: Emoji, Language Games and Political

Abstract

Are emoji political? In an increasing body of research, emoji have variably been viewed as emotional data or personality identi昀椀ers. However, little attention has been paid to the social and political import of emoji. Using a dataset of politically active Twitter users in Poland, including 334 members of parliament and their 1,288,950 followers, we ask whether emoji are used for political self-representation, and discuss the implications for political identity formation and mobilisation online. Adapting a new method of ideal point estimation, we identify patterns in the employment of emoji in user Twitter bios across a latent political space computed from a Twitter following network. We 昀椀nd that emoji are used as stand-ins for o昀툀ine political symbols such as 끹믤 , 꺈� and ✝️ . Additionally, we 昀椀nd that the use of emoji without recognisable political meaning, such as✌️ , ,

-

and is contingent on a users estimated political ideal point. Users on the le昀琀 are likelier to employ ✌️ and , while those on the right are likelier to employ

and . Using Ludwig Wittgenstein’s theory of language games, we argue that this points to the use of emoji for communication of both political values and a昀ect, and to the development of a new political language game of emoji. sociolinguistics, social media, networks, Polish politics.

1. Introduction

If you opened your Twitter app in October 2021, you might have seen a 昀氀ush of green or blue hearts on your timeline, as users added them to their bios in response to new anti-abortion legislation in Poland. Blue hearts denoted agreement, while green hearts indicated opposition a reference to the Argentinian ‘Green Wave’ abortion rights movement 4[ , 20 ]. Does the use of such emoji signal a new vernacular in online political communication?

CEUR Workshop Proceedings

We adapt a new method to estimate the political ideal points of Twitter accounts, based on their position in a latent network [ 5, 46 ], and thus identify patterns in the employment of emoji in user bios. As expected, use of끹믤 , 꺈� and ✝️ emoji appears politically contingent. More intriguingly however, so does the use of emoji without a direct political meaning, such a✌️s , , , . Users on the le昀琀 are likelier to employ ✌️ and , while those on the right are likelier to employ , . Using the theory of Ludwig Wittgenstein, the Austrian philosopher of language, we argue that this points to a new political language game.

The advent of emoji has been described as a “veritable paradigm shi昀琀” in human communication [12, p.VI]. While existing research has found that emoji use re昀氀ects patterns of collective and negotiated meaning, little attention has been given to the implications for collective identity formation. Yet this is signi昀椀cant, since emoji are both more reductive and more 昀氀exible than natural language. Although one emoji can have di昀erent meanings, natural language users can only pick ready-made emoji o昀 the shelf: they cannot adjust or edit them.

Linguistic research has focused on the discursive use of emoji for a昀ect [11], attentiongrabbing [ 19 ] and advertising [ 13 ]. More recent work has looked at their use in signalling political association [ 37 ] and social a昀케liation [ 3]. While much of this work has focused on emoji use in messaging and discussion posts, however, the increasing use of emoji for online self-representation has yet to be studied. We look speci昀椀cally at how this applies to Twitter bios.

Twitter bios ask for a self-representation “in 160 characters”. Given this space limitation, they require a concentrated form of communication 6[ 4, 24 ], and o昀er the online equivalent of a 昀椀rst impression [ 51, pp. 2]).

Pathak, Madani and Joseph describe a unique bio vernacular, which involves users listing multiple social and personal identities through single nouns delimited by a “.” or a “|”43[]. For example, former US President Barack Obama’s Twitter bio reads: “Dad, husband, President, citizen”. Meanwhile, a recent four-year longitudinal study observed users adding political words to their Twitter bios at a higher rate than any other word category 5[ 1 ].

To date, however, very little academic attention has been paid to the political content of Twitter bios. This paper will 昀椀ll the gap by analysing the use of emoji to express political values and self-representation on Twitter bios, and the implications of this for political mobilisation and polarisation [ 46, 5 ]. It is our hope that this will lay the groundwork for the use of emoji in expanding ideal point estimation.

2. Theoretical Background

2.1. A Language Games of Emoji

In 1953, Ludwig Wittgenstein’s Philosophical Investigations was published posthumously, outlining his observation that our experience of the world is shaped by language. In contrast to the dominant conception at the time – that language was a tool based on external rules - he described how words gain and regain meanings through use, calling this thelanguage-game. Wittgenstein’s revolutionary insight was that language is not simply a logic-based tool akin to mathematics, existing separately from people. Instead, words have no 昀椀xed meanings, but are given context-speci昀椀c meaning through action. How we experience actions and their meanings is de昀椀ned by the words we use about them.

Hence the idea of language games: “[speaking] language is part of an activity, a form of life” [62, pp. 15]. By forms of life, Wittgenstein means di昀erent groups and activities within our lives, around which we create unique language games which help give meaning to such activities. We talk about political processes, for example, di昀erently than about our favourite music; and we may talk about both di昀erently in di昀erent social contexts. Each distinct form of life we engage in develops a unique language game around it - and through these language games we construct the form’s meaning.

2.2. Sociolinguistics

Wittgenstein’s insights - that forms of life are shaped by language games, and that language itself is shaped by how we use it - gave way to further research in philosophy, sociology and linguistics. In linguistics, this 昀椀eld has become known as sociolinguistics. It looks at how social processes inform linguistic change [31], and how language use shapes social dynamics [ 22 ]. The latter research strand, sometimes called discourse analysis, has featured work analysing emoji usage, focusing primarily on emoji as a tool for emotional expression30[ , 17, 18, 23, 16, 42, 59 ]. This extends beyond traditional face emoji, whose emotional import is apparent. Riordan, for example, stresses how even object emoji, though having no direct emotional a昀ect, still perform “emotion work” [ 47, 26 ].

A common theme of the work so far concerns the potential for misunderstanding or interpretation divergence [ 42, 59, 17 ]. Approaching emoji as primarily “emotional data”, however, risks underplaying the social and political import of emoji usage5[3, pp. 1]. Recent work in computational sociolinguistics has, for example, highlighted the 2015 inclusion of skin-tone emoji modi昀椀ers to all Apple operating systems to observe how social groups use emoji to express identity [50, 49]. Further work has analysed cultural di昀erences in the use and interpretation of emoji [ 23, 54 ], and observed how emoji are used as stand-ins for political symbols during online campaigns [ 2 ]. This work recasts the literature’s focus on emoji “miscommunication”, and points towards the interpretation of emoji as negotiated and embedded in social group4s1[, 25, pp. 1].

2.3. Symbolic Boundaries

Beyond considering how emoji use is tied to speci昀椀c social groups, little attention has been paid to whether it plays a role in constructing social group identities. Social Identity Theory (SIT) posits that our sense of identity, rooted in the social group we are part of, is constructed through a “human capacity - rooted in language - to know ‘who’s who’ and ‘what’s what’”2[7, pp. 6]. As with Wittgenstein’s language games, SIT suggests that social identity is a process not something we have, but something we do. Language is central to this process of identity construction, and o昀琀en plays a role through symbolic boundaries [ 32, 57, 6 ] - lines that “include and de昀椀ne” a social group [ 33, pp. 1]. Symbolic boundaries are constructed iteratively through use of symbols such as clothing, images and linguistic patterns to de昀椀ne, and rede昀椀ne, in-group and out-group [32]. This process is known as boundary work. Group-based linguistic variation in use and interpretation of emoji indicates that emoji use could play a role in boundary work’s online performance.

This is signi昀椀cant since group boundaries play an important role in collective mobilisation, strengthening shared de昀椀nitions of “us/them” and bolstering “feelings of similarity and group membership” as a prelude to action [15, 58, 40, pp. 232].1 Similarly, language games enable group mobilisation. If emoji are used to communicate political a昀케liations and develop collective political identities, this becomes relevant for understanding online political mobilisation and group polarisation.

Unlike symbolic boundaries, however, the theory of languages games suggests that linguistic patterns underpinning group identity are not symbols of that group, but actions distinguishing that group from another. Rather than using an emoji to signal group membership, the very use of emoji enables the group to exist. Consequently, while symbolic boundaries view the tie between language and group identity as something which must be “recognised by outsiders for an objecti昀椀ed collective identity to emerge” [32, pp. 170], language games theory suggests the tie can occur unconsciously and contribute to group identi昀椀cation even it not recognised by outsiders. This provides a helpful framework for understanding emoji use, and the unique ways in which emoji such as and might assist collective identity building without directly appealing to symbolic boundaries2.

We aim to extend the work on emoji beyond their use in “emotion work” and o昀er a new theoretical framework for viewing emoji as a foundation for digital language games.

3. Methods

3.1. Data & Methods

To investigate the questions outlined Section2 we selected the case of the political Twitter sphere in Poland. Twitter is broadly used in Poland, but Polish language tweets make up a small proportion of Twitter content, and mostly produced by Polish users. This enables our sample to capture a maximally broad section of the population, while avoiding spillover e昀ects introduced by users from multiple countries. This is important because we aim to estimate patterns of language use within a single political community.

To produce estimates of ideological positions of a su昀케ciently large Twitter population involved in political online debates, we selected the ideal point estimation method for social media data [ 5 ]. In particular for Twitter data and multidimensional political scales in European settings, this method, calledideological embedding[46], leverages political homophily in online social networks and political survey data to produce interpretable scales of positions for large numbers of users along dimensions of political issues and ideologies. This method requires us to select a seed group of political references users, typically members of parliament in a country [45]. We annotated 334 members of parliament (MPs) in Poland in October 2020, building on the Twitter Parliamentary Database [ 60 ], and belonging to 9 political parties. We then collected 1Dubbed entitativity within SIT [ 7 ] 2This also enables the use of emoji to “communicate via coded workarounds”, for example the famo us emoji used by the far-right to signal group membership only to ‘those in the know’. See Albury1[] for a description of emoji as “o昀-label” uses of digital media. their 1.288.950 Twitter followers, and – following Ramaciotti Morales, Cointet, and Zolotoochin [46] 昀椀ltered those that had less than 25 followers or that had published less than 100 tweets, so as to minimise the probability of including bots in our dataset. We also excluded followers that followed less than 3 of our 334 MPs, so as to assure that a following link conveys an ideological signal, and is not due to other reasons (e.g., some MPs are celebrities in the public sphere). This resulted in 71.884 followers in our dataset.

3.2. Estimating and Interpreting Ideal Points

Following the methodology described in4[6], we compute an approximation of a multidimensional ideal point estimation in our MP-follower network using the Correspondence Analysis of the adjacency table of this bipartite network [36] (with MPs in columns and followers in rows, with values 0 and 1 indicating whether an MPs is followed by each user). The result of this operation is a multidimensional latent position for each MP and each one of the 71.884 followers in our dataset. Positions in this space explain how followers follow MPs by proximity: the closest they are, the higher the probability of one following the other. To interpret to which political issues and dimensions these dimensions are related, we will use the Chapel Hill Expert Survey (CHES) data [ 28 ]. The CHES data contains party positions, as evaluated by experts, along 51 dimensions of political issues and ideologies; 5 of our 9 political parties are present in the CHES data. Fig. 1 shows the distribution of MPs and followers along the 昀椀rst two dimensions of this latent space, with MPs colored according to the 5 parties available in the CHES data to which they belong.

To provide an illustration of the amounts of users that use emoji in our dataset, Fig2. shows the amount of users in our dataset according to their position along the 昀椀rst latent dimension, together with the proportion of those that have a Twitter bio (i.e., a pro昀椀le description), and those that include emoji on them.

4. Results

In this section we outline the results of the study, presenting them in response to the three research questions.

4.1. RQ1: Which political issue clusters arise from the network of political accounts and their followers on Twitter?

To answer this question we leverage the CHES data described in the previous section as detailed in [46]. For a given dimension of our latent space, we compare party positions along this dimension with party positions in the CHES dimensions.

To do this comparison, we attribute MPs the position of their parties on the one hand, and consider positions in the latent space in the other, to then assess the degree to which CHES and latent are aligned via correlation.

We systematically run a Pearson correlation between party positioning according to our 椀昀rst dimension, and according to all CHES dimensions and retain those were correlation is 2 n o i s n e m i d t n e t a L

Latent dimension 1

| | ≥ 0.9 . The CHES dimensions highly correlated with the 昀椀rst latent dimension of our MPfollower Twitter network are (with correlation value and p-value): attitudes towards elites and institutions ( =0.99, ≤ 0.001), anti-islam rhetoric ( =0.99, ≤ 0.001), attitudes towards EU integration ( =-0.98, ≤ 0.01), EU foreign policy ( =-0.97, ≤ 0.01), EU asylum policy ( =-0.95, ≤ 0.05) EU budget authority ( =-0.95, ≤ 0.05), towards multiculturalism (=-0.98, ≤ 0.01), attitudes against liberal immigration policy (=0.94, ≤ 0.05), attitudes towards economic growth over environmental protection (=0.94, ≤ 0.05), against ethnic minorities (=0.94, ≤ 0.05), towards nationalism ( =0.91, ≤ 0.05), towards law and order over civil liberties (=0.96, ≤ 0.01), ideological Le昀琀-Right ( =0.96, ≤ 0.01), and the importance granted to the question of immigration ( =0.90, ≤ 0.05). This is interpreted as follows. The 昀椀rst latent dimension obtained via the ideal point estimation procedure stands as an indicator of several political ideologies and issues that are highly aligned and that are measured in the CHES data.

Users with high values in the scale of this 昀椀rst latent dimension display high negative attitudes towards the EU, immigration, multiculturalism, Islam, and elites. They also display right-wing ideological positions (as measured by the CHES data). They attribute more importance to law and order than civil liberties and to economic growth over the protection of the environment.

Latent dimension 1

Figure 2: Amount of users in our dataset according to their position along the first latent dimension, together with the proportion of those that have a Twitter bio (i.e., a profile description), and those that include emoji on their them They are also more nationalistic. User with low values in the scale of this 昀椀rst latent dimension display positive attitudes towards the EU, they are more le昀琀-leaning, are less nationalistic, and more acceptance of liberal immigration policy and multiculturalism. While the 昀椀rst latent dimension of has an explained inertia of 5.1% (see the annex for more details), these correlations show this dimension already captures le昀琀-right di昀erences that can be leveraged in analyses of online activity. The user is referred to the Codebook of the CHES data for more detailed description of the substance of the CHES dimensions that attribute now meaning to our 昀椀rst latent dimension. Fig. 3 explicitly illustrates the computation of the correlation using the position of MPs of political parties on a few dimensions.

Fig. 3 illustrates that parties clustered on the lower end of latent dimension 1 include opposition party Civic Platform and the parties politically closest to it: Polish People’s Party and the socially liberal Modern party (Nowoczesna). We will call these parties ‘le昀琀’ wing for simplicity.3 Parties clustered on the higher end of latent dimension 1 include the incumbent populist right-wing PiS and the national conservative party Kukiz’15. Figure.3 below demonstrates the relative dispersion of MPs within each party, with smaller parties being more dispersed. Party centroids (average of MP positions) are shown in black. 3In the Polish political context, the ‘le昀琀’ platform (Civic Platform, Modern etc) refers to support for liberal social policies and free-market economics. While the ‘right’ platform (PiS, Kukiz’15) refers to support for conservative social policies and social welfare. Religion 昀椀gures heavily in this le昀琀-right division [ 8, 9, 63 ].

4.2. RQ2: Does the use of recognised political emoji in user bios vary across the

political network? We start by splitting latent dimension 1 into deciles. Twitter accounts are roughly evenlydistributed across deciles (see table 1). The decile to which an account is attributed based on its latent dimension score forms its ‘target’ class. Below, we present our analysis of tendencies in bio emoji usage across these deciles. We 昀椀rst report summary statistics. Total number of emoji and average per bio, by decile (le昀琀 → right) decile count emoji avg emoji Note: Top 5 most frequently used emoji, by % (corresponding to indices 1-5).

Interestingly, the Polish 昀氀ag is the most common emoji regardless of decile. This o昀ers a useful corroboration of research 昀椀ndings that this 昀氀ag is a nonpartisan symbol in Poland, unlike for example the English or British 昀氀ags [ 29 ]. In contrast, the EU 昀氀ag appears more partisan.

The total number of (non-MP) accounts in the sample is 71884, with a total of 52914 possessing bios (74%). Only users with bios are included in the analysis. This gives us a 昀椀nal sample of 10232 bio-accounts which include at least one emoji in their bio (19.3%), with average of 3 emoji per bio (sd: 3.1; variance: 9.6). The mode emoji usage is 1 (3571 accounts), followed by 2 (2238 accounts). The maximum number of emoji used in any bio is 71. Table1 demonstrates that the average number of emoji-per-bio is consistent across deciles, hovering around the sample average of 3.

The most frequently used emoji across the sample are끹믦 , 끹믤 and ❤️ . Below we outline the most frequently used emoji by decile. Most frequently used emoji by latent dimension decile (le昀琀 → right)

le昀琀 5 le昀琀 4 le昀琀 3 le昀琀 2 le昀琀 1 right 1 right 2 right 3 right 4 right 5 1 2 3 4 5 ️✌ 믦끹 믤끹 ️❤ 믦끹 믤끹 ️❤ 믦끹 믤끹 ⚡ ⚡ ⚽

⚽ The dominant symbols on the le昀琀 and right are 끹믤 and ✝️ respectively. This 昀椀ts strikingly within expectations from the literature, and popular conception, that the primary cleavage in Poland is based on attitudes towards the EU and Catholicism6[ 1 ]. Indeed, the incumbent PiS emphasise religion in their rhetoric, alongside an infamously anti-EU stance3[ 5, 56 ].

Beyond this, a social attitudes divide also emerges, with the emoji emerging as the most popular on the far le昀琀, fading out in the centre, before being replaced by the ✝️ . These two emoji seem to be representing the dominant social attitudes associated with both sides of the political cleavage. Curiously, the⚽ dominates among the centre and centre-right deciles.

Some particularly interesting cases emerge. One of the most popular emoji on the farthest le昀琀 is ⚡, the emoji昀椀ed lightening-bolt. This is also the symbol of the Polish feminist Women’s Strike movement (Ogólnopolski Strajk Kobiet, OSK). This movement gained prominence following national anti-abortion laws. Inclusion of this emoji in a bio, therefore, would serve to express that account’s pro-choice, feminist political stance. Prevalence of the emoji umbrella ☂️ among bios with emoji by ideological deciles.

le昀琀 5 le昀琀 4 le昀琀 3 le昀琀 2 le昀琀 1 right 1 right 2 right 3 right 4 right 5 ️☂ 17% 11% 0% 2% 0% 0% 0% 0% 0% 0%

A second symbol closely associated with the abortion rights movement in Poland is the umbrella, which became a symbol following protests which took place in the rain4. In line with expectations, the emoji昀椀ed umbrella also appears to be being used for political selfrepresentation by accounts on the le昀琀. For accounts on the farthest le昀琀, the umbrella emoji makes up 17% of all the emoji used in bios, as visible in3. See Appendix for further detail on⚡ and ☂️ .

The emoji described above, whose use is associated with a user’s political position, all correspond to recognised o昀툀ine political symbols, including 昀氀ags, religious and protest symbols. There are other emoji in the table above whose use is also associated with a user’s political position, but which do not correspond to known political symbols. These relate to our third research question.

4.3. RQ3: Does the use of unrecognised political emoji in user bios vary across the

political network? Unlike the 昀氀ags and political symbols explored above, the political meaning of some emoji visible in the table ( ,

& ✌️ ) are harder to explain. These emoji do not have recognisable political associations, but serve the purpose of communicating tone, a昀ect, or even body-language[ 18 ]. The striking di昀erence in the ‘body-language’ of emoji on either side of the political network, from strong and assertive to open-palmed and peaceful, indicates the use of emoji in delivering political a昀ect. This, we argue, points to the development of a language game of emoji, whereby 4See for example a media report byNotes from Poland. certain emoji carrying a particular tone or expressive mode become associated with a political stance.

To illustrate this further, we calculate Pearson correlations based on % use of each emoji by decile and where that decile resides on the latent dimension (ie. lower value deciles are associated with proximity to ‘le昀琀’ wing parties and vice versa). Below we present the 10 emoji whose use is most strongly correlated with accounts on the ‘right’ (ie. associated with higher latent dimension scores), sorted by e昀ect size (p≤0.005). with the right-leaning users.

Pearson correlations based on % of use of emoji by decile of the latent dimension for emoji associated ⛪ ️✝ ️➡ 믧끹 unicode :church: :latin cross: :right arrow: :hundred points: :wedding church: :folded hands: :oncoming fist: :Hungary:

:baby: :prayer beads: coef

pval

Remarkably, these emoji communicate multiple social and political values associated with the right in Poland. The use of emoji depicting church (both⛪ & , used interchangeably) for self-representation are both associated with a 9.2% and 8.8% increase in use for each decile shi昀琀 to the right. While a one decile shi昀琀 to the right is associated with a 9.1% increase in use of the latin cross. Other religious emoji emerge as part of users’ self-descriptions including prayer-beads and folded hands.

Curiously, use of the baby emoji emerges as strongly associated with accounts on the right, perhaps capturing the political salience of Poland’s record low birth rates, it’s feminist tie-ins, and PiS’s subsequent election pledge to increase birth rates5. The baby emoji is an interesting example both of political value signalling through emoji, but also a fascinating illustration of the incumbent party’s cornerstone policy through it’s supporters choice of emoji1[ 0, 39, 14 ].

Meanwhile, the presence of the Hungarian 昀氀ag as an emoji highly correlated with the rightmost deciles in latent Polish political Twitter space is testament to the political association between the right in these two countries (which are both associated with anti-EU, far-right governments). As described above, the two ‘body-language’ emoji here, th e and , both express a political a昀ect which is strong and assertive.

As expected, political symbols associated with the Polish le昀琀, including the rainbow, EU 昀氀ag 5See here for a discussion about the controversial Family 500+ social welfare programme which has been a cornerstone of the Law and Justice party’s political programme

Pearson correlations based on % of use of emoji by decile of the latent dimension for emoji associated 믦끹 믤끹 ️☮ ⚡ ♑ coef

pval and feminist lightening-bolt are highly correlated with accounts on the ‘le昀琀’ of the network. A one decile shi昀琀 to the le昀琀 is associated with an increase in percentage use of these emoji by 9.4% (rainbow), 9.2% (EU) and 9.1% (lightening-bolt). The emoji, commonly used on social media to signify veganism or environmentalism, echoes research on the association between veganism and politically le昀琀 attitudes[ 21 ] as well as the CHES environmentalism issue dimension captured in the network.

A one-decile shi昀琀 to the le昀琀 is also associated with a 9.4% and 9% increase in use of the medium skin tone and medium-light skin tone. The literature on skin-tone modi昀椀ers suggests that while their use doesn’t necessarily re昀氀ect the skin-tone of the user, the use of any skin-tone modi昀椀er in the 昀椀rst place does appear to be associated with skin-tone [ 48, 55]. These 昀椀ndings indicate that use of skin-tone modi昀椀ers is likewise associated with ‘le昀琀’ political positions in Poland. The Capricorn and Ugandan 昀氀ag emoji are more di昀케cult to interpret but could pose interesting questions for further research, particularly since interest in astrological signs has been associated with political attitudes in previous research3[ 4 ]. Beyond these, the peace symbol and the purple heart emerge as potential examples of political a昀ect. Below we develop a more in depth discussion about the potential mechanisms behind these emoji. 4.4. Analysis of the

emoji

was consistently associated with users on the right of the political spectrum. We argue that unrecognised political emoji such as this play a role in political language games online, by communicating political a昀ect. To further test whether this emoji is being used in communicating political values, we ran a frequency count to identify the single-word terms most frequently co-occurring with .6 For users who include this emoji to describe themselves, the table above shows the 22 Polish terms most frequently used alongside it. Among all single-word 6Stop words removed: the, in, of, to be terms used for self-representation in these bios, babieslivesmatter, God, catholic, fatherland, patriot, power and honour are among those appearing most o昀琀en. “Lewactwo” is derogatory in tone, similar to ‘le昀琀ies’, while “duda2020” refers to Andrzej Duda’s PiS Presidential campaign. See Appendix for equivalent table for a comparative emoji associated with the le昀琀: ⚡. The a昀ective impression of

in a bio imbues elements of Social Dominance Orientation and Big Five personality traits associated with right-wing political values4[ 4, 52 ]. However, the use of emoji to communicate this a昀ect, and the implications for nuance and polarisation, have extensive relevance for political identity construction online. 44 68 52 20 17 16 14 14 12 23 10 9 9 9 9 9 8 8 7 7 7 10

5. Conclusion

The dictionary of 3633 emoji that sits in our phones is a smorgasbord of values, ideals, roles and preferences, containing a wealth of normative political content. Analysing which we use to describe ourselves, and when, has the potential for a unique and alternative avenue for understanding political values and expression. This paper has found that emoji play a role in political self-representation, and that Twitter users in di昀erent political circles choose to put di昀erent emoji in their bio, even when those emoji are not recognisably connected with their political position. Emoji o昀er users a medium through which to associate a昀ect with political value, making use of language games to build stronger collective identity around shared political positions. However, to heed McLuhan’s famous maxim “the medium is the message”3[8, pp. 7], emoji are a comparatively reductive and in昀氀exible medium to expressing political values. Further multi-sample investigation would be necessary to understand the broader implications on polarisation when emoji, as opposed to natural language, are used to express political values.

Acknowledgments

This work has been funded by the “European Polarisation Observatory” (EPO) of the CIVICA Research (co-)funded by EU’s Horizon 2020 programme under grant agreement No 101017201. P.R. acknowledges support by the Data Intelligence Institute of Paris (diiP) through the French National Agency for Research (ANR) grant ANR-18-IDEX-0001 “IdEx Université de Paris” and SoMe4Dem (Grant No. 101094752) Horizon Europe project. Data declared in 19 March 2020 and 15 July 2021 at the registry of data processing ofFondation Nationale de Sciences Politiques (Sciences Po) in accordance with General Data Protection Regulation 2016/679 (GDPR) and Twitter policy. For further details and the respective legal notice, please visithttps: //medialab.sciencespo.fr/en/activities/epo./ [39] G. Meardi and I. Guardiancich. “Back to the familialist future: the rise of social policy for ruling populist radical right parties in Italy and Poland”. InW:est European Politics 45.1 (2022), pp. 129–153. [41] [42] [43]

H. Miller, D. Kluver, J. Thebault-Spieker, L. Terveen, and B. Hecht. “Understanding emoji ambiguity in context: The role of text in emoji-related miscommunication”. InP:roceedings of the International AAAI Conference on Web and Social Media. Vol. 11. 1. 2017, pp. 152–161. H. Miller, J. Thebault-Spieker, S. Chang, I. Johnson, L. Terveen, and B. Hecht. ““Blissfully Happy” or “Ready to Fight”: Varying Interpretations of Emoji”. In:Proceedings of the International AAAI Conference on Web and Social Media. Vol. 10. 1. 2016, pp. 259–268. A. Pathak, N. Madani, and K. Joseph. “A Method to Analyze Multiple Social Identities in Twitter Bios”. In: Proceedings of the ACM on Human-Computer Interaction. Vol. 5. Cscw2. 2021, pp. 1–35. [44] F. Pratto, J. Sidanius, L. M. Stallworth, and B. F. Malle. “Social Dominance Orientation: A Personality Variable Predicting Social and Political Attitudes”. InJ:ournal of Personality and Social Psychology 67.4 (1994), p. 741. [45] P. Ramaciotti and Z. Vagena. “Embedding social graphs from multiple national settings in common empirical opinion spaces”. In:2022 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM). Ieee. 2022, pp. 60–67. [46] P. Ramaciotti Morales, J. Cointet, and G. Zolotoochin. “Unfolding the dimensionality structure of social networks in ideological embeddings”. InP:roceedings of the 2021 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining. 2021, pp. 333– 338. [47] [48] [49] [50]

M. A. Riordan. “Emojis as Tools for Emotion Work: Communicating A昀ect in Text Messages”. In: Journal of Language and Social Psychology 36.5 (2017), pp. 549–567. A. Robertson, W. Magdy, and S. Goldwater. “Emoji skin tone modi昀椀ers: Analyzing variation in usage on social media”. In:ACM Transactions on Social Computing 3.2 (2020), pp. 1–25.

A. Robertson, W. Magdy, and S. Goldwater. “Black or White but Never Neutral: How Readers Perceive Identity from Yellow or Skin-Toned Emoji”. InP: roceedings of the ACM on Human-Computer Interaction 5.Cscw2 (2021), pp. 1–23.

A. Robertson, W. Magdy, and S. Goldwater. “Self-representation on Twitter using Emoji Skin Color Modi昀椀ers”. In: Proceedings of the International AAAI Conference on Web and Social Media. Vol. 12. 1. 2018. [51] N. Rogers and J. Jones. “Using twitter bios to measure changes in self-identity: Are Americans de昀椀ning themselves more politically over time?” In: Journal of Social Computing 2.1 (2021), pp. 1–13. [52] C. G. Sibley and J. Duckitt. “Personality and Prejudice: A Meta-Analysis and Theoretical Review”. In: Personality and Social Psychology Review 12.3 (2008), pp. 248–279. [55]

6. Appendix

References

[1] K. Albury . “ Sexual expression in social media” . In:J. Burgess , A. Marwick , & T. Poell The sage handbook of social media ( 2018 ), pp. 444 - 462 .

[2] [3] M. Alfano , R. Reimann , I. O. Quintana , A. Chan , M. Cheong , and C. Klein . “ The A昀케liative Use of Emoji and Hashtags in the Black Lives Matter Movement on Twitter” . In:Social Science Computer Review ( 2022 ). doi: 10 .1177/08944393221131928.

M. Alfano , R. Reimann , I. Quintana, M. Cheong , and C. Klein . “ The a昀케liative use of emoji and hashtags in the Black Lives Matter movement: A Twitter case study” . In: ( 2021 ).

[4] G. Artazo , A. Ramia , and S. Menoyo. “ A new feminist ethic that unites and mobilizes people: the participation of young people in Argentina's Green Wave” . InG:ender Development 29 . 2 - 3 ( 2021 ), pp. 335 - 350 .

[5] P. Barberá . “ Birds of the same feather tweet together: Bayesian ideal point estimation using Twitter data” . In: Political Analysis 23.1 ( 2015 ), pp. 76 - 91 .

[7] [10] [11] [12] A. M. Bhatt , A. Goldberg , and S. B. Srivastava . “A language-based method for assessing symbolic boundary maintenance between social groups” . In:Sociological Methods & Research 51.4 ( 2022 ), pp. 1681 - 1720 .

A. L. Blanchard , L. E. Caudill , and L. S. Walker . “ Developing an Entitativity Measure and Distinguishing It from Antecedents and Outcomes within Online and Face-to-Face Groups” . In: Group Processes & Intergroup Relations 23.1 ( 2020 ), pp. 91 - 108 .

[8] T. Bojarowicz . “ New axes of political cleavages in Poland a昀琀er 2005” . In: Regional Formation and Development Studies 22 .2 ( 2017 ), pp. 6 - 15 .

[9] I. Borowik. “Religion, politics, and social attitudes in transforming Poland: a conclusion” . In: Religion, Politics, and Values in Poland. New York: Palgrave Macmillan, 2017 , pp. 313 - 324 .

M. Cornejo-Valle and J. Ramme . ““We Don't Want Rainbow Terror” : Religious and FarRight Sexual Politics in Poland and Spain” . In:Paradoxical Right-Wing Sexual Politics in Europe. Cham: Palgrave Macmillan , 2022 , pp. 25 - 60 .

M. Danesi . “ Emoji and the expression of emotion in writing” . InT:he Routledge Handbook of Language and Emotion. Routledge , 2019 , pp. 242 - 257 .

Bloomsbury Publishing , 2017 .

[13] G. Das , H. J. Wiener , and I. Kareklas. “ To emoji or not to emoji? Examining the in昀氀uence of emoji on consumer reactions to advertising” . InJ:ournal of Business Research 96 ( 2019 ), pp. 147 - 156 .

[14] M. Dudek and T. Wojewodzic . “ Does a Demographic Crisis Threaten European and Polish Agriculture?” In: Wieś i Rolnictwo 1 ( 190 ) ( 2021 ), pp. 97 - 117 .

[15] C. F. Epstein . “ Tinkerbells and Pinups: The Construction and Reconstruction of Gender Boundaries at Work” . In:Cultivating Di昀erences: Symbolic Boundaries and the Making of Inequality . Vol. 232 . 1992 , pp. 233 - 238 .

[16] B. Felbo , A. Mislove , A. Søgaard , I. Rahwan , and S. Lehmann . “ Using Millions of Emoji Occurrences to Learn Any-Domain Representations for Detecting Sentiment, Emotion, and Sarcasm” . In: ( 2017 ). eprint: arXiv: 1708 . 00524 .

[17] C. L. Franco and J. M. Fugate . “ Emoji Face Renderings: Exploring the Role Emoji Platform Di昀erences Have on Emotional Interpretation” . In: Journal of Nonverbal Behavior 44.2 ( 2020 ), pp. 301 - 328 .

[18] L. Gawne and G. McCulloch. “ Emoji as digital gestures” . In:LanguageInternet 17.2 ( 2019 ).

[19] J. Ge and U. Gretzel . “ Emoji rhetoric: a social media in昀氀uencer perspective” . In: Journal of marketing management 34 . 15 - 16 ( 2018 ), pp. 1272 - 1295 .

[20] E. Graells-Garrido , R. Baeza-Yates , and M. Lalmas . “ Every colour you are: Stance prediction and turnaround in controversial issues” . In:12th ACM Conference on Web Science . 2020 , pp. 174 - 183 .

[21] T. Grünhage and M. Reuter . “ What Makes Diets Political? Moral Foundations and the Le昀琀-Wing-Vegan Connection” . In: Social Justice Research 34.1 ( 2021 ), pp. 18 - 52 .

[22] J. J . Gumperz, ed. Language and Social Identity . Cambridge University Press, 1982 .

[23] S. Guntuku , M. Li , L. Tay , and L. Ungar . “ Studying cultural di昀erences in emoji usage across the east and the west” . In: Proceedings of the international AAAI conference on web and social media . Vol. 13 . 2019 , pp. 226 - 235 .

[24] O. Haimson and A. Ho昀mann. “ Constructing and enforcing “authentic” identity online: Facebook, real names, and non-normative identities” . InF:irst Monday ( 2016 ).

[25] S. Hall . “ Cultural Studies: Two Paradigms” . In:Media, Culture & Society 2.1 ( 1980 ), pp. 57 - 72 .

[26] A. R. Hochschild . “Emotion Work, Feeling Rules, and Social Structure” . InA:merican Journal of Sociology 85.3 ( 1979 ), pp. 551 - 575 .

[27] R. Jenkins . Social identity . Routledge , 2014 .

[28] S. Jolly , R. Bakker , L. Hooghe , G. Marks, J. Polk , J. Rovny , M. Steenbergen , and M. A. Vachudova . “ Chapel Hill expert survey trend 昀椀le , 1999 - 2019 ”. In: Electoral studies 75 ( 2022 ), p. 102420 .

A. Kariryaa , S. Rundé , H. Heuer , A. Jungherr , and J. Schöning. “ The role of 昀氀ag emoji in online political communication” . In:Social Science Computer Review 40.2 ( 2022 ), pp. 367 - 387 .

[29] [31] [32] [33] [35] [36] [38] [30] L. K. Kaye , S. Rodriguez-Cuadrado , S. A. Malone , H. J. Wall , E. Gaunt , A. L. Mulvey , and C. Graham . “ How Emotional Are Emoji?: Exploring the E昀ect of Emotional Valence on the Processing of Emoji Stimuli” . In:Computers in Human Behavior 116 ( 2021 ), p. 106648 .

W. Labov . The Social Strati昀椀cation of English in New York City . Cambridge University Press, 2006 .

M. Lamont and V. Molnár . “ The study of boundaries in the social sciences” . InA:nnual review of sociology 28.1 ( 2002 ), pp. 167 - 195 .

M. Lamont , S. Pendergrass , and M. Pachucki . “ Symbolic Boundaries” . In:International Encyclopedia of Social and Behavioral Sciences . Vol. 2 . 2015 , pp. 850 - 855 .

[34] J. Lindgren . “ Who believes that astrology is scienti昀椀c?” In: Northwestern Public Law Research Paper 14 - 10 ( 2014 ).

Springer Nature , 2021 , p. 353 .

W. Lowe . “ Understanding wordscores” . In:Political Analysis 16.4 ( 2008 ), pp. 356 - 371 .

[37] J. Manfredi-S'anchez, A. Amado-Su'arez, and S. Waisbord . “ Presidential Twitter in the Face of COVID-19: Between Populism and Pop Politics” . In:Comunicar: Media Education Research Journal 29 .66 ( 2021 ), pp. 79 - 90 .

M. McLuhan . “ Understanding Media: The extensions of man. New American Library” . In: Inc., New York ( 1964 ).

[40] A. Melucci . Challenging codes: Collective action in the information age . Cambridge University Press, 1996 .

[53] L. Stark and K. Crawford . “ The Conservatism of Emoji: Work, A昀ect, and Communication” . In: Social Media+ Society 1.2 ( 2015 ).

[54] S. Sugiyama . “ The Emoji and the Management of Social Boundaries” . In: ( 2018 ).

M. Sweeney and K. Whaley . “ Technically white: emoji skin-tone modi昀椀ers as American technoculture” . In: First Monday ( 2019 ).

[56] A. Szczerbiak . “ Opposing Europe or problematizing Europe? Euroscepticism and 'Eurorealism'in the Polish party system” . In:Opposing Europe 2 ( 2008 ), pp. 221 - 242 .

[57] A. Tabouret-Keller and R. B. Le Page . Acts of Identity: Creole-Based Approaches to Language and Ethnicity . Cambridge University Press, 1985 .

[58] V. Taylor and N. E. Whittier. “ Collective identity in social movement” . In: ( 1992 ).

[59] G. W. Tigwell and D. R. Flatla . “ Oh That's What You Meant! Reducing Emoji Misunderstanding” . In: Proceedings of the 18th International Conference on Human-Computer Interaction with Mobile Devices and Services Adjunct . 2016 , pp. 859 - 866 .

[60] L. Van Vliet , P. Törnberg , and J. Uitermark. “ The Twitter parliamentarian database: Analyzing Twitter politics across 26 countries” . In:PLoS one 15.9 ( 2020 ), e0237073 .

[61] P. Vermeersch . Domestic Discourses on European Integration in Poland Before and A昀琀er 2004 : Ideology, Nationalism, and Party Competition . Tech. rep. CES Central Eastern Europe Working Paper No. 66 , 2008 .

[62] L. Wittgenstein . “ Philosophy investigations” . In: ( 1953 ).

[63] T. Zarycki . “ Politics in the periphery: Political cleavages in Poland interpreted in their historical and international context” . In:Europe-Asia Studies 52.5 ( 2000 ), pp. 851 - 873 .

[64] X. Zheng and A. Sun . “ Collecting event-related tweets from Twitter stream” . In:Journal of the Association for Information Science and Technology 70.2 ( 2019 ), pp. 176 - 186 .