3.1Affect and emotions

What is affect? How does it differ from emotions or feelings? In linguistic approaches, emotion is first considered an embodied, personal, cognitive, and psychological experience that is communicated either linguistically or through bodily expressions (Enfield and Wierzbicka 2002Enfield, Nick J., and Anna Wierzbicka 2002 “The Body in Description of Emotion.” Pragmatics and Cognition 10:½: 1–25. , 4–6). Researchers have defined emotions from several perspectives, such as the cognitive and social viewpoints, and they either accentuate individual experience or social experience of emotions (see, e.g., Bednarek 2008Bednarek, Monika 2008 Emotion Talk Across Corpora. Basingstoke England, New York: Palgrave Macmillan. , 4–12 for an overview). According to Edwards (1999Edwards, Derek 1999 “Emotion Discourse.” Culture & Psychology 5 (3): 271–291. , 282), emotions and affectivity are cognitively grounded or cognitively consequential in relation to objects or events.

Here, our starting point is communicative activities and how affectivity is expressed in them. To differentiate between affect and emotion, we turn to the phenomenology of joint action. According to Salmela and Nagatsu (2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. , 451), affect is a phenomenal state that has two types of realizations:

Emotions are felt evaluative responses to specific objects and events and they motivate the subject to act in accordance with evaluative content of the emotion; to fight or flee in danger, to retaliate or retribute when offended, to hide in shame, and so on. Feelings can be part of emotion, and they can be experienced as bodily sensations or intentional feelings directed at the particular object of emotion or as both kinds of feelings at the same time. However, not all feelings such as rapport or alienation are part of emotions.(Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. , 451; our underlining)

Communicative situations are social and cultural situations of joint activity (Linell 2009Linell, Per 2009 Rethinking Language, Mind and World Dialogically. Interactional and Contextual Theories of Human Sense-making. Charlotte, NC: Information Age Publishing., 202). In other words, social actors express affect or emotions that are indexically grounded in situations (see Edwards 1999Edwards, Derek 1999 “Emotion Discourse.” Culture & Psychology 5 (3): 271–291. ). Ochs (1996Ochs, Elinor 1996 “Linguistic Resources for Socializing Humanity.” In Rethinking Linguistic Relativity, ed. by John J. Gumperz, and Stephen C. Levinson, 407–437. Cambridge: Cambridge University Press., 420) explained this in detail:

In all communities, affective stances are socio-culturally linked to social acts, in the minds of speakers (illocutionary acts), of hearers (perlocutionary acts), or of both speakers and hearers. [… P]articular affects help to constitute the meaning of particular acts. Where these affects are indexed by a linguistic form, that form may also constitutively index associated social acts.(Ochs 1996Ochs, Elinor 1996 “Linguistic Resources for Socializing Humanity.” In Rethinking Linguistic Relativity, ed. by John J. Gumperz, and Stephen C. Levinson, 407–437. Cambridge: Cambridge University Press., 420)

Here, we consider affectivity and emotions as social acts that are indexical and situated in contexts. Moreover, in social situations, emotions are tied to interpersonal relationships and communicative activities (Linell 2009Linell, Per 2009 Rethinking Language, Mind and World Dialogically. Interactional and Contextual Theories of Human Sense-making. Charlotte, NC: Information Age Publishing., 201–203). In her approach to emotions as social and cultural practices, Ahmed (2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London., 10) does not consider emotions as individual expressions from “inside out.” Instead, she proposes a model she calls the sociality of emotions, in which “emotions create the very effect of the surfaces and boundaries that allow all kinds of objects to be delineated. The objects of emotion take shape as effects of circulation” (Ahmed 2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London., 10).

In joint communicative activities, emotions are relational: they may be expressed together, or one emotion may have an effect on the co-actors. Emotions create bonds between social actors; they are shared in a way that bond people together (“towardness”), or they separate social actors from each other (“awayness”; Ahmed 2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London., 8–9). In addition, emotions, especially negative ones, can function in such a way that differentiation or othering between social actors takes place (Ahmed 2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London., 1).

In their phenomenological account of joint action, Salmela and Nagatsu (2017)Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. considered small-scale, face-to-face situations, such as singing, dancing, and spectating team sports. According to the researchers, shared emotions give a sense of we-agency during and in consequence of joint actions (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. , 451). When participants have an experience in which they share a similar type of emotion, they have similar evaluative contents and affective experiences, and they are aware of this: “Phenomenologically, the evaluative content and affective experience of an emotion are typically intertwined and intentionally directed at the particular object of emotion. While the evaluative content of an emotion is necessarily intentional, the affective experience is only contingently so” (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. , 457). In addition, the evaluative content of an emotion contains concerns, such as norms or values, for example (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. , 457). Emotions also differ in intensity from the weakest type to moderate and the strongest shared emotions, in which the degree of collectivity and concerns are either private or collectively shared (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. ).

3.2Affective stance and positioning

According to Ochs (1996Ochs, Elinor 1996 “Linguistic Resources for Socializing Humanity.” In Rethinking Linguistic Relativity, ed. by John J. Gumperz, and Stephen C. Levinson, 407–437. Cambridge: Cambridge University Press., 421), in verbally expressed emotions, linguistic elements help in identifying acts that take place with expression of those emotions. Emotions can be expressed explicitly by lexical elements, such as sad or furious, but they can also be communicated in implicit ways (Edwards 1999Edwards, Derek 1999 “Emotion Discourse.” Culture & Psychology 5 (3): 271–291. , 279). Affect and emotions have been studied in a great range of linguistic studies since the seminal work of Biber and Finegan (1989)Biber, Douglas and Edward Finegan 1989 “Styles of Stance in English: Lexical and Grammatical Marking of Evidentiality and Affect.” Text 9 (1), 93–121. , which focused on overt lexical and grammatical markers of a speaker’s stance, such as adjectives, adverbials, hedges, and verbs. In her corpus linguistic approach, Bednarek distinguished between emotional talk (signaling function) and emotion talk (denoting function; Bednarek 2008Bednarek, Monika 2008 Emotion Talk Across Corpora. Basingstoke England, New York: Palgrave Macmillan. , 11). She gave the examples of first person use – Oh fuck (signal), I’m really angry (denote) – and other uses – And then he goes “Oh fuck” (signal) and And he was very angry (denote; Bednarek 2008Bednarek, Monika 2008 Emotion Talk Across Corpora. Basingstoke England, New York: Palgrave Macmillan. , 11–12).

To make another distinction, when expressing emotions, social actors may orient toward an object in the world and express their emotion regarding it (something is terrible, nice, bad). Alternatively, they may formulate their own subjective view on it: I hate it (Edwards and Potter 2017Edwards, Derek, and Jonathan Potter 2017 “Some Uses of Subject-side Assessments.” Discourse Studies 19 (5): 497–514. , 497–498). Edwards and Potter (2017)Edwards, Derek, and Jonathan Potter 2017 “Some Uses of Subject-side Assessments.” Discourse Studies 19 (5): 497–514. called these O-side (object) or S-side (subject) assessments. O-side assessments are intersubjective and shared, whereas S-side assessments “formulate the evaluation […] restricted to the judgement of the speaker” (Edwards and Potter 2017Edwards, Derek, and Jonathan Potter 2017 “Some Uses of Subject-side Assessments.” Discourse Studies 19 (5): 497–514. , 511–512), building up the speaker’s position. When using S-side assessments, speakers may manage communicative situations to avoid misunderstandings and disagreement while managing diverse views (Edwards and Potter 2017Edwards, Derek, and Jonathan Potter 2017 “Some Uses of Subject-side Assessments.” Discourse Studies 19 (5): 497–514. , 511–512). Either way, social actors express a stance with an orientation toward an object. We build on this distinction below (see Section 4).

Twitter discussion can be considered as written interaction in a digital context, situated and temporarily limited in its discussion on a topic. Although we use the term discussion here, the specific characteristics of this social media interaction should be considered, and it should not be compared to face-to-face interactions as such. Participants post tweets as their contributions to this large-scale interaction that can be interactive in the sense that they respond to topic, but they have a choice between directing their post explicitly to other participants with @ or publishing their post without any addressee. However, our data derives from large scale data, and it is not possible to account for this. We will describe it and its limitations in the following section.

4.1Large scale data and clustering of tweets

In our data, the #jesuisCharlie hashtag was used in 1.2 million tweets in 51 languages between 7.1.2015 (18:33 h) and 14.1.2015 (06:50 h). They were collected11.Collected by Marco T. Bastos and Raquel Recuero. with the yourTwapperKeeper application (Bruns and Liang 2012Bruns, Axel, and Yuxian Eugene Liang 2012 “Tools and Methods for Capturing Twitter Data During Natural Disasters.” First Monday 17 (4). http://​firstmonday​.org​/ojs​/index​.php​/fm​/article​/view​/3937​/3193. Accessed May 15, 2016. ). In this study, we concentrate on tweets written in French and apply a mixed-methods approach. This allows for the combination of a large-scale quantitative analysis revealing general tendencies found in the entire dataset to a detailed examination of linguistic instances in their usage contexts. The study design consists of two phases, as described below.

In the first phase, our method is applying the large-scale approach of clustering, an exploratory machine learning method used to find structure and groupings in previously unseen data (Kaufman and Rousseeuw 1990Kaufman, Leonard, and Peter J. Rousseeuw 1990 Finding Groups in Data: An Introduction to Cluster Analysis. New York: John Wiley. ; Divjak and Fieller 2014Divjak, Dagmar, and Nick Fieller 2014 “Clustering Linguistic Data.” In Corpus Methods for Semantics: Quantitative Studies in Polysemy and Synonymy, ed. by Dylan Glynn, and Justyna Robinson, 405–441. Amsterdam/Philadelphia: John Benjamins Publishing Company. ; Moisl 2015Moisl, Hermann 2015 Cluster Analysis for Corpus Linguistics. Berlin: De Gruyter Mouton. ). We use clustering to find thematic groupings in the tweets and group tweets with similar topics together into clusters. This is ensured by constructing a vector space representation for each tweet using word2vec (Mikolov et al. 2013Mikolov, Tomas, Kai Chen, Greg Corrado, and Jeffrey Dean 2013 “Efficient Estimation of Word Representations in Vector Space.” In ICLR Workshop.), a neural network model that learns to detect semantically similar words based on their usage contexts (Firth 1957Firth, John Rupert 1957 “A Synopsis of Linguistic Theory 1930–1955. In Selected Papers of J.R. Firth 1952–1959, ed by. Frank Robert Palmer, 168–205. London: Longman.; Gries 2012Gries, Stefan Th 2012 “Behavioral Profiles: A Fine-grained and Quantitative Approach in Corpus-based Lexical Semantics.” In Methodological and Analytic Frontiers in Lexical Research, ed. by Gary Libben, Gonia Jarema, and Chris Westbury, 57–80. Amsterdam/Philadelphia: John Benjamins Publishing Company. ). We hypothesize that this grouping will also tie together similar expressions of affectivity and reveal the most typical ways of expressing affect in the tweets. Further, for each cluster, we estimate the 30 most typical tweets.

We extracted the tweets written in French based on the language identification offered by Twitter and excluded retweets and tweets without any linguistic information from the data. This gave us the final dataset, which consisted of 108,236 tweets.

Before the clustering, the tweets were preprocessed with UDPipe (Straka and Strakovà 2017Straka, Milan, and Jana Straková 2017 “Tokenizing, POS Tagging, Lemmatizing and Parsing UD 2.0 with UDPipe.” In Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, 88–99. Vancouver, Canada: Association for Computational Linguistics. ) to obtain morphological and syntactic information on the data. As a second step, we excluded tokens belonging to part-of-speech classes with little linguistic content, namely adpositions, determiners, punctuation, numbers, conjunctions, auxiliary verbs, and symbols, from the tweets. After this, the tweets were vectorized using the French word2vec embeddings published by Ginter et al. (2017)Ginter, Filip, Jan, Hajič, Juhani, Luotolahti, Milan Straka, and Daniel, Zeman 2017 CoNLL 2017 Shared Task – Automatically Annotated Raw Texts and Word Embeddings. Prague: Charles University, LINDAT/CLARIN Digital Library at the Institute of Formal and Applied Linguistics (ÚFAL), Faculty of Mathematics and Physics. http://​hdl​.handle​.net​/11234​/1​-1989. To obtain vectors for whole tweets instead of individual words, we counted the average vectors from the word vectors belonging to the tweet. The clustering was done with KMeans Minibatch in Scikit learn. Different clustering solutions were compared, and the solution with 15 clusters using Euclidian distance was estimated as the best. Out of the 15 clusters, several clusters contained tweets that were posted to point out a link to a website, such as news events that unfolded. All these were excluded, and for the analysis, we kept eight topical clusters, with a total of 240 tweets. The topics found in these clusters were similar to those found in previous studies, which confirms our approach (Johansson et al. 2018Johansson, Marjut, Aki-Juhani Kyröläinen, Filip Ginter, Lotta Lehti, Attila Krizsán, and Veronika Laippala 2018 ”Opening Up #jesuisCharlie Anatomy of a Twitter Discussion With Mixed Methods.” Journal of Pragmatics 129: 90–101. ; Smyrnaios and Ratinaud 2017Smyrnaios, Nikos, and Pierre Ratinaud 2017 “The Charlie Hebdo Attacks on Twitter: A Comparative Analysis of a Political Controversy in English and French.” Social Media + Society 3 (1): 1–13. ). Here, we focus on the affectivity expressed in tweets in these clusters.

4.2Qualitative analysis

In the qualitative phase, we analyzed the 240 tweets22.We use tweets as such without any lexical or grammatical correction. that were included in the clusters and their expressions of affectivity.

The study of affect and emotions has to be considered on the three following levels: social and cultural practice (macrolevel), interactional practice of joint action (mesolevel), and language use and communicative acts in tweets (microlevel). At the microlevel, our analysis consisted of linguistic analysis of affectivity. It was studied regarding lexicogrammatical elements in terms of signaling or denoting emotions in the stances expressed by users (Bednarek 2008Bednarek, Monika 2008 Emotion Talk Across Corpora. Basingstoke England, New York: Palgrave Macmillan. ; Du Bois 2007Du Bois, John 2007 “The Stance Triangle.” In Stancetaking in Discourse. Subjectivity, Evaluation and Interaction, ed. by Robert Englbretson, 139–182. Amsterdam/Philadelphia: John Benjamins Publishing Company. , 163). At the mesolevel, we focused on how the users positioned themselves, and we analyzed the tweets as O-side or S-side assessments (Edwards and Potter 2017Edwards, Derek, and Jonathan Potter 2017 “Some Uses of Subject-side Assessments.” Discourse Studies 19 (5): 497–514. ). In addition, we studied whether the participants aligned or disaligned themselves with others (Du Bois 2007Du Bois, John 2007 “The Stance Triangle.” In Stancetaking in Discourse. Subjectivity, Evaluation and Interaction, ed. by Robert Englbretson, 139–182. Amsterdam/Philadelphia: John Benjamins Publishing Company. , 163). In this analysis, we distinguished between individual and collective emotions. At the macrolevel, we studied what kind of we-agency was expressed (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. ) and what kind of shared values and togetherness or awayness belonged to this experience (Section 5). We then focused on tweets in which users departed from the we-agency and the group belonging and what kind of affectivity and values were expressed (Section 6). In the end, we analyzed tweets that consisted of negative affectivity towards the we-agency and shared values (Section 7.)

5.1Solidarity and grief

In the previous studies (see Section 2), solidarity and freedom of speech were found to be the most common expressions during this event. We will explore this further in French-speaking tweets in order to consider how it forms what we call we-agency (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. ). In the tweets that belong to this category, the participants shared their commitment through the hashtag #jesuisCharlie, especially identifying themselves as Charlie:

In Examples (1) and (2), the tweeters identify themselves as Charlie by giving the index of time (depuis, since; encore et toujours, still and always). These tweets are S-side assessments in which the subjectivity is either individual (ex. (2)) or collective (ex. (1)). In Example (1), the stance is expressed through the personal pronoun nous (we) with a preposition (chez, at) indexing a local place, followed by an impersonal pronoun (on, we) that is inclusive of the writer. In Example (2), the participant uses the structure je suis (I am) to accentuate his/her commitment to Charlie. These are the kind of tweets which create bonds between networked users, creating what Ahmed calls towardness (Ahmed 2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London., 8–9).

The participants expressed their support for solidarity and freedom of speech:

In Examples (3)–(4), the stances are not subjective, although the writers are evaluating Charlie; rather, they represent O-side assessments. In Example (3), the meaning of “To be Charlie” is spelled out by this participant, followed by a negative rhetorical question (Qui ne s’y revendique pas? Who does not claim this?) that presupposes that all participants in this Twitter discussion are identifying with Charlie. At the end of this tweet, there is a communicative act – an order – that boosts the rhetorical question (Soyez Charlie! Be Charlie!). Thus, identification is strongly invited by this participant. In Example (4), the identification is performed collectively, equating Paris with Charlie (Paris est Charlie, Paris is Charlie) and cheering on Charlie (Vive Charlie, Long live Charlie). The affect here is sharing emotions with people who feel the same according to these participants.

The participants also expressed their sorrow and mourning toward the victims.

The Examples (5) and (6) are S-side assessments: there is a subjective stance expressed with the first-person pronoun (je, I and moi, me), and in 6, there is an inclusive impersonal pronoun on (we). In Example (5), the participant addresses the message toward the victims (vie perdus, lost lives), which is the object of her/his emotion (en mémoire de, homage; in the memory of, paying homage). Another hashtag, #MarcheRepublicaine, is used, where this writer announces she/he will be attending, J’y serais (I will be there). In Example (6), the participant signals disbelief (cauchemar, nightmare).

In the following examples, the affectivity is about shared values:

In Examples (7)–(10), the shared values – freedom of speech and solidarity – are the evaluative content of these tweets. They are O-side assessments, except for Example (9), and they take objects from national pride. In Example (7), the pen as a sign of freedom of speech is evoked, and in Example (8), there is mention of the French national anthem, the Marseillaise. In (9), the participant uses a first-person pronoun (j’étais, I was), with the adjective fier (proud) and the mention of nationality (Français, French). This emotion expresses this writer’s evaluation of a solidarity march and group belonging. In Example (10), there is an invitation: this participant uses a swearword at the beginning of her/his tweet (putain, fuck), thus signaling an emotion. He/she encourages all the tweeters to participate in the solidarity march.

In these examples, the social actors express affect that constitutes the core of the affectivity in this Twitter discussion. Participants share similar emotions that reveal shared values, primarily involving identifying with Charlie and manifesting group belonging. The participants are aware of one another’s positioning as they invite others to join and share the same emotions and values; thus, they align with each other. In this sense, participation in this large-scale Twitter discussion is a joint activity. The participants’ positioning shows moral values that are behind this strong manifestation of affect. However, most of these tweets in this section do not denote an emotion by naming it; instead, they signal them (cf. Bednarek 2008Bednarek, Monika 2008 Emotion Talk Across Corpora. Basingstoke England, New York: Palgrave Macmillan. ). The values not only comprise defending freedom of speech, but they also include defending democratic and national values, which are the values that create towardness. The tweets constitute the we-agency (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. ) and togetherness (Ahmed 2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London.).

6.1Shared values and anger against the other

There are tweets that express similar shared emotions and shared values as those analyzed in Section 5.1., but here, participants make the distinction between self and others manifested in S-side assessments. In other words, these types of tweets contain material in which social actors observe behaviors that are deviant from a straightforward commitment to the emotional content.

In Examples (11)–(14), the participants align themselves with shared emotions and values (freedom of speech, republic, liberty, future). In Example (11), the writer distinguishes between freedom of speech and terrorism, whereas in 12, the participant takes national values (la république, republic) as her/his object of evaluation and points out what she/he considers to be the threat to it (religions). In other words, these social actors point out the we-agency and togetherness (ex. (10)), but they also illustrate its boundaries: terrorism, religion, and sick people (the other). In other words, from the social and cultural perspective, there is towardness, but it is signaled explicitly by what breaks the bond. These Examples (11)–(14) are S-side assessments in which the participants refer to themselves either by repeating the hashtag #jesuisCharlie or employing other first-person expressions in the singular or plural. In Examples (12) and (13), the participants denote their emotions clearly, amour, haine, and choc (love, hate, and shock).

In the last example in this section, the tweet is similar to those in which the writers engage in self-identification. However, like in the Examples (11)–(14), here, the other is also pointed out. The other is the enemy:

In Example (15), there is an expression of shared values (liberté d’expression, freedom of speech) and that the writer is free (libre), but at the end of the tweet, he/she points to the political enemy, right-wing politician Marine LePen.

In these cases, the core of shared emotions is quite strong – they are comparable to the previous case, as are the values expressed here. They express themselves in S-side assessments identifying or supporting Charlie and the values that they associate with this publication or the event. However, the participants take different positions: they signal disalignment with others or other ideologies. Therefore, the affectivity that emerges, in addition to solidarity, freedom of speech, and commitment, is that of anger and hatred targeting what the other represents and threats to the values the participants want to claim. It expresses strong awayness from the other who does not support these values.

6.2Distancing from group belonging: Doubts and irony

There are tweets in which the participants share the same values but at the same time blame the other for not maintaining the shared emotions and values. In this respect, they point out explicitly the awayness they have observed (Ahmed 2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London., 8–9). They are critical about maintaining the togetherness and are sad that it is breaking up, as in the following example:

The Example (16) is an O-side assessment, while the Example (17) is an S-side assessment. In Example (16), the participant asks two rhetorical questions in which he/she wonders about people’s commitment. He/she indexes time, encore (still) and de passage (momentary), signaling the passing of this momentary towardness and engagement in the shared values. The core identification and group belonging are at stake here – qui sont Charlie (who are Charlie). In Example (17), the participant complains and explicitly denotes an emotion (marre, sick). He/she complains that the togetherness of shared emotions and shared values did not last long (n aura pas duré longtemps, did not last long). He/she refers to the shared values by indicating la belle unite (beautiful unity) and referring to his/her country (mon pays, my country).

There are also participants who do not accept the core identification and group belonging:

In Example (18), S-side assessment, this participant does not accept the core identification: he/she initially denies this by expressing an order to those who commit to it (Faut arrêter avec vos #JeSuisCharlie, Stop with your #IamCharlie). Although he/she denies of being Charlie (je m’appelle pas Charlie, my name is not Charlie), he/she commits to the shared values by indicating nationality with the use of inclusive on (we; on est tous français, we are all French). The Example (19) is an O-side assessment in which the user weighs if one should be Charlie with the construction je suis (I am). However, in Example (20), all the utterances build an S-side assessment and allows the participant to be ironical. He/she enumerates three hashtags in the beginning of the tweet that name the killed police officer (Ahmed), the perpetrators (Kouachi), and Charlie and identifies with them all (comme ça quoi qu’il arrive j’ suis sûr d’ être dans le bon camp, whatever happens I am sure I’m in the right camp). Here, there is no marking of the shared emotions or the shared values. In sum, the participants express their doubts and ironical stance against the others who do not share the same emotions by signaling the emotion; none of them denote emotions explicitly.

7.1Irritation, anger, and repulsion

Some of the tweeters considered that not everyone shares the same values, and thus, they threaten democracy and togetherness. They tweet about the awayness (Ahmed 2015Ahmed, Sara 2015 The Cultural Politics of Emotion. Routledge, New York & London., 8–9). Their tweets are othering: the persons who take as the objects of their emotion by which they express their concerns (Salmela and Nagatsu 2017Salmela, Mikko, and Nagatsu, Michiru 2017 “How Does It Really Feel to Act Together? Shared Emotions and The Phenomenology of We-Agency”. Phenomenology and the Cognitive Sciences 16, 449–470. , 457). Especially, they consider that freedom of speech is not shared by politicians; if it was, they should be talking about it in relation to other matters as well:

In Examples (21)–(23), O-side assessments, the participants comment on the ongoing discussion across media while mentioning either political figures (Raïf, Dieudonné) or a philosopher (Michel Onfray). In Example (21), this participant invites others to action for another cause; in Example (22), there is a rhetorical question about the limits of freedom of speech, and, actually, this tweet can be interpreted as being against hate speech. In Example (23), the participant notes the emotions the philosopher has expressed, aligning with it questioning of others’ surprise (tu m’étonnes, you surprise me). They observe the awayness of the moment that united users together.

There are tweets that involve disalignment from the political ideologies in France or French politics:

Example (24) is one of the clearest examples of S-side assessments that contains a denoted emotion. The participant expresses the affect from his/her point of view with utterances with je (I) that express his/her malaise (encore plus mal, even more sick) and distrust (je ne fais plus confiance, I don’t trust anymore) of the French political left. The socialist politics are also under attack in Example (25), where the participant bashes the president at the time, François Hollande, in an O-side assessment. Here, there is an expression of anger and repulsion (politique de merde, his damn politics). In other words, the affectivity is turned against those who do not contribute to the togetherness or we-agency but instead fake it or eat away at it.

7.2Disgrace and condemnation

In the last set of examples, the social actors express feelings of disgrace about three different elements linked with this event:

In Examples (26) and (27), the emotion is denoted as disgrace (honte, honteux) in O-side assessments. In (26), the writer condemns the inappropriate opinions of their co-participants. In (27), the writer is indignant that the media has not covered a similar, poignant news event that has taken place in Africa. In both tweets, the repulsion is about an observation of clear offense against the shared value of freedom of speech. These social actors disalign from those who have transgressed this value. In the last Example (28), an O-side assessment, the participant condemns ISIS, accusing them of destroying the past and comparing them with the Nazis. They condemn the kind of behaviors that create awayness and violate most clearly the core of democratic values of the we-agency and togetherness.