the emoji definitions. We further replaced German umlauts
(a¨, o¨, u
¨) and converted all words to lowercase.
In this section, we first present high-level descriptive statistics about the use of flag emoji in our data
set. We then present results specific to research questions that we formulated in the beginning.
Descriptive Statistics
Overall, we were surprised to find that the national flag emoji are among the most frequently used
emoji in both countries, and they are more frequent than the widely popular anthropomorphic emoji.
Figure 2A and 2B show the 10 most commonly used emoji in political communication on Twitter in
Germany and United States in our data set. The national flag is the second and third most used emoji
in Germany and United States, respectively. The flag of the European Union, which is a political and
economic union of 28-member states including Germany, is the second most used flag in Germany.
This demonstrates that in specific contexts such as political communication, topic related nonan-
thropomorphic emoji are popular ones. Interestingly, anthropomorphic emoji and symbols/markers
both appear frequently in case of Germany, and in United States symbols are the prevailing emoji.
The emoji, which is most frequently used in both Germany and United States, is mostly used for
point to external weblinks in a tweet.
Investigating the proportion of accounts that used emoji at all, we find that in Germany, 78.8%
accounts used emoji and 29.3%used the national flag emoji at least ones during our study period.
We also find that emoji usage varies greatly between different accounts. For example, for some
accounts, more than 90%of tweets contained an emoji and 60%of the tweets contained the national
flag emoji as compared to many accounts that did not use any emoji in period. In United States,
82.9%accounts used an emoji and 43.6%used a flag emoji at least ones. In the United States, for
some accounts, more than 80%of tweets contained an emoji and 70%of the tweets contained the
national flag emoji.
Research Question 1: How does the national flag emoji usage differ across political parties?
To understand how the emoji usage varies across parties, we analyzed the percentage of tweets
that contained national flag emoji for each account. Figure 3A and 3B show the result for Germany
and United States, respectively.
Figure 2. Panel A (left): Ten most used emoji in political communication in Germany. Panel B (right): Ten most
used emoji in political communication in United States in our data set. Note: National flag is the second most
used emoji in Germany and third most in the United States.
Kariryaa et al. 9
In both countries, right-wing parties use emoji and flag emoji more than the left-wing parties. In
Germany, 4.86%tweets from the far-right AfD, but only 0.08%tweets from the far-left Die Linke
contained the emoji. This pattern also holds for United States and Republicans and Democrats,
even though the ideological distance between Republicans and Democrats is closer than the ideo-
logical distance between AfD and Linke (The Political Compass, n.d.). Overall, Republicans use an
emoji in 4.77%tweets and the national flag emoji in 0.81%tweets, whereas the Democrats used
emoji in 3.76%tweets and national flag emoji in 0.25%tweets.
Research Question 2: How do the external events impact the flag emoji usage?
In this section, we study the impact of external events including major national and sporting
events on the flag emoji usage in the two countries. Figure 4A and 4B show the daily percentage of
tweets with and emoji for Germany and United States, respectively. In Germany (Figure 4A),
the peaks around external events are distinguishable, and they are usually twice as large as the
surrounding data. The highest peak is observed on June 23, 2018, when the German soccer team won
against Sweden in a game crucial to stay in the FIFA World Cup. On German Unity Day (October 3),
a local spike can be observed in 2018 but not in 2017. Another spike can be seen on January 22, with
22 German flags used on the anniversary of ´
Elys´ee Treaty between Germany and France. In general,
around FIFA World Cup (June–July 2018), we observed higher national flag usage than around
national elections (September 2017).
Figure 3. Panel A (top): Each dot represents an account’s percentage of tweets with emoji and any emoji.
Panel B (bottom): Each dot represents an account’s percentage of tweets with emoji and any emoji.
10 Social Science Computer Review XX(X)
For the United States (Figure 4B), the flag emoji usage is more closely related to the national
events, and peaks are notably higher than the baseline. We observe the dominating spikes on the July
4, 2017 and 2018 (Independence Day), and local spikes on November 10, 2017 and 2018 (Veterans
Day), January 31, 2018 (after the State of the Union delivered by the president on the previous day),
and May 28, 2018 (Memorial Day).
Since the national flag usage in the United States is linked to the national events, it is possible that
the flag usage is an overt reference to national identity. However, in case of Germany due to
conflicted relationship with the national flag, the usage is probably driven more by the context of
the text instead of the external event and hence the smaller differences.
Research Question 3: How does the meaning of national flag emoji differ across parties?
Since flag emoji are so frequently used, our third research question explores the meaning of flag
emoji in greater detail. We trained independent word2vec models for every party and one additional
model combining data from all parties. Table 3 shows which words and emoji are the most similar to
Figure 4. Panel A (top): Daily percentage of tweets containing the from all German parties. Panel B
(bottom): Daily percentage of tweets containing the from both American parties.
Kariryaa et al. 11
the emoji. Note that this similarity is based on cosine similarity, ranging from 1.0 for dissim-
ilarities and 1.0 for the highest possible similarity.
At a high level, we find that in Germany the national flag is mostly used in the same context as the
country name Deutschland, words relevant to the current affairs in Germany and flags of neighboring
countries ( , , ) and European Union ( ). However, there are differences in the parties along their
political and economic agenda. Linke uses the national flag in context of social issues such as
#EUNAVFORMED in context of military mission to stop human smuggling in the Mediterranean,
#le2511 in support of Anti-fascist movement (Antifa) in Leipzig, and #germandronewars to oppose the
development of autonomous drones for wars. For Linke, we had only 31 tweets with German flag,
hence the results should be interpreted with care. Gru
¨ne uses the national flag in context of environ-
mental issues such as #erneuerbare (renewable energy), Innensta¨dten (with regard to the pollution in
inner cities), and #greenfinance. Additionally, words and hashtag related to social issues (#integration,
Anker-Zentren) and foreign policy (#sanktionen, #irland and Sweden) also appear in the similar
context. In case of SPD, flags of neighboring countries ( , , , , ) and emoji used for
empowerment and support for communities ( , , , ) appear in similar context. For FDP, we
observe words related to trade and economic issues such as Innovation, Belastungen (charges),
abzubauen (reduce), and Subventionen (subsidy). In case of CDU/CSU, we see the flag used in
the similar context along with words related to economic and social issues (bauen, belohnt, kosten-
¨nstiger, and Schutzbedu
¨rftige). Finally, AfD is a far-right (Chase & Goldenberg, 2019) political
party with a focus on German national identity. AfD has the largest number of tweets with German flag
in our data set. Due to AfD’s strong anti-refugee stance, it is unsurprising that AfD uses Deutschland
(Germany), Europa (Europe), Zukunft (future), Vaterland (homeland), Sozialsystem (social system),
Heimat (home), Haftanstalten (prison), and Fluechtlingsstrom (refugee flood) in a similar context as
the national flag. In general, the context in which the national flag used by political parties in Germany
varies with their social, environmental, foreign, and economic policy.
Unlike Germany, in the United States (Table 4), other flags and emoji are predominantly
used in similar context as the national flags. Interestingly, compared to Germany both U.S.
parties use the national flag more strongly in the context of electoral politics than policy. In
case of the Republicans, we observe country name (#usa), campaign slogan (#maga), party
symbol ( ), and the flags of other countries ( , , , , ). For the Democrats, we
find party and election-related symbols ( ,), flags of neighboring country ( )andemoji
that show solidarity with queer and Black people ( , , ). sometimes also refers to
the Purple Heart, the oldest military decoration still given in the United States, awarded to
those wounded or killed while serving.
The results are well in line with our expectation that different parties would use flag emoji in
different context. And they are in agreement with predictions of the symbolic interaction theory
regarding divergent meanings across countries. While in Germany, the use of the national flag emoji
function as an overriding symbol of the country for parties to align with their policy preferences, in
United States, the use of the national flag appears to be more strongly related to the context of
electoral competition.
Research Question 4: Is the national flag associated with higher engagement?
For this section, we operationalized engagement as the number of retweets. Additionally, to
investigate whether the presence of an emoji leads to a higher engagement with content, we calcu-
lated the median number of retweets of all tweets and compared the number to the median number of
retweets that contain either an emoji, a flag emoji or the national flag. In our data set, the favorites
(sometimes referred to as likes) and retweets are highly correlated with a Pearson correlation
12 Social Science Computer Review XX(X)
Table 3. Results for Research Question 3, 10 Most Similar Words With Their Cosine Similarity to the Emoji.
All Parties (2,729) Linke (31) Gru¨ne (102) SPD (563) FDP (167) CDU/CSU (233) AfD (1,633)
1 #deutschland
#eunavformed (0.99) #sanktionen (0.99) (0.96) innovationen (0.99) (0.95) deutschland (0.91)
2 frankreich (0.7) bzw (0.99) anker-zentren (0.99) (0.94) belastungen (0.99) ausnahmen (0.95) europa (0.9)
3 deutschland (0.7) barcelona (0.99) flaechendeckend (0.99) integrationsrates (0.93) abzubauen (0.99) gewahrt (0.95) zukunft (0.87)
4(0.69) schwarzen (0.99) #greenfinance (0.99) (0.92) kitas (0.99) stabil (0.95) #deutschland (0.86)
5 #frankreich (0.69) kompass (0.99) #erneuerbare (0.99) (0.92) subventionen (0.99) schutzbeduerftige
vaterland (0.85)
6 europa (0.69) @zdebelhubertus (0.99) innenstaedten (0.98) (0.92) zb (0.99) entlastet (0.94) badewanne (0.85)
7(0.67) #le2511 (0.99) #irland (0.98) (0.91) fahrverbote (0.99) legale (0.94) sozialsystem (0.84)
8 #europa (0.66) kriterien (0.99) schweden (0.98) (0.9) tempo (0.99) bauen (0.94) heimat (0.83)
9 grossbritannien
monika (0.99) #integration (0.98) (0.88) mut (0.99) belohnt (0.94) haftanstalten (0.83)
10 (0.65) #germandronewars
(0.87) #deutschland (0.99) kostenguenstiger
Note. Independent word2vec model was trained from all tweets of each party and an additional model for all tweets from Germany. For each party, count of tweets with is shown next to the
party name. SPD ¼Social Democratic Party of Germany; FDP ¼Free Democratic Party; CDU/CSU ¼coalition of Christian Democratic Union of Germany and Christian Social Union of
Germany; AfD ¼Alternative for Germany.
coefficient of .95 for Germany and .89 for United States. And the overall trends described here for
retweets also hold for favorites.
The results for Research Question 4 are shown in Figure 5A and 5B. In Germany, for all parties’
tweets that contain national flag emoji have a higher median engagement. The highest impact is seen
for the Gru
¨ne, FDP, and AfD, where we see more than a 5-fold increase in the median retweet count
for tweets with . The lowest impact can be observed for the ruling parties, CDU/CSU and SPD. To
provide additional support that a difference in engagement between tweets with a national flag emoji
and tweets without a national flag emoji exists, we used Mood’s (1950) median test. It is a non-
parametric test that is used to test whether the medians of two samples are identical. We used the
test’s implementation available in the SciPy Stats package with default parameters (SciPy Version
1.2.1 Reference Guide, 2019). We found a significant difference for all political parties in Germany,
Linke (median with flag ¼14, median without flag ¼4; p¼.002), Gru
¨ne (median with flag ¼10.5,
median without flag ¼2; p< .001), SPD (median with flag ¼3, median without flag ¼1; p< .001),
FDP (median with flag ¼6, median without flag ¼1; p< .001), CDU/CSU (median with flag ¼1,
mean ¼5.87, SD ¼12.59, median without flag ¼1, mean ¼3.67, SD ¼17.92; p¼.005), and AfD
(median with flag ¼157, median without flag ¼15; p< .001). For Linke, we should consider that
there are only 31 tweets with a flag emoji in the data set.
Compared to Germany, in the United States, the effect of national flag on the engagement of a
tweet is minuscule. For Democrats, the tweets with a national flag have a lower median retweet
count (median: 6.0) as compared to tweets without a national flag (median: 7.0). Even though the
differences are small, the results are statistically significant (Mood’s median test; p< .001). For
Republicans, the trend is opposite. The median number of retweets for tweets with national flag
(median: 5.0) is higher than that for tweets without the national flag emoji (median: 4.0). For
Republicans also, the result is significant (Mood’s median test; p< .001).
The results indicate that in Germany, a tweet with a national flag is associated with higher
engagement on Twitter as compared to tweet without the national flag. In the United States, it is
associated with a slight increase in engagement for Republicans and a slight decrease for Democrats.
Table 4. Results for Research Question 3, Most Similar Words With Their Cosine Similarity to the Emoji.
All Parties (2,342) Democrats (718) Republicans (1,624)
1(0.7) (0.76) (0.78)
2(0.7) (0.73) (0.75)
3(0.68) (0.72) (0.75)
4(0.66) (0.72) (0.72)
5(0.65) (0.71) (0.72)
6(0.65) (0.69) #usa (0.71)
7 #maga (0.64) (0.69) (0.71)
8(0.62) #mapoli (0.67) #maga (0.7)
9(0.61) #unionstrong (0.67) (0.68)
10 #votebluewi (0.61) #nelsonsneighbors (0.67) (0.66)
Note. Independent word2vec models were trained from all tweets of both parties and an additional model for all tweets from
United States. For each party, count of tweets with United States is shown next to the party name.
14 Social Science Computer Review XX(X)
Figure 5. Panel A (left): Median retweet count for all tweets, tweets with an emoji, tweets with emoji, and tweets with emoji for Germany
parties. Panel B (right): Similar figure for U.S. parties. Note: Please note that in case of Germany, Alternative for Germany is on a different scale
than the rest of the parties.
Discussion and Limitations
In the article, we study the role of national flag emoji in online political communication. We
calculate the prevalence of national flag emoji in the online communication, study the differences
in its usage across parties, explore the impact of external events on its usage, analyze the meaning of
the national flag emoji for various political parties, and measure the impact of the national flag emoji
on engagement with the audience.
We find that national flag emoji are among the most frequently used emoji in online political
communication in both Germany and United States. While a likely explanation for this is the
ubiquitous use of national flag in the United States, the results are noteworthy for Germany
since use of the national flag as a national symbol is highly contested (Dohmen et al., 2006;
Heinrich, 2003; Stehle & Weber, 2013). It is also interesting to note that for both countries, flag
emoji are more popular than facial emoji (such as ,,and ) which are often reported to
be the most popular in online communication (Hu et al., 2017; Miller Hillberg et al., 2018;
Na’aman et al., 2017).
For evaluating the impact of external events, we examined the flag emoji usage on days of
national importance such as Independence Day, Veterans Day and Memorial Day in United States,
and Day of Unity in Germany. In both countries, we observed increase in percentage of tweets with
national flag on days of national importance. Flags are a common phenomenon in sporting events,
and national events and that seems to impact their online usage as well. While comparing Germany
and the United States, we observe two notable differences. Firstly, in Germany, both national and
sports events result in similar increase in the national flag emoji usage, while in United States, a
striking increase can be observed for national events but not for sports events. Secondly, in United
States, we see up to 5-fold increase in percentages of tweets with national flags on days of external
events, while only 2-fold increase is the case of Germany. Germany and United States have different
relationship to the national identity and national symbols and probably this leads to different impact
of the external events on national flag usage.
An interesting observation to emerge from the data was in the context of meaning of the flag. In
the face of the highly conflicted relationship of Germans with national symbols and a proven
reluctance of politicians to appear as literal or symbolic flag wavers, the question arises: Has
Unicode given a new meaning to the German flag by including it as an emoji? We found that in
Germany, parties use the flag emoji in the context of their policy stances. In a way, they are thus
using the flag not as a symbol to align themselves with traditional values of national identity
automatically associated with the flag. Instead, they appear to attempt to redefine the flag as being
aligned with their policy goals. Thus, they use the flag as a symbol of national unity which is then
given meaning to by the policy content they are proposing. The national flag can thus not automat-
ically be read as a signifier for tradionalist, right-wing, or conservative values of national identity.
Over time, this more relaxed attitude toward the symbol of the national flag might also translate into
uses of the actual physical manifestation of the flag.
In the United States, national flags appear more strongly in the context of electoral competition.
In both countries, we also observe the national flag is also used in the similar context as the country
name (#usa and #deutschland). Overall, we found different uses of the national flag based upon
specific context and political leaning.
The most striking result to emerge from the data is that in Germany, tweets with are associated
with significantly higher engagement for all political parties. For AfD, Gru
¨ne, and FDP, we observed
more than 5-fold increase in engagement for tweets with German flag, which is remarkably higher
when compared to political parties from the United States. Our results indicate that even though the
use of the national flag in Germany as a national symbol is highly contested, national flags in
Germany remain a powerful symbol when it comes to interaction with online content.
16 Social Science Computer Review XX(X)
In the United States, we observe that tweets with national flag are associated with a slight
increase in engagement for Republican party and slight decrease for Democratic party. This
connects to Kalmoe and Gross (2016) who found that the American flag has a pro-Republican
effect on the people identifying as Republicans but offers no advantage to the Democrats. Our
study finds a similar pattern for the Republicans but also observes slight disadvantage for the
Although we have found that in Germany parties across the political spectrum were using the
national flag emoji successfully to redefine its meaning and to foster engagement with their content,
we also found it a powerful symbol on the political far right. In fact, political parties on the far right
of the political spectrum use flag emoji with higher frequency and usually for these parties, tweets
with flag emoji are associated with higher engagement. These two findings taken together mean that
political parties on right and especially parties endorsing nationalistic thought might overall benefit
most from the introduction of flag emoji in the Unicode character set. These findings point to
important challenges for Unicode consortium and the designer of emoji, who might have unseeingly
skewed online communication in favor of some parties. In the recent years, right-wing nationalism is
on rise globally (Duara, 2018; Snyder, 2019; “The right-wing nationalists shaking up Europe,”
2019). Flag emoji provide an easy to use tool to represent the nation and through it express
nationalistic thought, in a media through which it is easy to reach masses. This research has thrown
up many questions in need of further investigation. Further work needs to be done to establish the
effectiveness of flag emoji in expressing nationalistic ideology and its role in increasing nationalism
around the globe and the successful contestation of these ideas through uses of flags across the
political spectrum.
There are various factors that influence engagement with a tweet. Some of these factors are
related to the authors’ timing, while others are related to the context. In this article, we do not
investigate the effect of these factors on engagement with a tweet. For example, while it is unlikely,
it is possible that all the flag tweets containing national flag might be written in only acclaimed
contexts (e.g., sports or foreign affairs), and people could be responding to the context rather than the
flag emoji itself. Difference in context might explain why the effect of the flag on engagement in the
United States is smaller—because it might be used in a wider range but often more mundane of
contexts as compared to Germany. In this study, we do not investigate this factor and cannot
comment on its impact the engagement with a tweet.
We started out to determine the role of national flag emoji in political communication in Ger-
many and the United States. One of the more significant findings to emerge from this study is that
presence of flag emoji is associated with notable higher engagement with tweets in Germany. The
second major finding is that meaning of national flag emoji depends on political and cultural context.
The results also show that the usage of flag emoji greatly varies between parties and in general
political parties on the right use of flag emoji more often than the parties on the left. Furthermore,
flag emoji usage in online communication is associated with external events of national importance.
These findings demonstrate that as symbolic interactionists would expect, the uses and meaning of
national flags as a symbol vary across cultural and political contexts and have different conse-
quences as expressed in stimulating audience reactions. Here the divergent contextual uses of
national flags in tweets of political actors depending on their ideological leaning are especially
interesting and offer promising perspectives for future research.
Authors’ Note
We would like to thank Stefanie Walter for her input and feedback on this research. We are also grateful to the
two anonymous reviewers for their comments that helped us improve the manuscript. We thank our colleagues
Daniel Diethei, Gian-Luca Savino for the discussion and Daria Vladimirovna Soroko for the help with the
Kariryaa et al. 17
Data Availability
This study is based upon analysis of tweets from Twitter. Due to Twitter’s terms of use, we can’t release the
complete data set, but we release the list of tweet IDs, for noncommercial research purposes, that we considered
in this study, along with their relevant attributes such as party affiliation and count of emoji. This information
is sufficient for all analysis except for training a word2vec model. With the tweet IDs, a replica of the data set
can be created through the Twitter API to train the word2vec model. The partial data set can be obtained from
the first author or accessed via the open-source repository (
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or
publication of this article.
The authors disclosed receipt of the following financial support for the research and/or authorship of this article:
This research was supported in part by the Volkswagen Foundation through a Lichtenberg Professorship.
Software Information
We used custom code for this study. The code is written in Python3 using Jupyter notebooks, Pandas, Numpy
and Scipy packages. The code can be obtained from the first author or accessed via the open-source repository
Barbieri, F., Kruszewski, G., Ronzano, F., & Saggion, H. (2016, October). How cosmopolitan are emojis?
Exploring emojis usage and meaning over different languages with distributional semantics. Proceedings of
the 2016 ACM on multimedia conference (pp. 531–535).
Barbieri, F., Ronzano, F., & Saggion, H. (2016). What does this emoji mean? A vector space skip-gram model
for Twitter emojis.
Becker, H. S. (1988). Herbert Blumer’s conceptual impact. Symbolic Interaction,11(1), 13–21.
Becker, J. C., Enders-Comberg, A., Wagner, U., Christ, O., & Butz, D. A. (2011). Beware of national symbols.
Social Psychology,43(1), 3–6.
Bernstein, R. (2006, June 18). In World cup surprise, flags fly with German pride. The New York Times. https://
Billig, M. (1995). Banal nationalism. Sage.
Blumer, H. (1966). Sociological implications of the thought of George Herbert mead. American Journal of
Sociology,71(5), 535–544.
Boykoff, J., & Carrington, B. (2019). Sporting dissent: Colin Kaepernick, NFL activism, and media framing
contests. International Review for the Sociology of Sport.
Butz, D. A., Plant, E. A., & Doerr, C. E. (2007). Liberty and justice for all? Implications of exposure to the US
flag for intergroup relations. Personality and Social Psychology Bulletin,33(3), 396–408.
Carter, T. J., Ferguson, M. J., & Hassin, R. R. (2011). A single exposure to the American flag shifts support
toward Republicanism up to 8 months later. Psychological Science,22(8), 1011–1018.
Chan, E. Y. (2017). Exposure to the American flag polarizes democratic-republican ideologies. British Journal
of Social Psychology,56(4), 809–818.
Chase, J., & Goldenberg, R. (2019, October 28). AfD: What you need to know about Germany’s far-right party
[Newspaper]. DW.COM.
18 Social Science Computer Review XX(X)
Dhir, A., Khalil, A., Kaur, P., & Rajala, R. (2019). Rationale for “Liking” on social networking sites. Social
Science Computer Review,37(4), 529–550.
Dohmen, T. J., Falk, A., Huffman, D., & Sunde, U. (2006). Seemingly irrelevant events affect economic
perceptions and expectations: The FIFA World Cup 2006 as a natural experiment (IZA Discussion Papers
2275). Institute of Labor Economics (IZA).
Duara, P. (2018, August 31). Nationalism and development in Asia. UNU-WIDER.
Durkheim, E. (1953). Value judgements and judgements of reality (D. F. Pockock, Trans). Sociology and
philosophy (pp. 80–97). The Free Press. (Original work published 1911)
Ehrlinger, J., Plant, E. A., Eibach, R. P., Columb, C. J., Goplen, J. L., Kunstman, J. W., & Butz, D. A. (2011).
How exposure to the confederate flag affects willingness to vote for Barack Obama. Political Psychology,
32(1), 131–146.
Eisner, B., Rockta¨schel, T., Augenstein, I., Boˇsnjak, M., & Riedel, S. (2016). Emoji2vec: Learning emoji
representations from their description. ArXiv Preprint ArXiv:1609.08359.
Elias, N. (1992). Studien u¨ ber die Deutschen: Machtka¨mpfe und Habitusentwicklung im 19. und 20. Jahrhun-
dert (Vol. 1008). Suhrkamp.
Evans, H. K., Brown, K. J., & Wimberly, T. (2018). “Delete Your Account”: The 2016 presidential race on
Twitter. Social Science Computer Review,36(4), 500–508.
Ferguson, M. J., & Hassin, R. R. (2007). On the automatic association between America and aggression for
news watchers. Personality and Social Psychology Bulletin,33(12), 1632–1647.
Harr´e, R. (2002). Material objects in social worlds. Theory,Culture & Society,19(5–6), 23–33.
Hassin, R. R., Ferguson, M. J., Shidlovski, D., & Gross, T. (2007). Subliminal exposure to national flags affects
political thought and behavior. Proceedings of the National Academy of Sciences,104(50), 19757–19761.
Hayes, R. A., Carr, C. T., & Wohn, D. Y. (2016). One click, many meanings: Interpreting paralinguistic digital
affordances in social media. Journal of Broadcasting & Electronic Media,60(1), 171–187.
Heinrich, A. (2003). The 1954 Soccer World Cup and the Federal Republic of Germany’s self-discovery.
American Behavioral Scientist,46(11), 1491–1505.
Hu, T., Guo, H., Sun, H., Nguyen, T. T., & Luo, J. (2017). Spice up your chat: The intentions and sentiment
effects of using emoji. ArXiv Preprint ArXiv:1703.02860.
Huntington, S. P. (2004). Who are we? The challenges to America’s national identity. Simon & Schuster.
Janssen, J., & Verheggen, T. (1997). The double center of gravity in Durkheim’s symbol theory: Bringing the
symbolism of the body back in. Sociological Theory,15(3), 294–306.
Jaskulowski, K. (2016). The magic of the national flag. Ethnic and Racial Studies,39(4), 557–573.
Jungherr, A. (2015). Analyzing political communication with digital trace data. Springer.
Jungherr, A. (2016). Twitter use in election campaigns: A systematic literature review. Journal of Information
Technology & Politics,13(1), 72–91.
Jungherr, A., Rivero, G., & Gayo-Avello, D. (2020). Retooling politics: How digital media are shaping
democracy. Cambridge University Press.
Kalmoe, N. P., & Gross, K. (2016). Cueing patriotism, prejudice, and partisanship in the age of Obama:
Experimental tests of U.S. flag imagery effects in presidential elections. Political Psychology,37(6),
Kemmelmeier, M., & Winter, D. G. (2008). Sowing patriotism, but reaping nationalism? Consequences of
exposure to the American flag. Political Psychology,29(6), 859–879.
Khan, M. L. (2017). Social media engagement: What motivates user participation and consumption on You-
Tube? Computers in Human Behavior,66, 236–247.
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature,521(7553), 436.
Kariryaa et al. 19
Lee, S.-Y., Hansen, S. S., & Lee, J. K. (2016). What makes us click “like” on Facebook? Examining
psychological, technological, and motivational factors on virtual endorsement. Computer Communica-
McLaughlin, B., & Velez, J. A. (2019). Imagined politics: How different media platforms transport citizens into
political narratives. Social Science Computer Review,37(1), 22–37.
Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013). Efficient estimation of word representations in vector
space. ArXiv Preprint ArXiv:1301.3781.
Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. (2013). Distributed representations of
words and phrases and their compositionality. Advances in Neural Information Processing Systems
(pp. 3111–3119).
Miller Hillberg, H., Levonian, Z., Kluver, D., Terveen, L., & Hecht, B. (2018). What I see is what you don’t get:
The effects of (not) seeing emoji rendering differences across platforms. Proceedings of the ACM on
Human-Computer Interaction,2(CSCW), 124.
Mood, A. M. (1950). Introduction to the theory of statistics. McGraw-Hill.
Mueller, J. E. (1970). Presidential popularity from Truman to Johnson 1. American Political Science Review,
64(1), 18–34.
Na’aman, N., Provenza, H., & Montoya, O. (2017). Varying linguistic purposes of emoji in (Twitter) context.
Proceedings of ACL 2017, Student Research Workshop, 136–141.
Pennington, J., Socher, R., & Manning, C. (2014, October). Glove: Global vectors for word representation.
Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP)
(pp. 1532–1543).
The Political Compass. (n.d.). Retrieved February 3, 2020, from
Rehurek, R., & Sojka, P. (2010). Software framework for topic modelling with large corpora. In Proceedings of
the LREC 2010 Workshop on New Challenges for NLP Frameworks.
Roßmann, R. (2016, December 14). Wenn Bilder sto¨ ren [When images interfere]. Su¨ ddeutsche Zeitung. https://
Sahlgren, M. (2005). An introduction to random indexing. In Methods and applications of semantic indexing
workshop at the 7th international conference on terminology and knowledge engineering. http://www.diva-
Schatz, R. T., & Lavine, H. (2007). Waving the flag: National symbolism, social identity, and political
engagement. Political Psychology,28(3), 329–355.
Schill, D. (2012). The visual image and the political image: A review of visual communication research in the
field of political communication. Review of Communication,12(2), 118–142.
Schroeder, R. (2018). Social theory after the internet: Media, technology, and globalization. UCL Press.
SciPy Version 1.2.1 Reference Guide. (2019, February 8).
Scissors, L., Burke, M., & Wengrovitz, S. (2016, February 27–March 02). What’s in a Like? Attitudes and
behaviors around receiving Likes on Facebook. Proceedings of the 19th ACM conference on computer-
supported cooperative work and social computing (pp. 1501–1510).
Smith, A. D. (1991). National identity. University of Nevada Press.
Smith, W. (1975). Flags through the ages and across the world. McGraw-Hill.
Snow, D. A. (2001). Extending and broadening Blumer’s conceptualization of symbolic interactionism. Sym-
bolic Interaction,24(3), 367–377. JSTOR.
Snyder, J. (2019, February 25). The Broken Bargain: How nationalism came back. https://www.foreignaffairs.
20 Social Science Computer Review XX(X)
Spiegel Online. (2006, April 21). Germany’s Patriotism Problem: Just Don’t Fly the Flag.
Spierings, N., Jacobs, K., & Linders, N. (2019). Keeping an eye on the people: Who has access to MPs on
Twitter? Social Science Computer Review,37(2), 160–177.
Stehle, M., & Weber, B. M. (2013). German soccer, the 2010 World Cup, and multicultural belonging. German
Studies Review,36(1), 103–124.
The right-wing nationalists shaking up Europe. (2019, November 13). BBC News.
Unicode 6.0.0. (2010, October 11).
User timelines. (2016, June 19).
Wohn, D. Y., Carr, C. T., & Hayes, R. A. (2016). How affective is a “Like”? The effect of paralinguistic digital
affordances on perceived social support. Cyberpsychology, Behavior, and Social Networking,19(9),
Author Biographies
Ankit Kariryaa is a PhD candidate in Human Computer Interaction research group at the University of Bremen
in Germany. He received a master’s degree in intelligent systems from Bielefeld University in Germany. His
current research focuses on volunteered geographic information and political communication on social media.
His research has been presented at reputable conferences such as ACM Conference on Human Factors in
Computing Systems (CHI).
Simon Rund´
eis a bachelor student in the University of Bremen. His bachelor thesis focuses on hate speech and
emoji usage in online political communication.
Hendrik Heuer is a PhD candidate in human-computer interaction and machine learning and a research
assistant at the Information Management Research Group at the University of Bremen in Germany. He is also
a member of the Institute for Information Management Bremen GmbH (ifib) and the Centre for Media,
Communication and Information Research (ZeMKI). He received graduate degrees in human-computer inter-
action and design from the Royal Institute of Technology, Stockholm, Sweden, and Aalto University, Helsinki,
Finland. His current research is focused on the user experience of machine learning, especially trust and
algorithmic transparency.
Andreas Jungherr is an assistant professor at the Department of Politics and Public Administration at the
University of Konstanz. His work focuses on political communication and the effects of digital technology on
politics. He approaches these questions through the lens of computational social science and digital methods.
He is the author of Analyzing Political Communication With Digital Trace Data: The Role of Twitter Messages
in Social Science Research (Springer, 2015) and Retooling Politics: How Digital Media Is Shaping Democracy
(with Gonzalo Rivero and Daniel Gayo-Avello, Cambridge University Press, 2020).
Johannes Scho
¨ning is a Lichtenberg Professor and Professor of Human-Computer Interaction (HCI) at the
University of Bremen in Germany. In addition, he is the codirector of the Bremen Spatial Cognition Center
(BSCC) and member of the Technologie-Zentrum Informatik und Informationstechnik (TZI). His research
interests lie at the intersection between (HCI), geographic information science, and ubiquitous interface tech-
nologies. His work has been published in numerous reputable international journals and conferences.
Kariryaa et al. 21
... Despite growing attention to non-verbal cues, political parties' visual communication strategies have remained largely unexplored. To be sure, several studies examine political parties' election posters and online communication (Schill, 2012), paying attention to the display of national (Kariryaa et al., 2020) and European flags (Dumitrescu and Popa, 2016). Moreover, recent scholarship has leveraged the concept of branding to study political parties' choice of candidates, political platforms, and communication strategies (Nielsen andLarsen, 2014, Needham andSmith, 2015;Grimmer and Grube, 2017;Pich and Newman, 2020). ...
Full-text available
Despite growing attention to electoral brands, political scientists have largely remained “color blind,” neglecting parties’ chromatic choices. Moreover, scholars dedicated limited attention to how party organizations converge on the use of similar structures and communication strategies, thereby engaging in a process of institutional isomorphism. We seek to simultaneously fill both gaps by examining the logos of more than 300 parties in 35 democracies during the latest political elections. Our findings show that a strong relationship exists between ideology and the use of certain color hues: left-wing party logos mainly display hues at the red end of the color spectrum, while blue hues prevail among right-wing parties. Likeminded parties’ chromatic isomorphism, however, is moderated by country and party-specific factors. Notably, the correlation between color hue and ideology is stronger in Western Europe and among older parties.
... Political communication strategy is not only limited to political marketing activities but also in local government, it is related to the exchange of messages with the community that is sustainable in decision making (Kariryaa et al., 2020). ...
... Az emócióelemzés képi ága azokon a leírásokon alapul, amelyeket Ekman és Friesen (1969) adott arról, hogyan jelenik meg a hat alapérzelem az emberi arcon. Manapság pedig az emotikonok és az emojik politikai használatának vizsgálata dívik (Bódi-Veszelszki, 2006;Kariryaa et al., 2020). ...
... • User's association with national symbol [133]: In relation to political interests, we study the usage and meaning of national flags, the most influential and widely used national symbols, for politicians and political parties, and analyze their association with user engagement. ...
Full-text available
User-generated content (UGC) has been on the rise with the emergence of Web 2.0 and it is the driving force behind prominent online platforms such as Wikipedia, OpenStreetMap, Facebook, and Twitter. UGC has led to numerous innovations in academia and industry and has transformed our world in multiple ways. While the positive impact of UGC is abundant, there is limited research on its negative impacts. In this thesis, we study the impact of the UGC from the perspective of the users, who are both its creators and consumers. This thesis has four main contributions, detailed below. The first contribution improves our understanding of the impact of the UGC on the geo-privacy of the user. We investigate the accuracy of localness methods used for the categorization of UGC at the city, county, and state scale. Through a user study, we show that some methods can assess the location of a content producer with very high accuracy. Thus, the research establishes a standard for localness methods as well as highlights the impact of UGC on the geo-privacy of users. The second contribution improves our understanding of the impact of national symbols in UGC on political communication. Through a study of tweets by politicians and political parties in Germany and the USA, we analyze the role of flag emoji, looking into their usage, meaning, and association with the audience engagement. The results show that flags remain an influential symbol in online communication and they are associated with significantly higher engagement for most political parties, which helps improve to our understanding of UGC in politics. In the third contribution, we present a tool for limiting the impact of UGC on the online privacy of the users. UGC producers often reveal personal information about themselves and online passwords commonly contain personal information. Current password meters do not consider personal information and, therefore, their users are susceptible to guessing attacks. We present the MoiPrivacy password meter, that extends a neural network- and heuristic-based approach and considers a user's personal information while calculating the password strength and feedback. Through a user study, we find that MoiPrivacy significantly limits the inclusion of personal information in passwords, thus limiting the negative impacts of UGC on online security. In the fourth contribution, we present a deep learning-based approach for identifying individual trees in sub-meter satellite imagery at a very large scale. This approach was used to map the crown size of each tree >3 sq. m in a land area spanning 1.3 million sq. km in the West African Sahara, Sahel, and sub-humid zone. We detected over 1.8 billion individual trees, or 13.4 trees per ha, with a median crown size of 12 sq. m along a rainfall gradient from 0 to 1000 mm. Our assessment suggests a way to monitor trees outside forests globally and to explore their role in mitigating degradation, climate change, and poverty. While content generation is associated with some adverse impacts on the user, it also offers an opportunity for large scale UGC-based citizen science platforms. In the future, large scale citizen platforms might be crucial for tackling global challenges including climate change and shrinking biodiversity and the presented approach could be crucial for bootstrapping such platforms.
... This makes understanding the user perspective of ML-based curation an important and timely research area. Especially since research by Kariryaa et al. (2020) showed that even the usage of particular emoji is affected by its usage context on social media platforms. Therefore, understanding how ML-based curation systems shape the experience of users is an important concern. ...
Full-text available
Users are increasingly interacting with machine learning (ML)-based curation systems. YouTube and Facebook, two of the most visited websites worldwide, utilize such systems to curate content for billions of users. Contemporary challenges such as fake news, filter bubbles, and biased predictions make the understanding of ML-based curation systems an important and timely concern. Despite their political, social, and cultural importance, practitioners' framing of machine learning and users' understanding of ML-based curation systems have not been investigated systematically. This is problematic since machine learning - as a novel programming paradigm in which a mapping between input and output is inferred from data - poses a variety of open research questions regarding users' understanding. The first part of this thesis provides the first in-depth investigation of ML-based curation systems as socio-technical systems. The second part of the thesis contributes recommendations on how ML-based curation systems can and should be explained and audited. The first part analyses practitioners' framing of ML by examining how the term machine learning, ML applications, and ML algorithms are framed in tutorials. The thesis also investigates the beliefs that users have about YouTube and introduces a user belief framework of ML-based curation systems. Furthermore, it demonstrates how limited users' capabilities for providing input data for ML-based curation systems are. The second part evaluates different explanations of ML-based systems. This evaluation uncovered an explanatory gap between what is available to explain ML-based curation systems and what users need to understand such systems. Informed by this explanatory gap, the second part of this thesis demonstrates that audits of ML systems can be an important alternative to explanations. This demonstration of audits also uncovers a popularity bias enacted by YouTube's ML-based curation system. Based on these findings, the thesis recommends performing audits to ensure that ML-based systems act in the public's interest.
Retrospective rhetoric is considered to be one of the most influential political master frames used in response to the multiple crises in Europe. However, there are significant knowledge gaps in this area. First, there is no comprehensive analytical tool for studying the display of nostalgia in politics. Second, there is limited knowledge regarding the role of social media in disseminating nostalgic narratives. Third, there is a lack of data on citizens’ responses to nostalgia during election campaigns. This article addresses these gaps by identifying the patterns of political nostalgia in the Facebook communications of the Hungarian political forces running for seats in the 2019 European parliamentary election. In addition to a qualitative content analysis of the nostalgic posts, the paper investigates their reception by the social media users who reacted to the messages. The data was gathered during the seven-week official campaign period until the election on May 26. The study confirms that right-wing, nationalist politicians expressed nostalgia more often than left-leaning leaders, and the analysis also shows that nostalgic messages are not very efficient in terms of user engagement and the numbers of emoji responses. Our results therefore mitigate the concerns about the influence of political nostalgia in Hungary.
Full-text available
Twitter is credited for allowing ordinary citizens to communicate with politicians directly. Yet few studies show who has access to politicians and whom politicians engage with, particularly outside campaign times. Here, we analyze the connection between the public and members of parliament (MPs) on Twitter in the Netherlands in-between elections in 2016. We examine over 60,000 accounts that MPs themselves befriended or that @-mentioned MPs. This shows that many lay citizens contact MPs via Twitter, yet MPs respond more to elite accounts (media, other politicians, organized interests,…), populist MPs are @-mentioned most but seem least interested in connecting and engaging with “the” people, and top MPs draw more attention but hardly engage—backbenchers are less contacted but engage more.
Full-text available
This article examines political transportation—the construction of mental models that represent a political world and result in the absorption and positioning of oneself within the constructed world. Specifically, we propose that when citizens become immersed in the political narrative crafted by a politician, they become more likely to see the political world as personally relevant and, subsequently, become more committed to supporting that candidate. Further, the degree of political immersion should depend upon which media platform a campaign message is delivered through. These expectations were tested using an experiment where partisans were exposed to a campaign message delivered in the form of a television ad, a political e-mail, or a series of tweets. Results demonstrate that Twitter was the least likely to lead to political immersion. Further, results provide support for our theoretical model, where there is an indirect effect of campaign messages on political attitudes and behavior through immersion and perceived personal relevance. Taken together, this study demonstrates the utility of applying the concept of narrative transportation to politics.
Full-text available
While there has been a growth in the number of published studies about how candidates for the U.S. House and Senate use Twitter, candidates for president have been largely ignored. In this article, we examine the way the two 2016 presidential candidates communicated on Twitter. Using a content analysis of all tweets sent from Hillary Clinton and Donald Trump’s accounts from July 1 to Election Day, we explore whether the two candidates used this social network in the same ways, stressed similar policy issues, and were equally likely to “go negative” online.
Cambridge Core - Computing and Society - Retooling Politics - by Andreas Jungherr
In August 2016, Colin Kaepernick, a quarterback on the San Francisco 49ers of the National Football League, sat in protest during the national anthem. He made it clear that his stance was a political statement against racialized oppression and police brutality carried out against people of color in the USA, and in doing so he became a polarizing cultural figure. This article uses content analysis to examine newspaper coverage of the emergence and evolution of Kaepernick’s political activism over a critical two-year period, from August 2016 through August 2018. First, we identify the dominant frames that media adopted when covering his protests and their aftermath. Then we examine who got to speak in the articles: which sources tended to predominate and how did this inflect the principal frames? Finally, we explore whether Kaepernick’s activism deepened discussions over police brutality against African Americans or racial inequality in the USA. We conclude that the print media’s coverage was largely favorable to Kaepernick even as much of the coverage reduced the protest from being about racial injustice in the USA to being framed, reductively, as an “anthem protest.”
Emoji are popular in digital communication, but they are rendered differently on different viewing platforms (e.g., iOS, Android). It is unknown how many people are aware that emoji have multiple renderings, or whether they would change their emoji-bearing messages if they could see how these messages render on recipients' devices. We developed software to expose the multi-rendering nature of emoji and explored whether this increased visibility would affect how people communicate with emoji. Through a survey of 710 Twitter users who recently posted an emoji-bearing tweet, we found that at least 25% of respondents were unaware that the emoji they posted could appear differently to their followers. Additionally, after being shown how one of their tweets rendered across platforms, 20% of respondents reported that they would have edited or not sent the tweet. These statistics reflect millions of potentially regretful tweets shared per day because people cannot see emoji rendering differences across platforms. Our results motivate the development of tools that increase the visibility of emoji rendering differences across platforms, and we contribute our cross-platform emoji rendering software to facilitate this effort.
The “like” feature is popularly utilized by online social media users for different reasons including socializing, giving feedback and giving or seeking attention as well as for pure affection. The “like” function is a gamified element of social networking sites used billions of times per day. Despite its widespread use in the social media space, little is known about the different factors that influence Facebook users’ “like” continuation intention or the game mechanics of “like.” To address this relevant issue, a cross-sectional survey was administered with 728 adolescent Facebook users (12–18 years old). This study utilized the theory of planned behavior to investigate the role of attitude (hedonic motivation, reciprocal benefit, and social presence), subjective norms (primary influence and secondary influence), and perceived behavioral control (self-efficacy and habit) in influencing the continuation intention of “like” as well as the influence of self-efficacy and habit on the game mechanics of “like.” This investigation addresses the urgent need to understand better the postadoption issues as well as the intentions to use specific features of social media. The results suggest that social presence, primary and secondary influence, self-efficacy, and habit significantly predicted Facebook “like” continuation intention. Furthermore, self-efficacy and habit significantly predicted the game mechanics of “like.” Different theoretical and practical implications of the study are presented and discussed in light of prior information systems literature.
Conference Paper
We propose two novel model architectures for computing continuous vector representations of words from very large data sets. The quality of these representations is measured in a word similarity task, and the results are compared to the previously best performing techniques based on different types of neural networks. We observe large improvements in accuracy at much lower computational cost, i.e. it takes less than a day to learn high quality word vectors from a 1.6 billion words data set. Furthermore, we show that these vectors provide state-of-the-art performance on our test set for measuring syntactic and semantic word similarities.