Sage Journals: Discover world-class research

Abstract

There is growing research interest in the sharing of emotions through social media. Usually centred on ‘newsworthy’ events and collective ‘flows’ of emotion, this work is often computationally driven. This article presents an interaction-led analysis of small data from Twitter to illustrate how this kind of intensive focus can ‘thicken’ claims about emotions, and particularly empathy. Drawing on Goffman’s work on ritual, we introduce and then apply the idea of ‘empathy rituals’ to exchanges about emotional distress on Twitter, a platform primarily researched using big data approaches. While the potential of Goffman’s work has been explored in some depth in relation to digital performances, its emotional dimension has been less fully examined. Through a focus on Twitter conversations, we show how reading small data can inform computational social science claims about emotions and add to sociological understanding of emotion in (digital) publics.

Keywords

Emotions Goffman rituals Twitter

This article explores how Goffman’s work on ritualisation can illuminate the emotional dimensions of online interactions on public platforms. By framing interpersonal exchanges based on conventions of empathy¹ as ‘empathy rituals’, we argue that Goffman’s work provides a lens through which to consider a particular type of ‘public’ emotion work – everyday exchanges about emotional distress on Twitter. We suggest that understanding more about the emotion work involved in such exchanges may offer insight into the collective and aggregate expression of emotion online and a sociological and empirical basis for analysing people’s expression of ordinary emotional distress online. Unlike existing studies of Twitter which tend to focus on grand events through hashtags and the ad hoc publics they help to create (Bruns and Burgess, 2011; Bruns and Highfield, 2013), we look at the platform through the underexplored area of small conversations (Cogan et al., 2012).

We show how small exchanges can help us to make sense of the aggregate emotional transitions identified through computational social science. This involves stepping back from debates about how to detect and provide support to people in distress who are using social media (O’Dea et al., 2015) to offer a sociologically informed analysis of responses² that are already happening in these spaces and which are part of the banal rather than topicalised Internet (Hine, 2015). While not the lens for this particular analysis, we are sympathetic to calls to engage with the political economy of big data (Frade, 2016) and for a more historical, and hence sociologically imaginative, approach to digital practices (Uprichard, 2012). Our start point, however, is to treat everyday ritual encounters on social media every bit as seriously – and as worthy of our sociological imagination – as sociologists since Goffman have treated offline encounters.

We understand Twitter – like other digital spaces – to be revealing of the social (Weller et al., 2014). Our particular interest is in empathy, the wider social significance of which has (re)surfaced in recent debates³ on race and gender on Twitter and offline. Empathy in this context can be read as a type of emotional capital – an emotionally valued asset or skill that is stratified and is, in its turn, stratifying (Reay, 2004). As with other types of capital evident on social media,⁴ emotional capital is intertwined with issues of identity and power and can challenge or reinforce offline hierarchies of who and what is understood to matter. Using Goffman’s understanding of ritual, we step back from this patterning of empathy to explore how empathy plays out at an interactional level on Twitter. Sharing emotion (implicitly and explicitly) is part of our everyday ‘getting by’ (Brownlie, 2014) and so it is important to understand more about such sharing and its potential risks. A focus on empathy in the context of Twitter is also timely because the digital turn within sociology foregrounds the complexity of how we can read emotion in and through text. Indeed, Puschmann et al.’s (2014: 426) point that Twitter ‘signifies something to those who engage in it that is difficult to describe only in terms of the data that is produced’ speaks to broader challenges of documentary interpretation.

We start by sketching out some literature on sharing emotions through social media, including how Goffman’s work has been drawn on to date, before outlining the research methodology and presenting an analysis of empathic interactions on Twitter through the lens of ritualisation.

Tweeting Emotion

A body of work focused on the sharing of emotions through social media, including Twitter, has emerged in the last five years, much of it informed and driven by computational methods (Kivran-Swaine et al., 2014; Yuan et al., 2016). Kim et al. (2012: 495), in their machine learning analysis of emotion in Twitter conversations, for example, found that conversational partners usually express the same or more positive emotion than that found in the original tweet they are replying to⁵ and that tweets containing sympathy, apology or complaint were the ones most likely to elicit emotional content from others. Other research has looked at the influence of user characteristics, including gender (Kivran-Swaine et al., 2014), and at how technological features shape online sharing (Bazarova et al., 2015). Kivran-Swaine and Naaman (2011), for instance, highlight the relationship between the sharing of emotional content and the properties of Twitter users’ social networks: those with many followers may be more likely to share emotions, but were less likely to do so if they had a network with higher density,⁶ suggesting there are costs to sharing in public or semi-public fora (Bucholtz, 2013: 23). How we imagine our audience, then, along with network norms, shapes emotional sharing (Litt and Hargittai, 2016).

It is increasingly clear that there is variation in how emotion is shared across digital spaces (Zappavigna, 2012: 77–78) and that platforms, including Twitter, are themselves heterogeneous. Developed as a text message platform, Twitter was closely tied to and inspired by mobile phone culture and the potential for ‘mobile intimacy’ (Hjorth and Lim, 2012). Indeed, this ‘as it happens’ sharing is linked to a ‘norm of authenticity’ – though such sharing is, in practice, shaped by what it is acceptable to say but also by why people choose to disclose publicly in the first place. This might include trying to align one’s sense of public and private selves as well as achieving support from an identified network (Veletsianos and Stewart, 2016; Zappavigna and Martin, 2018). So, while sharing of emotions on social media might intensify existing confessional culture (Bauman, 2000), ‘public’ digital disclosures have complex roots and cannot be assumed to always be about visibility: people may disclose to many, and do so in ephemeral ways, in order to avoid the visibility of disclosing to the few in a sustained fashion (Brownlie, 2018).

Research on the role of social media in relation to the markers of mental health and suicidality has also emerged in recent years, again, typically led by big data (Burnap et al., 2015; Cavazos-Rehg et al., 2016; De Choudhury et al., 2013). Where sharing of emotions on social media has been conceptualised in the social sciences it tends to be through understandings of affect (Garde-Hansen and Gorton, 2013). There is not space to engage with the wide range of theoretical positions subsumed under the affective turn (Gregg and Seigworth, 2010) but, in common with the diffuse ‘ontology of fluidity, mobility and change’ (Frade, 2016: 866) that informs big data analysis, affect in the context of social media has tended to involve a concern with how emotion flows between human bodies and between humans and non-humans (Hillis et al., 2015). While this is important, the need also to focus on relationships, interactions (and discourses) – and their role in constituting and interpreting affect – has been recognised for some time.⁷ It is to the interactional dimensions of online emotion that we now turn.

Goffman’s Platform Performances

Alongside the computational social science and affect approaches that have tended to dominate analysis of emotions on social media are sociologically informed conceptualisations of social media, and the Internet more broadly, including some that have drawn on Goffman (Hogan, 2010; Knorr Cetina, 2009; Murthy, 2012; Robinson, 2007). This work engages with a range of Goffman’s ideas, including ‘presentation of self’ and ‘encounters’, to inform our understanding of interactions based on ‘response presence’ rather than ‘co-presence’ (Knorr Cetina, 2009: 74).

Writing in a pre-digital age, long before social media, Goffman (1967) introduced the idea of ‘platform performances’ – performances which take place before an audience and position most of us as ‘vicarious watchers’. Ritual, we suggest, is an underused yet productive concept for thinking through the emotional and relational aspects of such platform performances⁸ in a digital context. While Goffman (1981: 17) himself was unsure about the term because of its association with ‘otherworldliness and automaticity’, Manning (1989: 365) notes that when using ‘ritual’ Goffman was referring both to the ‘smooth running of everyday encounters’ and ‘the honouring of the selves who people them’ – an honouring that includes saving the face of those we engage with.⁹ Existing work on Goffman and social media, including Twitter, tends to focus on ritual in the former sense (see, for example, Murthy, 2012) and, as a consequence, the emotional and relational import of ritualisation for understanding digital interactions is sidelined. So, while sociology’s reliance on Goffman in general is sometimes lampooned,¹⁰ we suggest that thinking about the emotional implications of Goffman’s work for digital contexts still has some way to go.¹¹

In his writing on supportive and remedial interchanges, Goffman (1971: 63) focused on the generic relationship between ‘doer and recipient’, though others have attempted to place a more specific relationship at the heart of his work. Manning (1989) suggests that rules are for strangers not friends (indeed friends often affirm their relationship through rule-breaching) and yet friends, in some situations, reserve the right to be treated as strangers. The complexity of how to manage such relational double-footing is writ large on social media platforms such as Twitter. While Manning pulls out the implications of Goffman’s work on rituals for particular relationships, Goffman (1981: 18) made clear the broader significance of emotions to rituals: it is, he argued, through the latter that the former are safeguarded. Interaction is based on emotional and subjective experience and, as such, all interactions allow for the possibility of empathy but also risk its absence (Goffman, 1983: 9). Scheff’s (2013) recent notion of the ‘Goffman/Cooley conjecture’ reminds us that the risks of everyday interactions that come to be regularised through ritual are indeed affective ones.

Later in the article we make the empirical case for reading interpersonal exchanges on Twitter through the emotional and relational dimensions of Goffman’s work. We argue that – contra the vogue for computational reading of emotions at an aggregate level – there is a need to return to small interchanges happening in digital public spaces to remind ourselves of the emotional gamble of sharing online.

Methodology: Following Feelings Online

Initial optimism about the potential of following the ‘imaginations, opinions, ideas, and feelings of hundreds of millions of people’ (Manovitch, 2012 in Burgess and Bruns, 2012) through researching digital data has given way to pragmatism about the methodological difficulties of doing so. The social phenomena studied using Twitter data tend to be those that have been made into events – made grand – through hashtags and retweets. Smaller discussions and everyday practices of empathy that do not solidify into a ‘Twitter event’ cannot be analysed (or found) with the same ease (Rogers, 2014).

Building on an innovative approach,¹² we used a multi-stage process to collate tweets and to capture responses to these so that Twitter data could be read as conversations. First, we collected tweets that were suggestive of emotional distress, using key phrases that we found (through initial searches) to be commonly used to express what appeared to be hopelessness, despair, low self-worth and, in some cases, suicidality.¹³ These searches were not foolproof: they were likely to include tweets that were not suggestive of suicidality or emotional distress, as well as missing many tweets that were. While more sophisticated search terms for the detection of emotional distress may be possible, our aim was exploratory – to begin to understand how identified examples of emotional distress are responded to on Twitter.

We processed these data to exclude retweets, producing an initial dataset of 279,005 tweets. As our primary research interest was in responses, and to avoid spotlighting potentially vulnerable people, the initial dataset was not the focus other than to provide a context for the responses.

The next step involved collecting ‘conversations’ from the 15,846 tweets that had at least one reply.¹⁴ A conversation refers to an initial tweet including the search terms above (with retweets removed) in combination with replies to that tweet and any replies to replies. We then focused on a three-week period from 25 September 2015 to 15 October 2015, to move further towards a sample appropriate for qualitative analysis. Conversations where the initial tweet did not appear to be about emotional distress and where the initial tweet was in the middle of a conversation (i.e. it began with an @) were then excluded, and we entered a second stage of manual filtering to exclude instances where any of the conversation was in a language other than English, and where the thread was missing conversational partners due to tweet or account deletion. Ultimately, 398 conversations were coded and analysed in NVivo.

These data are, therefore, dyadic or small-group conversations resulting from individual expressions of emotional distress, as opposed to broader hashtags such as #mhchat (a hashtag used to collect discussions around mental health on Twitter). We are presenting a partial and decontextualised look at responses to emotional distress online through a qualitative and sociologically informed analysis. Nevertheless, it is an approach consistent with Twitter’s own diversity of use and provides a qualitative baseline for further analysis.¹⁵

There has been much written about the difficulties of interpreting the emotional content of social media text (Thelwall and Kappas, 2014) including statements of suicidality (O’Dea et al., 2015). Approaches such as sentiment analysis are error-prone and Twitter data are seen as peculiarly ‘noisy’ (Kim et al., 2012). But reading emotion is never simply a technical exercise in accessing the ‘correct’ emotion, but an epistemological and ontological investigation of how emotions come to be constituted and reconstituted through their expression (Brownlie, 2014).

Analysing Twitter data qualitatively, however, involves various other methodological challenges. To arrive at a meaningful volume of text, a relatively large sample (in qualitative research terms), needs to be generated. At the same time, we have little or no contextual data about those who are writing the tweets (Bail, 2014). Most of the conversations we looked at were dialogical rather than involving multiple actors, and many appeared to be between people with an existing relationship. Close readings of these ordinary conversations suggest they mainly involve younger people. This is reflected in the conversational style adopted and the substantive focus of the concerns raised (for instance, problems with college courses). Such an assumption would be consistent with what we know about the general Twitter population (Murthy and Eldredge, 2016) and how social identity impacts on the way people tweet (Nguyen et al., 2013). Some conversations also offer an indication of gender, through the pronouns and names used, but again this is only an impression. While manual tracing of tweets might fill in some demographic gaps, we decided against this: ethically, because of the retrieval of personal information required, and analytically because meanings do not become apparent through knowledge of a participant’s background – we cannot simply read one off the other (Stenvoll and Svensson, 2011: 572).

Likewise, it is difficult to know to what extent users were pseudonymous in the sample, though a sizable proportion had non-standard names. There has been a shift towards the use of real names on social networking sites, and in the case of Facebook a variably enforced policy on this. The turn towards real names and stable identities creates the problem of context collapse (Marwick and boyd, 2011), which in turn may stifle talk about emotions. Twitter allows pseudonymous participation as well as the creation of multiple accounts. It is plausible that those Twitter users engaged in talk about emotions may be more likely to use pseudonyms. As Van Der Nagel (2017: 312) has documented, pseudonymity and platform or profile differentiation continue to be used to ‘deliberately compartmentalise identities’, and potentially in this case to open up spaces for talk about emotions.

Our aim is to map some features of how expressed distress is negotiated in conversation. Taking as our start point that there is no way round the interpretive, we focus on being as explicit as possible about our ‘decision trail’ (Cheek, 2004: 1147). In doing so we recognise the need to maintain privacy while researching feelings that are publicly posted but perhaps imagined as intimately shared. When drawing on individual tweets, we follow Markham (2012) and paraphrase extracts rather than quoting directly.

Empathy Interchanges on Twitter

Saving (Digital) Face

I’m close to giving up

- what’s wrong? are you alright? xx

– I’m just tired xx

— okay love, let me know if there is anything I can do xxx

–– thank you xxx (Conversation 86)¹⁶

Exchanges such as this are common on Twitter and in the conversational dataset. In their large-scale analysis of emotion on Twitter, Kim et al. (2012) refer, in passing, to a similar pattern: a person who was advised to pray after tweeting about their grief, replied, ‘Not really religious, but thanks man.’ This interaction also ended with a statement from the ‘responder’ that he/she was ‘here if you want to talk’. Indeed, Kim et al. (2012) argue on the basis of their computational analysis that this speaks to a general pattern: Twitter users tend to ‘accommodate’ each other’s emotions, and the expression of particular emotions, including sympathy, ‘influences’ others to become more positive. Discussing emotional transitions, they conclude that a shift away from talking about distress could signal that the person who tweeted about their upset is feeling better, perhaps because of having tweeted.

In our data, positive emotion is very much in evidence in response to emotional distress. Most responses in our sample of nearly 400 conversations, including the extract from the conversation above, involve expressions of love and affection: ‘I love you’ or simply ‘hugs’ and textual kisses (‘xxx’). These are often accompanied by images or links to visual and audio clips – of favourite celebrities, musicians or films – chosen either as a source of distraction or as symbolic of love and comfort.

Direct expressions of empathy are also common including through acronyms such as IKR (I know, right?). These are either cognitively framed (‘I know’; ‘same’; ‘I can relate’) or are more embodied (‘I feel you man’; ‘HUGS’; ‘free hug coupons’ and ‘pats on back’). Despite Twitter’s restricted character count, some of these suggest considerable depth of shared knowledge, for instance, about feelings of powerlessness in the face of depression or the physical symptoms of anxiety.

Extract 1

I tell myself that depression’s sight depends on movement and if I don’t move it can’t see me then I stay in bed all day

- its a rattlesnake tru (Conversation 17)

Extract 2

just by thinking you feel one triggers it more and you feel it in your chest … so scared!

- so true. and then the medicine makes it even worse. you can’t win (Conversation 5)

While Kim et al.’s (2012) computational study works with an understanding of emotion that is interactional and fluid, it falls short of exploring how empathy, such as the above, is expressed and responded to on Twitter; nor does it offer ways of reading ‘platform performances’ that go beyond surface descriptions of a transition from negative to positive emotion. In other words, it stops short of a theoretically informed big data approach, one that engages with ‘the unspoken or implicit meanings that occur in-between words’ (Bail, 2014: 467). Drawing on the relational and emotional dimensions of Goffman’s work on rituals we suggest, offers a theoretically informed approach towards such emotional transitions, further examples of which are outlined below.

Extract 3

When you fuck up so bad; depressed, empty; alone. When is something gonna work out for me?

- are you okay?

– I’ll be ok :/ just going through something. thank you tho (Conversation 40)

Extract 4

I can’t deal with this stress anymore. I’m so close to giving up

- if ya want to chat about it, pop me up x

– think I’m just going to go to bed but thanks anyway x

— ok (Conversation 3)

Extract 5

I don’t know how to be happy anymore.

[…]

- what’s up sweetie?

– just another day. you?

— I’m just at work. Cheer up alright. Always here to chat if you want.

–– I’ll be okay, thanks though (Conversation 4)

On the one hand, what is taking place here is basic rule-following to achieve smooth interaction¹⁷ as rejecting an opening statement on social media about distress is akin to ‘declining an extended hand’.¹⁸ However, unlike Goffman’s (1981: 18–19) typical interchange, where the person making an opening remark signals that their statement is a request and apologises for interrupting, the opening statements above are unusually explicit. Belying the clichéd nature of how the interactions then unfold, these are significant emotional processes and Goffman offers a framing for understanding them through the ritual of face-saving.

For Goffman (1967: 19), face-saving rituals are ‘sacred’ because they involve interactions ‘through whose symbolic component the actor shows how worthy he is of respect or how worthy he feels others are of it’. Goffman suggests there are four classic moves involved in face-saving. First, a ‘misconduct’ is identified and a challenge is made on this basis; second, an offering or a chance to correct the misconduct and re-establish equilibrium occurs. This might involve showing that an expression that provoked anxiety was ‘really a meaningless event, or an unintentional act, or a joke not meant to be taken seriously, or an unavoidable, “understandable” product of extenuating circumstances’ (1967: 20). For Goffman, it is this second move that allows the ‘suspect’ person to show that, in fact, ‘he can take the role of the other and that the rules of conduct are in place’ (1967: 21). In other words, a remedial interchange has taken place.

This relates to Smith’s (1759) observation that while we wish for ‘fellow feeling’ it is through ‘mutual sympathy’, that we come to view others, and then our own emotions, as ‘impartial spectators’. The exchanges above can be read as examples of face-saving where an expression of ‘fellow feeling’ is followed by actors becoming ‘impartial spectators’ of their own performances.

Goffman’s third move involves ‘acceptance’ by the receiver of the ‘suspect’s’ new framing, which, in turn, allows the exchange to be concluded through the fourth move of an expression of gratitude, a thanks to those who have given ‘the indulgence of forgiveness’ (Goffman, 1967: 22). Goffman calls this process, from acknowledgement of threat to face to re-establishing ritual equilibrium, an interchange. Its length and intensity, he suggested, is adapted to the persistence and intensity of the threat (1967: 13). At the end of such an interchange we have a clearer sense of what Goffman calls the ‘line’ – how an actor evaluates their own situation and that of others.

In the exchanges above, an initial statement of distress (‘I feel empty/distressed/like I’m drowning’) tends to be met by a query (‘are you okay?’). If one takes the expression of despair as in some way troubling or a breach of everyday interaction, the request for clarification can be read as a challenge to this. After a request for clarification, the original tweeter then either seeks to offer reassurance and/or to back off from their initial statement. In Goffman’s (1967: 43) terms, this back-tracking or ‘venturing nothing’ can be read as an ‘offering’ which, if accepted (through, for instance, statements of ‘being there’ or of empathy¹⁹) allows the ritual to be concluded with an expression of gratitude – ‘I’ll be fine, thanks though’; ‘but thanks for the offer’; ‘thanks for asking tho’.

One variation is when the responder replies not with a plea for clarification but by protesting. Statements such as ‘no one would care if I took my life’, for instance, may provoke a response such as ‘I would care a lot actually’ (Conversation 77). Protests can also take the form of injunctions not to say certain things – so, for instance, tweets in response to suicidal tweets, include – ‘stop with that’; ‘stop that. Right now’; ‘stop the emo’. Such exchanges are highly contingent: they can be disruptive of the expression of distress yet at the same time they can open up the possibility of new forms and spaces of support as in the exchange below:

Nobody would even know if I died. Or care.

- No! Please don’t say that. I would care

– thanks [names tweeter]

— always lovely, if you need someone to talk to you know I’m always around! I’ll send you my new number. (Conversation 88)

Such protest exchanges, not surprisingly, appear to happen between people with an existing relationship, though the expression of distress can create new ‘situational obligations’ (Goffman, 1963). In the above exchange, for instance, the responder advises they are always available, although it is only through this Twitter exchange that a new contact number is offered.

Expressions of gratitude, which as is evidenced above are a core part of empathy rituals, constitute the third largest category of coded data, after ‘love and affection’ and ‘empathy’. The relationship between gratitude, resilience and well-being has been explored in positive psychology (Watkins et al., 2014) and popular culture (Kaplan, 2016) and there have been calls for gratitude to be investigated online in the context of crises (Shaw et al., 2013). There is, however, little sociological investigation of gratitude, on or offline. We suggest that gratitude, when expressed as part of an interpersonal exchange around emotional distress in a public Twitter, helps to end or draw a line under a ritual exchange or, at least, discourage further exchange, while also acknowledging the recognition that has occurred. Gratitude, though, is not just about interpersonal relationships but promoting wider sociability (Smith, 1759). Harpham (2004) makes a similar point when he argues that although gratitude is individually expressed, it is only possible through being part of a community. Expressions of gratitude on Twitter help both to acknowledge and maintain social bonds, while at the same time allowing us to manage how much we share through these bonds.

Understanding why the above expressions of face-saving, and closure, might be needed returns us to the point raised earlier about the ambiguity of why emotional distress is expressed in public contexts in the first place. Elsewhere, we suggest that emotional expression on public Twitter may be less a cry for help than a means of diffusing emotion, or of safely expressing emotion without fear of being singled out for response (Brownlie, 2018). One reason the need to save face arises is that the actual and intended audience for a tweet may be at odds. In other words, saving face may become necessary because people are trying on/out different faces without necessarily having an expectation of response. This is less about ‘crowd sourcing’ emotional support – seeking out the (emotional) wisdom of the crowd (Surowiecki, 2004) – than an investment in the idea of a faceless crowd allowing for certain ‘faces’ to be rehearsed, particularly in relation to still stigmatised feelings of depression and loneliness.

Few of those who tweeted about their distress demonstrated a sense of audience, nor what the intention of sharing the distress was.²⁰ Tweets such as ‘I knew you’d show up’ were, therefore, the exception. For instance, one expression of distress – ‘I DON’T THINK ANYONE UNDERSTANDS HOW HARD THIS IS’ – is met with a specific response from a known other – ‘if you need someone, you’ll always have me’. While this direct response is acknowledged, the possibility remains that the purpose of the initial tweet was not to receive empathy from a known (or unknown) other but rather to declare a (presumably strong)²¹ feeling to oneself or everyone else. In other words, the audience in such cases might well be the self – Twitter as diary (Murthy, 2012) – or, conversely, all the user’s followers and, therefore, no one in particular. Presumably, too, given the initial strong feeling was about isolation, being responded to empathically might produce feelings associated with being contradicted.

In many cases, the interaction, at least on public Twitter, peters out after the expression of gratitude, or after an offer to take the conversation to another space.

I feel so depressed and I can’t fix it. Nothing is right.

- you know you can DM²² me if you ever need to speak to someone x

– thank you so much, you’re such a sweetheart x

— aw, it’s nothing x (Conversation 180)

It is impossible to know whether such interactions end because they are satisfactorily concluded, on Twitter or elsewhere, or because the exchange was unwelcome in the first place – perhaps because sharing was an end in itself, or the responder was not the imagined audience. Murthy (2012) has suggested that having a conversation on Twitter can be like sitting in a room not knowing who is going to appear through the door, or indeed who is listening behind it. This, he argues, is consistent with the main purpose of Twitter being to publish content rather than foster networks. In the case of dialogic exchanges about distress on public Twitter, however, both publishing and fostering of networks is happening at the same time, and herein lies the ambiguity of such interchanges. Goffman’s (1983) framing allows the above exchanges to be read in such a way that the work of ‘self-sustained restraints’ is recognised as well as the ambiguity around why people choose to tweet about their distress. It seems reasonable to at least consider that those who tweet their distress are reserving the right, as Manning put it, to be treated as strangers, and for the rules of interaction relating to strangers not to be breached.

This Goffman-led analysis questions assumptions, therefore, that such exchanges represent a straightforward move towards empathy. While it could be argued that drawing on Goffman leads to a more cynical or instrumental reading of online interactions, this need not mean an assumption that we are never genuinely invested in our performances and/or that we only ever have our own interests at heart. As Goffman (1959: 17–18) notes, we are all on a spectrum from acting sincerely to cynically, from having complete belief to no belief in our performances. Below, we add to our understanding of the nature of platform performances through surfacing some of the norms which shape the sharing of emotion.

Surfacing Norms

The ambiguities mentioned above reflect and reinforce the relative lack of direct discussion of norms surrounding the sharing of emotional distress on Twitter. In our data, such norms surface through apparent breaches of expected practice, for example expectations about the length of time it takes to respond to a distressed message – ‘why on earth are you bothering to reply to this now?’ While the synchronous/asynchronous distinction is not always clear-cut (Rettie, 2009), on Twitter there is an expectation of near-immediate response. As with other platforms (Buehler, 2017), there are also norms about what belongs on Twitter and hence how emotional support should be sought and offered.

Today was awful – so much stress. I want to give up, but I’d feel so depressed about it.

- oh no, what happened bub?

– things that should not be talked about on Twitter

— alright then (Conversation 120)

‘Gatherings’ on social media – as offline too – happen from moment to moment, and evidence of one’s attachment to them depends on a ‘capacity for involvement’ which Goffman (1963: 247) argues must be ‘immediate and continuous’. It is not surprising that there are norms around who should be providing support, including an assumption that those who know us best will be the first port of call. The following exchange, for instance, is in response to a tweet: ‘who can help me now? I feel like I’m going to drink again.’ This conversation suggests a triage of support people imagine should be in place:

- you must have friends who can help

– not in my hometown, and no one else is answering

— dm me (Conversation 180)

We came across only one admonishment of lack of response in nearly 400 conversations: ‘I guess you’re the only one who cares enough to make sure I’m okay’ (Conversation 20). Instead, the most common way of reinforcing norms about what can be shared, and who should respond, is through requests or suggestions that the discussion is moved to another space. These alternative spaces might be offline – so there are frequent suggestions that people ‘come over’ or ‘visit’. Other alternative spaces are constituted online, through private channels – ‘dm me’; ‘dm is always open’; ‘text me’; or ‘do you want to take it to pm?’²³ Both these online and offline spaces provide the potential for further communication and are consistent with the sentiment that responders are ‘there’ if the person wishes ‘to talk’. This suggests that for some who respond, conversation on public Twitter itself is a precursor to, or something other than, ‘being there’ emotionally in other spaces. These discussions about where, and with whom, emotional distress is shared are how those who respond to tweets do the boundary work of becoming ‘an audience’ to emotional distress (Litt and Hargittai, 2016: 1).

Drawing on Goffman’s insight that ‘those who happen to be in perceptual range of the event will have some sort of participation-status relative to it’ (Goffman, 1981: 3, cited in Murthy, 2012), Murthy highlights that, just as we are sensitive to the expectation of response from particular people rather than every audience member, we also manage our membership of such audiences through what we choose to be attentive to and how we respond. Even when we acknowledge we have seen a tweet, therefore, we may respond in a manner that suggests ‘this is not for me’ or, as in the case of the extract above about drinking, ‘it might be for me but only once you’ve tried others’. Being part of a potentially large-scale interaction gives audience members more room for manoeuvre (Goffman, 1967: 131).

Norms also become apparent through responses which appear to disrupt emotional expression. Jokes or flippant comments are much more common in small groups than dialogical chat. A common response to a tweet about ‘feeling empty’, for instance, is ‘to eat something’. It is impossible to know from these exchanges if the initial statement was an ‘authentic’ sharing of distress and/or if the responders are, as Manning (1989) suggests, doing friendship by breaching interaction rules. What is clear, is that humorous responses provoke humorous replies. Other responses which have the potential to disrupt emotional expression involve direct advice. This is not to suggest no advice is given: there are suggestions that people should talk to a professional, take medication, count their blessings, listen to music, as well as more flippant advice such as to eat ice cream. More directive tweets, however, tend to be resisted:

I’m so over this. I drive in in the morning and then feel depressed until I shut my eyes at night.

- Im sorry :( Can you cope by playing music?

– nope [explains why not]

— try thinking about what you’re grateful for, it will help you remember the positives. It could work?

–– I only want to sleep

—– then stay focused on that. I know it helps me.

—— I try, but it’s more complex than that, and I don’t want to go into it.

——- I understand and I don’t want to push it. But I’m here if you wanna talk to someone (Conversation 72)

Mirroring the fact that most expressions of distress are not linked to specific problems but are vague expressions of disquiet or unease – ‘idk.²⁴ It’s just feelings’ – general encouragement rather than specific advice tends to be a more common response: ‘keep breathing’, ‘don’t quit’, ‘keep swimming’, ‘keep afloat’; ‘hang in there’; ‘things will get better’; ‘stay strong’. These stock responses – the key features of which are perseverance, resilience, keeping on – are not disruptive in the way more specific advice might be and, like Hallmark messages, are potentially less risky in interactional terms than direct advice-giving. Despite their blandness, these responses are regulating the emotional content of those expressing distress on public Twitter, yet we noted only one overtly negative response to such statements:

i feel like I’m drowning

- keep swimming

– jfc²⁵ (Conversation 200)

A more common response to such ‘Hallmark-like’ comments is to be opaque in turn. The tweets we have been concerned with begin as anything but vague but they become more so once they are acknowledged by others as distress. This could be read as a version of ‘vaguebooking’. A portmanteau of ‘vague’ and ‘Facebook’ (West, 2015: 15), this refers to online postings which are akin to ‘noisy silences’ (Linde, 2008: 196). For some, these practices are criticised because they disregard the audience: either one should post direct content to named others (for instance, through private messages) or vent to all. Others see the value in such vagueness or lack of transparency, and suggest it can allow emotion to be expressed without the risk to privacy or respect from specific disclosure. In offline encounters, there are ways to maintain the protection afforded by vagueness through, for instance, accepting emotional support incidentally or ‘by the by’ (Brownlie and Spandler, 2018) or through the guise of practical help. It is difficult to transfer such incidental, unobtrusive support online; hence the need, at least in the public space of Twitter, for vagueness by those who post and for tact on behalf of the audience (Goffman, 1959).

Conclusion

What, then, can we conclude from this micro-analysis of everyday conversations about emotional distress on Twitter? We have suggested that these practices of one-to-one or one-to-a-few sharing of intimate stories in ‘public’, on Twitter, are a type of emotion work that can contribute to our understanding of aggregate patterns on Twitter as well as to the emotion work we do in other public spaces. Goffman’s conceptualisation of ritual has allowed us to offer a sociologically and empirically informed study of this work, highlighting interactional practices that would certainly be lost through big data methods but also possibly through affect driven approaches where the focus is more on how emotion circulates across interactions on social media than how it is constituted within them. These practices include expressions of empathy, love and affection but also face-saving (rooted in our becoming spectators of our own digital sharing) and the governing of emotional expression through protest, humour and injunctions to ‘carry on’. At the same time, other practices are notably absent, including, most saliently, direct advice-giving.

These practices are the online equivalent of the scaling down and up of ‘talk’ and listening that needs to be managed for ‘involvement’ rather than ‘civil inattention’ to take place (Goffman, 1967). Calibrating ‘platform performances’ in both the digital and Goffman sense of the term is complex because of ambiguity about who our audiences are, and what constitutes appropriate sharing. Norms about what can be shared, where and with whom, shape attention and response but they are, as we have seen, rarely made explicit. There is, then, a vagueness about these norms but also about the way emotional distress is negotiated. At the same time, to paraphrase Garfinkel (1967), there would appear to be good reasons for sharing vaguely, not least that it allows people to try on or out different faces without the commitment of private one-to-one interaction. In other words, it allows friends to be given the right to be treated as strangers.

People’s reasons for writing about emotional distress cannot be read off tweets in any straightforward way and the methodology here is limited in this respect. It is, however, an area worth further investigation, not least because considerable attention is being paid to identifying and responding to emotional distress on Twitter, including computationally, but with little interest in how emotional distress in this space is already being responded to in the everyday. The need to focus on this banal sharing is pressing because in mapping computationally large-scale shifts in emotion online (including assumed shifts at an aggregate level from negative to positive emotions), we may miss the complexity of such transitions and misread what these shifts mean. The interactional ‘lines’ that so fascinated Goffman are drawn even in the most fleeting and clichéd of interchanges on Twitter, and interpreting them through the conceptual lens of empathy rituals reveals the affective gamble of these everyday digital interactions.

Footnotes

Acknowledgements

Thank you to Dr Nishanth Sastry (KCL) and Dr Dmytro Karamshuk for their work on the Twitter dataset and to Dr Karen Gregory for offline sharing. Thank you also to Natalie Bazarova and colleagues at the Social Media Lab, Cornell University, for the visiting scholarship which allowed space for this article to be developed.

Funding

Thank you to the funders of the Shared Space and Space for Sharing project: the ESRC, AHRB, EPSRC, Dstl and CPNI (Grant No.ES/M00354X/1).

Notes

Julie Brownlie is a senior lecturer in Sociology at the University of Edinburgh. Her research and teaching interests include the sociology of emotions, relationships and documentary analysis. She is currently researching trust and empathy online as part of the ESRC’s EMoTICON programme. Recent publications include Ordinary Relationships: A Sociological Study of Emotions, Reflexivity and Culture (Palgrave, 2014) and ‘Thinking sociologically about kindness’ (with Simon Anderson, Sociology).

Frances Shaw is a postdoctoral researcher at the Black Dog Institute in Sydney where she is examining the ethics and politics of social media and mobile device interventions for the diagnosis and prevention of mental illness. Previously she was a research fellow at the University of Edinburgh researching the expression of emotional distress on social media, and how trust and empathy are established in online spaces. Her primary research interests include digital ethics, social media cultures, digital methods, health cultures, digital embodiment and the self.

References

Bail

(2014) The cultural environment: Measuring culture with big data. Theory, Culture & Society 43(3): 465–482.

Bail

(2016) Emotional feedback and the viral spread of social media messages about autism spectrum disorders. American Journal of Public Health 106(7): 1173–1180.

Bauman

(2000) Liquid Modernity. Cambridge: Polity Press.

Bazarova

Choi

Schwanda

et al . (2015) Social sharing of emotions on Facebook: Channel differences, satisfaction, and replies. In: Proceedings of the 18th ACM conference on computer supported cooperative work & social computing, Vancouver, BC, Canada, 14–18 March, pp. 154–164.

Brownlie

(2014) Ordinary Relationships: A Sociological Study of Culture, Emotions and Reflexivity. London: Palgrave.

Brownlie

(2018) Looking out for each other online: Digital outreach, emotional surveillance and safe(r) spaces. Emotion, Space and Society 27: 60–67.

Brownlie

Spandler

(2018) Materialities of mundane care and the art of holding one’s own. Sociology of Health and Illness 40(2): 256–269.

Bruns

Burgess

(2011) The use of Twitter hashtags in the formation of ad hoc publics. In: Proceedings of the 6th European Consortium for Political Research (ECPR) general conference 2011, University of Iceland, Reykjavik, August.

Bruns

Highfield

(2013) Political networks on Twitter: Tweeting the Queensland state election. Information, Communication & Society 16(5): 667–691.

10.

Bucholtz

(2013) Diffused intimacy: Trust and self-disclosure in online relationships. Media Studies 1: 23–39.

11.

Buehler

(2017) Seeking emotional support on Facebook. Social Media and Society 3(3): 1–11.

12.

Burgess

Bruns

(2012) Twitter archives and the challenges of ‘big social data’ for media and communication research. M/C Journal 15(5). Available at: http://journal.media-culture.org.au/index.php/mcjournal/article/view/561Driscoll/0.

13.

Burnap

Colombo

Scourfield

(2015) Machine classification and analysis of suicide-related communication on Twitter. In: Proceedings of the 26th ACM conference on hypertext & social media, Guzelyurt, Northern Cyprus, 1–4 September, pp. 75–84.

14.

Cavazos-Rehg

Krauss

Sowles

et al . (2016) A content analysis of depression-related tweets. Computers in Human Behavior 54: 351–357.

15.

Cheek

(2004) At the margins? Discourse analysis and qualitative research. Qualitative Health Research 14(8): 1140–1150.

16.

Cogan

Andrews

Bradonjic

et al . (2012) Reconstruction and Analysis of Twitter Conversation Graphs. New York: ACM Press, 25–31.

17.

Collins

(2004) Interaction Ritual Chains. Princeton, NJ: Princeton University Press.

18.

De Choudhury

Gamon

Counts

et al . (2013) Predicting depression via social media. In: Proceedings of the seventh international AAAI conference on weblogs and social media, Cambridge, MA, USA, 8–11 July, pp. 128–137.

19.

Frade

(2016) Social theory and the politics of big data and method. Sociology 50(5): 863–877.

20.

Garde-Hansen

Gorton

(2013) Emotion Online: Theorizing Affect on the Internet. Basingstoke: Palgrave Macmillan.

21.

Garfinkel

(1967) Good organisational reasons for ‘bad’ clinic records. In: Garfinkel

Studies in Ethnomethodology. Englewood Cliffs, NJ: Prentice-Hall, 186–207.

22.

Goffman

(1959) The Presentation of Self in Everyday Life. New York: Anchor Books.

23.

Goffman

(1963) Behaviour in Public Places. New York: The Free Press.

24.

Goffman

(1967) Interaction Ritual: Essays in Face-to-Face Behavior. Brunswick, NJ: Aldine Publishing Company.

25.

Goffman

(1971) Relations in Public: Microstudies of the Public Order. New York: Basic Books.

26.

Goffman

(1981) Forms of Talk. Philadelphia, PA: University of Pennsylvania Press.

27.

Goffman

(1983) The interaction order: American Sociological Association. 1982 Presidential Address. American Sociological Review 48(1): 1–17.

28.

Gregg

Seigworth

(eds) (2010) The Affect Theory Reader. Durham, NC: Duke University Press.

29.

Harpham

(2004) Gratitude in the history of ideas. In: Emmons

McCullough

(eds) The Psychology of Gratitude. Oxford and New York: Oxford University Press.

30.

Hemmings

(2005) Invoking affect. Cultural Studies 9(5): 548–567.

31.

Hillis

Paasonen

Petit

(2015) Networked Affect. Cambridge, MA: MIT Press.

32.

Hine

(2015) Ethnography for the Internet: Embedded, Embodied and Everyday. London: Bloomsbury Academic.

33.

Hjorth

Lim

(2012) Mobile intimacy in an age of affective mobile media. Feminist Media Studies 12(4): 477–484.

34.

Hogan

(2010) The presentation of self in the age of social media: Distinguishing performances and exhibitions online. Bulletin of Science, Technology & Society 30(6): 377–386.

35.

Kaplan

(2016) The Gratitude Diaries: How a Year Looking on the Bright Side Can Transform Your Life. New York: Dutton.

36.

Karamshuk

Pupavac

Shaw

et al . (2017) Towards transdisciplinary collaboration between computer and social scientists: Initial experiences and reflections. In: Fu

Luo

J-D

Boos

(eds) Social Network Analysis: Interdisciplinary Approaches and Case Studies. Boca Raton, FL: CRC Press.

37.

Kim

Bak

(2012) Do you feel what I feel? Social aspects of emotions in Twitter conversations. In: Proceedings of the sixth international AAAI conference on weblogs and social media, Dublin, Ireland, 4–7 June, pp. 495–498.

38.

Kivran-Swaine

Naaman

(2011) Network properties and social sharing of emotions in social awareness streams. In: Proceedings of the ACM 2011 conference on computer supported cooperative work, Hangzhou, China, 19–23 March, pp. 379–382.

39.

Kivran-Swaine

Ting

Brubaker

et al . (2014) Understanding loneliness in social awareness streams: Expressions and responses. In: Proceedings of the eighth international AAAI conference on weblogs and social media, Ann Arbor, Michigan, 1–4 June, pp. 256–265.

40.

Knorr Cetina

(2009) The synthetic situation: Interactionism for a global world. Symbolic Interaction 32(1): 61–87.

41.

Lamm

Silani

(2014) Insights into collective emotions from the social neuroscience of empathy. In: Von Scheve

Salmella

(eds) Collective Emotions: Perspectives from Psychology, Philosophy, and Sociology. Oxford: Oxford University Press.

42.

Linde

(2008) Working the Past: Narrative and Institutional Memory. Oxford: Oxford University Press.

43.

Litt

Hargittai

(2016) The imagined audience on social network sites. Social Media + Society 2(1): 1–12.

44.

Manning

(1989) Ritual talk. Sociology 23(3): 365–385.

45.

Markham

(2012) Fabrication as ethical practice: Qualitative inquiry in ambiguous internet contexts. Information, Communication & Society 15(3): 334–353.

46.

Marwick

boyd

(2011) I tweet honestly, I tweet passionately: Twitter users, context collapse, and the imagined audience. New Media & Society 13(1): 114–133.

47.

Murthy

(2012) Towards a sociological understanding of social media: Theorizing Twitter. Sociology 46(6): 1059–1073.

48.

Murthy

Eldredge

(2016) Who tweets about cancer? An analysis of cancer-related tweets in the USA. Digital Health 2: 1–16.

49.

Nguyen

Gravel

Trieschnigg

et al . (2013) TweetGenie: Automatic age prediction from tweets. ACM SIGWEB Newsletter 4(4): 1–6.

50.

O’Dea

Wan

Batterham

et al . (2015) Detecting suicidality on Twitter. Internet Interventions 2(2): 183–188.

51.

Papacharissi

(2012) Without you, I’m nothing: Performances of the self on Twitter. International Journal of Communication 6(18): 1989–2006.

52.

Puschmann

Bruns

Mahrt

et al . (2014) Epilogue: Why study Twitter? In: Weller

Bruns

Burgess

et al . (eds) Twitter and Society. New York: Peter Lang, 425–432.

53.

Reay

(2004) Gendering Bourdieu’s concepts of capitals? Emotional capital, women and social class. Sociological Review 52(2): 57–74.

54.

Rettie

(2009) Mobile phone communication: Extending Goffman to mediated interaction. Sociology 43(3): 421–438.

55.

Robinson

(2007) The cyberself: The self-ing project goes online, symbolic interaction in the digital age. New Media & Society 9(1): 93–110.

56.

Rogers

(2014) Debanalising Twitter: The transformation of an object of study. In: Weller

Bruns

Burgess

et al (eds) Twitter and Society. New York: Peter Lang.

57.

Scheff

(2013) Goffman on emotions: The pride–shame system. Symbolic Interaction 37(1): 108–121.

58.

Shaw

Burgess

Crawford

et al . (2013) Sharing news, making sense, saying thanks: Patterns of talk on Twitter during the Queensland floods. Australian Journal of Communication 40(1): 23–39.

59.

Skeggs

(2005) The making of class and gender through visualizing moral subject formation. Sociology 39(5): 965–982.

60.

Smith

(1759) The Theory of Moral Sentiments. London: Printed for A Millar, A Kincaid and J Bell.

61.

Stenvoll

Svensson

(2011) Contestable contexts: The transparent anchoring of contextualization in text-as-data. Qualitative Research 11(5): 570–586.

62.

Stewart

(2017) Academic Twitter and academic capital: Collapsing orality and literacy in scholarly publics. In: Thomson

Lupton

Mewburn

(eds) The Digital Academic: Critical Perspectives on Digital Technologies in Higher Education. London: Routledge, 79–93.

63.

Surowiecki

(2004) The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies, and Nations. New York: Doubleday.

64.

Thelwall

Kappas

(2014) The role of sentiment in the social web. In: Von Scheve

Salmella

(eds) Collective Emotions: Perspectives from Psychology, Philosophy, and Sociology. Oxford: Oxford University Press.

65.

Uprichard

(2012) Being stuck in (live) time: The sticky sociological imagination. Sociological Review Monograph, Live Research Methods ( Back

Puwar

(eds)) 60(1): 124–138.

66.

Veletsianos

Stewart

(2016) Discreet openness: Scholars’ selective and intentional self-disclosures online. Social Media + Society 2(3): 1–11.

67.

Van Der Nagel

(2017) From usernames to profiles: The development of pseudonymity in internet communication. Internet Histories 1(4): 312–331.

68.

Von Scheve

(2018) A social relational account of affect. European Journal of Social Theory 21(1): 39–59.

69.

Watkins

Uhder

Pichinevskiy

(2014) Grateful recounting enhances subjective well-being: The importance of grateful processing. Journal of Positive Psychology 10: 91–98.

70.

Weller

Bruns

Burgess

et al . (2014) Twitter and society: An introduction. In: Weller

Bruns

Burgess

et al . (eds) Twitter and Society. New York: Peter Lang, xxix–xxxviii.

71.

West

(2015) Responding (or not) on Facebook: A sociolinguistic study of Liking, Commenting, and other reactions to posts. PhD Thesis, Georgetown University, USA.

72.

Wetherell

(2012) Affect and Emotion: A New Social Science Understanding. London: SAGE.

73.

Yuan

Zhong

Zhang

et al . (2016) Who will reply to/retweet this tweet? The dynamics of intimacy from online social interactions. In: Proceedings of the ninth ACM international conference on web search and data mining, New York, ACM, pp. 3–12.

74.

Zappavigna

(2012) The Discourse of Twitter and Social Media. London: Continuum.

75.

Zappavigna

Martin

(2018) #Communing affiliation: Social tagging as a resource for aligning around values in social media. Discourse, Context & Media 22: 4–12.

Empathy Rituals: Small Conversations about Emotional Distress on Twitter

Abstract

Keywords

Tweeting Emotion

Goffman’s Platform Performances

Methodology: Following Feelings Online

Empathy Interchanges on Twitter

Saving (Digital) Face

Extract 1

Extract 2

Extract 3

Extract 4

Extract 5

Surfacing Norms

Conclusion

Footnotes

Acknowledgements

Funding

Notes

References