#Online harms or benefits? An ethnographic analysis of the positives and negatives of peer‐support around self‐harm on social media

BACKGROUND
There is emerging evidence of the potentially detrimental impact of social media on young people's mental health. Against this background, online self-harm content has been a recent focus of concern across academia, policy and the media. It has been argued to encourage or even cause acts such as self-cutting through mechanisms of contagion. However, little is known about why a young person might engage with such content or about its impact on behaviour or well-being.


METHODS
Online ethnographic observation of interactions around self-harm on Twitter, Reddit and Instagram: collection and analysis of 10,169 original posts and 36,934 comments, both written and pictorial, at two time-points in 2018 and 2019. Ten in-depth semi-structured interviews exploring engagements with self-harm content on social media.


RESULTS
Our data show that peer support is the central component of online interactions around self-harm. Young people accessing such content are likely to already be self-harming; they may turn to social media to understand, and seek help for, their actions and feelings in a context of offline stigma and service support gaps. This paper engages with the mechanisms, complexities and impact of this peer-support, reflecting on the benefits and dangers to caring for oneself and others through social media.


CONCLUSIONS
Self-harm content is a fraught issue at the centre of current debates around risks and opportunities for child and adolescent mental health in the digital age. Whilst the importance of supporting young people's online safety is clear, moves to eradicate self-harm content must be undertaken with caution so as not to cause unintentional harm. Our research highlights a need to think beyond a model of contagion, instead attending to other mechanisms of harm and benefit. In so doing, it challenges prevailing attitudes towards online communication about self-harm and accepted approaches to managing this.

In a 2019 letter to Facebook, Instagram, Twitter, Snapchat, Pinterest, Google and Apple, the UK Secretary of State for Health and Social Care Matt Hancock suggested that online self-harm content 'leads to self-harm and promotes suicide' (Lumley, 2019). In asking these providers to tackle such content, his words echoed a wider framing of it as causing acts such as self-cutting and burning. Particularly attributed to the perceived harmful impact of 'graphic' (BBC, 2019a) imagery of selfharm wounds, this attribution of causality led the UK government and others to call on Instagram to block access to 'graphic content,' which it did in 2019 (BBC, 2019b).
Academic discussions have also argued that social media may normalise, glamourise or reinforce offline behaviours, such as self-harm and suicide (Lewis & Baker, 2011;Whitlock, Powers & Eckenrode, 2006). The role played by social media in 'spreading' self-harm through 'contagion' (Lupariello et al., 2019) has been discussed in terms of a potential copycat influence exerted in particuar by viewing self-harm images (Arendt, Scherr, & Romer, 2019). This model of impact draws on wider understandings of an element of social contagion to self-harm itself (Jarvi et al., 2013), observed in adolescents (Nock et al., 2009) and young adults (Yates et al., 2008), for example. Social contagion has also long been discussed as a factor in suicide (Daine et al., 2013;Haw et al., 2013). Analyses have noted the influence on suicidal behaviour in adolescents of the media (Dunlop et al., 2011) as well as exposure to suicide clusters (Insel & Gould, 2008), with social learning posited to drive imitations (Jarvi et al., 2013). That media reporting of suicide has the potential for contagion underpins the World Health Organization's guidelines for reporting suicides in the media (WHO, 2008). Against this background, it has been suggested that an important component of understanding online portrayals of suicide is to examine the extent of contagion with Carlyle et al. (2018) recently arguing for the application of the WHO suicide guidelines also to social media. However, a model of social contagion may be assuming certain harms whilst blinding us to other, more nuanced, mechanisms of harm on social media. It may also too linearly frame the relationship between social media and self-harm or suicidal ideation as working in one direction: from social media to self-harm/suicide. Understanding the selfharm or suicide content necessitates also considering the other direction too: from self-harm or suicidal ideation to social media. This is supported by a recent review of internet use and suicide attempts in adolescents. The authors found an independent association between 'problematic use of social media/internet and suicide attempts in young people' but concluded that 'the direction of causality, if any, remains unclear' (Sedgwick et al., 2019). There is therefore a need to contextualise motivations for seeking online content in offline lived experiences. This points to the potential applicability of a model of 'assortive relating' (Joiner, 2003) to a consideration of self-harm content, which would challenge a model of contagion and align with the wider literature on mental health communities online.
The recent UK All Party Parliamentary Enquiry into Young People's Mental Health (RSPH, 2019) acknowledged the emotional support and sense of community and belonging that social media can engender. This has been shown in relation to mental health (Hanley et al., 2019;Naslund, Aschbrenner, Marsch, & Bartels, 2016;Tucker & Lavis, 2019) , with groups formed around a range of conditions, such as eating disorders (Lavis, 2016). It has been suggested that social media offers a space where people can gain a sense of belonging (Brown et al., 2018;Dyson et al., 2016), share their experiences (Rodham, Gavin & Miles, 2007;Whitlock, Powers & Eckenrode, 2006), reduce feelings of isolation (Ziebland & Wyke, 2012) and challenge stigma around mental health diagnoses (Webb, Burns, & Collin, 2008). These potential benefits have been considered in relation to self-harm content (Duggan et al., 2012;Dyson et al., 2016;Seko et al., 2015;Shanahan et al., 2019) and that pertaining to suicide (Carlyle et al., 2018). Sustained communities of support formed around self-harm have existed since the advent of the 21st century (Winter & Lavis, 2020a), and it has been suggested that acceptance, belonging, intimacy and community found online may all mitigate destructive desires (Daine et al., 2013).
However, with notable recent exceptions (Shanahan et al., 2019), much of the literature that has explored these potential positives of online self-harm interactions draws on studies of more static entities such as websites and forums rather than the fragmented, heterogeneous and dynamic current landscape of social media (Baker & Fortune, 2008;Jones et al., 2011;Smithson et al., 2011). Evidence regarding the harms and benefits of young people's engagements with self-harm content on social media is therefore very limited. There also remains a lack of knowledge regarding why and how young people engage in online self-harm discussions and what they post (Shanahan et al., 2019).
In seeking to fill these gaps in the current literature, this paper aims to contribute to the emerging evidence base regarding opportunities as well as risks in the relationship between social media and adolescent mental health. This is crucial to a consideration of how to sensitively tackle real, rather than assumed, harms and to harness positives in the development of appropriate and effective interventions, offline or online.

Methods
This paper draws on a 15-month study (2018-2019) funded by the Wellcome Trust, which comprised online ethnography on Twitter, Reddit and Instagram, at two timepoints -in May 2018 and January 2019 -alongside semi-structured interviews. We initially intended to conduct one round of data collection, but a second was decided upon to examine the extent to which selfharm content had been altered by Instagram's banning of graphic content.

The appropriateness of online ethnography
Underpinned by the interpretive qualitative framework of medical anthropology (Lambert & McKevitt, 2002), which emphasises the social, cultural and structural dimensions of a given phenomenon, online ethnography was employed. This is a well-established method, suitable for engaging with online interactions and communities, and is adaptable to textual or visual content (Hine, 2000;Winter & Lavis, 2020b). It offered an appropriate way to explore self-harm interactions as contextually embedded in social media users' offline lives, mental health experiences and social worlds.
Scoping the existing literature during study design suggested that the contradictions in the current evidence base regarding whether self-harm content is helpful or harmful are, in part, methodological; scholars have often explored one social media site or online space without comparing content both within and between these. Or, they have analysed images without attending to the accompanying comments (Shanahan et al., 2019), which may give rise to different interpretations (see Winter & Lavis, 2020b). Analysing a single online community or one form of expression risks siloed understandings that do not reflect the diverse ways that young people may engage with self-harm content.
We therefore collected data across Instagram, Twitter and Reddit., These platforms were selected to gather data representative of different communities, conversations and forms of expression around self-harm: Typically, Instagram is pictorial, Reddit is text-based, and Twitter comprises short posts. On Instagram, we collected and analysed pictures as well as the conversations that unfurled from each post. Reddit facilitates sustained discussions where people share their indepth experiences, often with the same group of people, over time. Whilst primarily text-based, there is some sharing of images here too. Subreddits are web-forums within Reddit formed around a particular topic, and we collected all those generated around the topic of self-harm. The shorter forms of expression, due to limited character counts, on Twitter provided insights into how interactions occurred within these confines. Across the platforms, all forms of expression were collected and analysed, such as memes, gifs, textual descriptions, emojis, emoticons and textual actions (e.g. sending hugs or kisses).
Identifying and defining self-harm content As our interest lay in the ways in which self-harm is defined and discussed across social media, we sought to engage with any imagery and conversations that pertained to self-harm, not limiting data collection to clinical definitions or our own assumptions regarding what this term means. We therefore first mapped the landscape of self-harm content by inputting the hashtags 'self-harm' and 'self-injury'. We explored the content that this produced across the three sites, noting the original post, associated conversations and all hashtags placed alongside these original ones. This set in motion a cyclical process of continually searching for and re-entering the associated hashtags, tracing these back and forth across the three platforms, noting misspellings, colloquialisms or vernaculars. This process revealed interactions about self-harm that were initially hidden, such as those attached to hashtags that ostensibly would not denote self-harm. Although commonly attached to self-harm related discusions, these hashtags also lay outside the content blocks on platforms such as Instagram.
The research team then reviewed all the hashtags and subreddits to identify and select those most prevalent across the three platforms, balancing this against a need for maximum variation to the data collected.

Ethnographic data collection
Once hashtags, subreddits and key terms had been selected, all data joined with these were collected at two time-points: May 2018 and January 2019. 10,169 original posts were downloaded, alongside all the comments accompanying each post, both textual and visual, which amounted to 36,934 comments. Self-harm hashtags often appear alongside those denoting other mental health experiences, such as anxiety or eating disorders and sometimes suicide. Posts were included whether only tagged with each selected self-harm hashtag or whether this appeared alongside others. For each post included in the study, all the accompanying comments were analysed, whether explicitly 'about' self-harm or not, to analyse the conversations that coalesce around self-harm content. From each post's comments, we also identified new hashtags/ terms/subreddits, continuing data collection through an iterative process of re-inputting these to explore new content. At each time point, we reached a point of saturation, whereby adding further hashtags or key terms provided no further insight into online discussions of self-harm. Throughout ethnographic data collection, we also recorded observations, in the form of fieldnotes, of interactions in real time. These noted common linguistic tropes and interactional dynamics. The research team compared their observations of the online communities throughout data collection to reduce potential researcher bias in interpretation.

Data management and analysis
The diversity of online forms of expression collected called for an analytic technique adaptable to different media, including imagery. Downloaded content was therefore analysed using iterative thematic analysis (Braun and Clarke, 2006). Concurrent with data collection, data from each hashtag (across the platforms) were analysed independently. We ensured that the original posts and comments remained associated with each other in the management of the data to maintain conversational context. After each period of data collection had ended, triangulation occurred, with analysis conducted across the hashtags to look for comparisons, patterns and disconnects in content, motivations and forms of expression.

Identifying posts by young people
In this research, we employed the WHO definition of young people as aged between 10 and 24 years. Whilst it is difficult to ascertain the exact age of each social media user, a key part of data analysis was an assessment of whether posts were by young people. This was determined through the content of conversations (e.g. whether conversations included descriptions of being in high school or college, homework, parents and living at home) and, if available, bios of the poster, as well as the language and linguistic tropes of posts (e.g. the presence of slang terms used primarily by young people). This latter method was modelled on the 'language of age' defined by Schwartz et al. (2013), whose study demonstrated linguistic distinctions between age groups on social media.

Semi-structured interviews
Alongside ethnographic data collection, semi-structured interviews were conducted with ten young people who use or have used social media to engage with self-harm content. Participants were recruited through an invitation posted on a range of social media platforms. The inclusion criterion was current or previous use of social media to seek or post self-harm content. For seven interviewees, this was current, and for three, it was prior. All the participants were female and aged over 18.
Three interviews were conducted over email, two on WhatsApp, and five via Skype. They were designed to last approximately an hour but were often longer at the direction of participants. Topic guides were semi-structured, with participants being invited to talk about any aspects of their experience of social media and/or self-harm they felt it important to highlight to the researcher. In line with the epistemological framework of the study, this aimed to maintain an openness to participants' concerns throughout data collection. It engendered a reflexive process of co-examination between participant and researcher of why and how people engage with self-harm content in the wider contexts and meanings of their lived experiences. Like the online content, interview data were also analysed using iterative thematic analysis.

Triangulation of the two data sets
Conducting and analysing the interviews alongside collection and analysis of the online data allowed triangulation to take place in which themes in one data set were tracked across the other. Being careful not to unethically impute similarities, this enabled us to trace the multi-directionality of relationships between offline self-harm experiences and social media engagements. The findings presented in this paper derive from this analysis of both data sets. Whilst the quotations that are included derive from interviews in order to adhere to our ethical stipulations not to include direct quotations from social © 2020 The Authors. Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for Child and Adolescent Mental Health.
Self-harm and social media media, they illustrate the key findings that have emerged across both data sets.

Ethics
The study received ethical approval from the University of Birmingham (ERN_17-1386). There are ongoing ethical debates about the use of social media data in research, with questions regarding the boundaries between 'public' and 'private' information contested (Henderson, Johnson & Auld, 2013). We defined social media posts as public if they were visible without needing to sign in.
Consideration of ethics also extends to the presentation of findings; it has been suggested that a study's data usage needs to reflect the level of confidentiality, privacy and traceability that social media users may reasonably expect (BPS, 2017). Whilst young people engaging with self-harm content know that posts are publicly visible, self-harm is a sensitive topic and those posting about it may already be vulnerable. Many may turn to online communities to seek support, disclosing their experiences in a place where they feel safe and anonymous. Therefore, to respect their privacy and the intimacy of the information shared, the paper contains no direct quotes or photographs from social media or any users' names. Where quotations from online data are included in speech marks, these are terms that are frequently used across social media rather than attributable to a specific user. Additionally, to reduce the traceability of our data and adhere to the ethical approval no subreddits, search terms or hashtags are noted.
Also, to respect the anonymity of social media users, we did not collect demographic information from the interview participants except a statement of being over 18 on the consent form; this was key to participants agreeing to take part. The presentation of findings therefore does not include any identifiable information. Interview quotations are also not attributed to a specific participant in order to render them unrecognisable to one another as well as a reader. informed written or audio recorded consent was obtained from all interview participants.

Results
From offline to online: motivations for seeking selfharm content on social media Many recent cross-societal discussions of self-harm content have framed this as potentially causing selfharm. However, data from both the online ethnography and the interviews strongly suggest that selfharm commences prior to seeking such content: I think there's a misconception that people will, like, get into self-harm because they see these pictures [online] and I really don't think that's true. (Interview Participant) The social media conversations we analysed frequently included what might be termed 'origin stories'; young people recount having begun to selfharm and then outline their reasons for turning to social media. A moment of realisation that what they were doing was 'self-harm' is often framed as a key motivating factor for going online; they wish to understand their actions, especially in these first stages of self-harming. I would use social media to reach out or ask questions on sites such as Yahoo, just to confirm that I wasn't a 'freak' like people would tell me. It was a call for help. (Interview participant) As is clear from this quotation, internalised stigma can pervade quests for self-understanding and information. The frequency of words like 'freak' and 'loony' in posts elucidate the prevalence amongst users of being misunderstood offline. Entire threads are dedicated to recounting others' reactions to selfharm.
Extract from fieldnotes, 2018: Throughout the data gathered so far on Reddit and Instagram people describe how negative interactions offline have led them to go online -frequently writing that their parents saw their self-harm and an argument ensued. The responses show the frequency of such experiences with individuals explaining that their parents or 'normies' (meaning someone who does not self-harm) also 'don't understand,' unlike those online, they say. The original posts are emotive, often showing a moment of crisis in which, after such arguments, people go online to seek support not only for their self-harm but also to cope with the reactions of others.
One subreddit with a teenage community (as shown by the experiences and spaces mentioned, e.g. school, parents) focused on reactions from parents. Online participants recounted being threatened with being thrown out of the house or admitted to hospital.
That parental help-seeking is described as a 'threat' links to narratives of experiences with healthcare professionals, which are replete with descriptions of feeling 'dismissed', 'ignored', or 'upset' across social media. In line with this, one of our interviewees was told she was a 'silly girl' by a GP, and another remembered a nurse in A and E telling a doctor, 'oh she's always here.' Importantly however, there are also many posts attached to self-harm hashtags describing an, often desperate, desire for professional support. Resonating throughout these are descriptions of encountering limited resources and waiting lists. We observed threads in which young people sought information on how to receive professional support for their mental health, asking for ways to be 'boosted' up waiting lists or 'taken seriously.' Such posts lead other young people to respond with suggestions for what to say to a GP to increase the chance of a referral to specialist mental healthcare. Social media also fills a gap between primary and secondary services; young people poignantly describe having been referred to child and adolescent mental health services (CAMHS) and then facing long waiting lists with no intervening support. In relation to this, one interviewee said: Rather than saying, like, 'if you ask for help, the help is out there', that's not entirely true so instead, we should say 'if you ask for help, the help may be difficult to get' but that's not a reflection on how much you need it, it's a reflection of how tight services are at the moment and how resources are being diverted to other areas, it's not your fault and it's not something that you should have to tackle on your own'. (Interview Participant) Thus, a clear motivation for seeking online selfharm content is to gain the support of other young people with shared experiences in the context of offline service lacks: As a teenager I spent all my free time searching for help and support online because I just didn't have a healthy outlet or anyone to talk to. I was desperate to find people who could explain what was going on and tell me what I needed to do because I felt so lost and had no idea. (Interview participant) There is an ebb and flow to this support-seeking; the day-to-day need for a support 'framework'an available community who listensis interlaced with moments of urgent need incited by the act of selfharm itself or a feeling of being in crisis. Posts attached to a variety of hashtags ask for help because a young person has an urge to cut, or less frequently, burn her/himself. Support to not go through with self-harm is sought, as is reassurance and comfort in the face of an un-resistible urge. The complex mix of emotions that accompany self-harm, such as shame, relief, sadness and disappointment, is apparent. I was really stressed and just trying not to self-harm and, sort of, it was good to have acknowledgement of what I was experiencing. And to be told that my emotions were real even if I didn't act on them. (Interview Participant) A young person may also turn to social media in a moment of extreme emotional crisis, such as to say that 'everything is falling apart now.' All these moments illustrate the immediacy of online support, with users being able to receive instant responses to their distress.
It is important to note that support seeking is therefore not limited to self-harm itself. Young people frequently use self-harm hashtags to recount distressing life experiences, both historic ones that may have been factors in the development of their self-harm and others from ongoing day-to-day life. Or as one interviewee put it, when asked why she engages with self-harm content: it's not so much talking about what you are doing, but talking about the context that you are doing it in.
As such, whilst self-harm related hashtags have widely been framed in media discussions, for example, as simply attached to images or descriptions of self-harm, they emerge clearly from our data as contextualised in young people's lived experiences. It is against this background that the next sub-section will explore the forms that peer-support takes on social media.

Online interactions: giving and receiving peersupport
The previous section has outlined how social media offers a young person a space to gain support from other young people with similar experiences. Crucially, this peer-support is not simply one component of the discussions attached to self-harm hashtags; rather, it is their central lynchpin.
I think a lot of people like the reason they go to these groups is because they want to know that they're not alone and that it matters what happens to them. And having someone else acknowledge what you're going though and to say that they care about you and to show that they see you, it helps a lot to feel like you're, like you matter, and you're not just drifting through the world disconnected. (Interview Participant) Peer-support takes various forms. Counter to wider perceptions of users encouraging one another to injure themselves, attached to self-harm hashtags, are often exchanges of information around how to reduce the urge to self-harm or to replace the act: Extract from Fieldnotes, 2018: Visual techniques -Reddit and Instagram: There are images of arms and legs covered in drawings/paintings of flowers, butterflies, or cartoon animals with captions discussing how this painting on the skin is used as a physical and visual coping strategy.
Distraction techniques -In response to 'I don't know what to do' or other crisis moments, we see people suggesting reading a book, talking to someone, or going for a walk Sensory techniques -Across all three platforms we are reading posts about how to replace the physical sensation of self-harm with other less 'harmful' techniques. Holding an ice-cube on the skin, snapping an elastic band, or drawing red lines with a permanent marker.
In response to the crisis moments or sharing of distress discussed above, across Twitter, Instagram and Reddit, others also respond to the poster by asking if they would like to talk privately, such as on a messenger, by sending virtual 'hugs' or emojis (particularly hearts to show affection towards to the poster), or by offering to virtually 'sit' with them in real time. Although, when seen in isolation, these acts might seem small or throwaway, they are frequently described as 'lifesaving' and, after the crisis has abated, thanks are exchanged. As such, seemingly small acts like sending an emoji must be recognised as meaningful enactments of care that forge valued relationships across distances: There are some people I've connected with who I'd consider friends, and it all starts from someone reaching out, or prolonged/frequent interactions. (Interview Participant) Through this relationality, sustained support communities are built up. This is particularly apparent from frequent discussions on social media of users who have been quiet for an unusual amount of time: Extract from Fieldnotes, 2018: Reddit -Someone in the community posts 'Has anyone heard from ***, do you know if they're ok?' Responses then discuss the last time they spok to the individual. Later on in the subreddit we see a post from the user explaining why they were away, which was due to a hospital stay. The responses from the community are welcoming, 'good to have you back', 'glad you're ok'. This contrasts to users' descriptions of the negative interactions in their offline lives.
The emotional reciprocity that underpins these supportive communities that form around self-harm hashtags emerged clearly from interviews: I try to make sure to go out of my way to reach out to people who seem like they are struggling, to try and be someone they can vent to or talk to, and I try and share some of the coping techniques I find work for me if I feel that it's appropriate for what they are going through. Additionally, if people have written something that is very emotional or personal to them, I will leave positive comments and check up on them via private messages. (Interview Participant) It is against this background of the various forms of peer-support that our data clearly show the need to re-assess the meaning of graphic imagery and recognise its connection to support-seeking.
'Graphic imagery' describes, as one interviewee put it, 'people posting right after they've literally torn apart their arms'. It comprises photographs of selfharm wounds, usually on the arms but sometimes on other parts of the body such as the legs. In our data set, still imagery is more prevalent but there are also videos of self-cutting. Both were primary targets of Instagram's content ban in 2019, in which the platform blocked access to the hashtags it deemed most likely to be attached to such content. Although this meant that certain, more 'obvious', hashtags no longer existed during our second round of data collection in 2019, those which used terms less obviously about self-harm remained in use and still had graphic content attached to them. In tandem, the banning of the more obvious hashtags had removed the supportive discussions that we had seen attached to those hashtags in 2018. This is because, whilst such graphic imagery is visually shocking, it is crucial to recognise that it sharedand after the ban continues to sharethe same online spaces and hashtags as peer-support.
Importantly, graphic imagery is often posted to ask for support when this may not feel articulatable in words. Some young people post photographs of their own self-harm but others re-post found images of the self-harm wounds of anonymous others. This latter underscores the need to think about how this imagery functions as well as what it depicts. It is used to alert other young people online to the need for help. The discussion above noted that self-harm hashtags are attached to discussions of distress and lived experiences, not just the act of self-harm. This is also the case in relation to the support-seeking enacted through posting a graphic image, with support sought not only for self-harm but also to the contexts that give rise to it. Posting a graphic image that is not even one's own does not necessarily even signify that an act of self-harm has taken place but rather that a young person feels at risk of injuring themselves or in great distress.
Across all three platforms, the imagery leads others to ask how the poster is, opening up an opportunity to be listened to: People do it [post graphic content] because they want someone to accept them [. . .] they want someone to take them seriously. If you are a regular self-harmer people stop caring. Or, not stop caring but people become so desensitised to it that it's no longer a crisis, it's just everyday [. . .] they want someone who isn't quite so accustomed to it who will say 'oh my god, let me support you'. (Interview Participant) However, as this interviewee makes clear, whilst a young person posting such imagery may be motivated by a need for support they feel unable to ask for verbally, there is a risk that such imagery, as well as the conversations that contextualise it, may deleteriously impact others; the next sub-section will explore this.
From online to offline: the value, impact and ambivalence of peer-support The value of the peer-support to young people engaging with self-harm content is clear. Across social media, there are articulations of gratitude for having been listened to or helped. I think just having people respond when you're feeling down, like kind of helps to confirm that I am a real person. That others care about me. (Interview  Participant) Extract from Fieldnotes, 2018: We have continually seen images of self-harm and how others respond to these with supportive comments, emojis, and questions such as "are you okay?" Often the original poster responds by saying how they were feeling at that moment or how they are doing now. The images open up a space for them to then share what they have been going through, and for others to listen. Throughout these threads the original poster thanks individuals "for listening to me." Such peer-to-peer interactions can potentially save a young person's life in a time of crisis: One of my online friends she was going through a really bad suicidal crisis. I guess about 4 or 5 years ago. And I would send her messages like when things were getting rough and then just talk to her, or like even just send her cute pictures of cats to distract her. (Interview Participant) However, set against the positives of receiving support, our data illustrate that providing support to another young person can be a heavy responsibility to bear. The reciprocity described above as core to the creation of sustained communities of care, can be difficult, with interviewees echoing statements also found on social media by describing feeling they are 'not helping enough'.
Coupled with this, there is a clear sense across our data sets that listening to another's distress or painful life story can impact a young person's mental health, and potentially re-trigger an urge to selfharm: I think the option of putting like a content warning and filtering posts that you want to see or don't want to see is also very helpful because I wouldn't like if I was having a rough day I wouldn't just bring that up to anyone but if I go online I can put a content warning and then people who feel able to read about it can read about it, and if they don't and they don't have to worry about upsetting them. (Interview Participant) As this participant suggests, across Instagram, Twitter and Reddit posts are tagged by the posters themselves with TW or Trigger Warning, to give others the option to view the post or scroll past it. This elucidates the self-surveillance and self-censorship undertaken to protect others within the community. Crucially, such warnings are not only attached to selfharm imagesto 'graphic imagery'but also to textual descriptions of distress such as accounts of 'having a rough day' as this participant puts it. This shifts, and extends, a narrower understanding of triggering as specifically linked to viewing self-harm imagery, which inheres in discussions of 'graphic' content as causal; it is the distressing story of another person that emerges from our data as triggering, whether that is depicted or described.
Even where seeing photographs of another person's wounds precipitates an episode of self-harm, this is described by interviewees as ensuing from the reading of another's distress from their wound. Participants describe feeling overwhelmedand therefore triggeredby this distress, not by the image per se. In describing how she does not find anonymous imagery triggering because it is 'separated from a person that you feel you know,' one interview said of tumlr: tublr [is] probably the worst of the worst in terms of self-harm content because you follow people. It's not just photos where you can disconnect from someone. The thing about tumblr is that it's a story, it's a narrative. It normalises it because you see it's part of someone's life [. . .] you feel much more of a connection. (Interview Participant) It is, however, important to attend to this interviewee's mention of normalisation. Whilst not necessarily causal, self-harm content, both textual and pictorial, may have the effect of entrenching or exacerbating selfharm. One interviewee noted that to maintain online support, it can feel like you need to keep showing that you need it, which can make self-harm worse; she said, 'everything becomes more and more extreme if you want someone to notice and the Internet can perpetuate that cycle', and she gave this as a reason for posting increasingly graphic imageswhether your own or found imagery. Engaging with self-harm content also, another interviewee said, 'can isolate people and make them think that they're the only people who can understand', which has the risk of precluding offline help-seeking.
This risk is also posed by a further form that support takes. Alongside the coping strategies described above, there is a plethora of information around 'how to self-harm safely'. Advice around cutting 'safely' is shared, such as how to sterilise a blade or what to put in a first aid kit with items including bandages, and alcohol wipes noted. We also followed conversations that unfurled from young people posting that they wished to cut on their back or other areas of the body more difficult to reach and therefore care for; respondents unanimously discouraged this due to safety concerns. Likewise, in response to imagery of cuts that may be deemed 'graphic' due to their severity, there are comments such as 'that's too deep' or 'that needs stitches', with encouragement to go to A and E. Whilst such advice may 'show you different ways of selfharming', as one interviewee put it, like the coping strategies it needs approaching contextually in analysis.
That 'how to self-harm safely' is framed by those offering and receiving such advice across our two data sets as a form of support links back to the motivation of going online to seek understanding from other young people. Such advice is part of the mutual recognition on social media of the meaning of self-harm; young people acknowledge that for one another it may be deeply meaningful, functioning as a coping strategy in the face of unendurable distress or painful life events: I think it's important not to shame or judge people if they do self-harm. There's no right way to deal with your emotions and it's a big coping mechanism for some people. (Interview Participant) It is also this recognition of what self-harm means to one another that underpins the support that is seen in response to a young person articulating that they have self-harmed: People have posted when they've relapsed with selfharm, and I've seen a lot of support for that as well. Not in the sense of encouraging it, but in the sense that we understand it happens and that recovery is never straightforward. Having somewhere judgement-free to talk about self-harm can be really helpful, I think, as it can still seem so taboo to people. (Interview Participant) This does not, however, preclude the widespread congratulations across platforms that greet a young person who posts that they have managed to resist the urge to self-harm, whether two hours or two years 'clean': Extract from fieldnotes, 2018: In the title of their posts people are writing about the length of time that they have been clean, such as "3 years clean!" with a description about their journey to get there. It could be one day clean, with discussions of how this short period without self-harming is an achievement, or accounts of not having self-harmed for years. Whatever the period of time, others respond "That's amazing", "well done", and "keep going". I am proud that I have been clean from self-harm for over two years now. I have posted about it on more than one occasion, though. I think it's important to let people know that recovery is possible and that it isn't a cut and dried thing. I may not have indulged in my urges, but I still get them quite a lot. (Interview Participant) Thus, to understand the meaning and impact of self-harm content, as well as young people's motivations for engaging with this, it is imperative to acknowledge its contexts and complexities; 'graphic' content, 'safety' advice, general chat and emotional support all share the same spaces, appear side-by-side in the same conversational threads, and are attached the same hashtags. The discussion section will now unpick these complexities and examine what they tell us about the positives and negatives of peer-support around online self-harm content.

Discussion
In contrast to widespread understandings of a causal relationship from social media to self-harm, this research has shown that young people accessing online self-harm content are likely to already be selfharming. Aligning with a model of 'assortive relating' (Joiner, 2003), this demonstrates the need to recognise how motivations for engaging with self-harm content are embedded in pre-existing offline contexts of distress and/or mental health difficulties. Stigma, both societal and internalised, is a clear motivation, as are offline service gaps, such as waiting lists for child and adolescent mental health services. This latter echoes a wider recognition of the value of the 'immediacy' of social media, especially at crisis moments (Tucker & Lavis, 2019). Whilst online and offline support seeking may overlap, the former is filling gaps in the latter. Moreover, although engaging with self-harm content can hinder offline help-seeking, this is not an inevitable outcome. There are myriad posts expressing a, sometimes desperate, desire for professional support. Responses include signposting towards services, as well as advice regarding what to say in order to 'be taken seriously' by professionals. These conversations also highlight how the two key motivations of stigma and offline gaps combine in the stigmatisation that young people recount having experienced in healthcare services.
These findings illustrate the need for a franker consideration of how self-harm is responded to in service settings, as well as by society more broadly. They also question the emphasis of recent advice to psychiatric professionals to ask young people about their social media use, such as by the Royal College of Psychiatrists (RCPSYCH, 2020). Rather than placing the focus of such discussions on the impact of social media on a young person, more reflection needs to be given to how and why they go online, and therefore what their support needs and lacks are.
Underpinning this clinical implication is the finding that central to self-harm content is peer-support; both seeking and offering support emerged from this research as a key reason for looking for, posting and commenting on self-harm imagery and text. Young people turn to social media to gain an understanding of their self-harm and gain help from others with similar experiences. That they may first do this in the early stages of self-harm indicates a potential opportunity for intervention. Crucially, this would need to replace the valued tenets of peer-support; social media offers a young person in distress the opportunity to be listened to without judgement, and to have painful, intimate and stigmatised experiences validated.
Previous literature shows self-harm to potentially have multiple embodied meanings, differing between people (Csordas & Jenkins, 2018), and a recent analysis of YouTube videos focused on self-harm recovery uncovered multiplicity also in understandings of self-harm (Ryan-Vig, Gavin & Rodham, 2019). Our data sets, in contrast, contain little discussion of the functions and meanings of selfharm, whilst instead illustrating a clear mutual recognition that it has meaning. What is played out across posts are the complexities of living with, or perhaps 'through' (Lavis, 2018) self-harm, and the felt need to have these recognised by others. Against the background of wider lacks in understanding and stigma in society, young people may feel that their experiences can only be understood by others online. Such discussions also suggest there may be a lack of space to explore or tackle these complexities and the ambivalent meaning and value of self-harm in a formal context of services.
There is, thus, depth and intimacy (Andreassen et al., 2018) to the relationships that form around self-harm content, with communities of care functioning as 'lifelines' in times of crisis as well as offering sustained support day-by-day to aid young people to cope with both self-harm and what may underlie it. This finding aligns with work that has recognised the benefits of online interactions in relation to self-harm (Baker & Fortune, 2008;Shanahan et al., 2019;Smithson et al., 2011). That young people use social media platforms to both give and receive peer-support has also been found in suicide research (Luxton et al., 2012;Sindoni, 2019). However, our data point to the need for further investigation of the differences between online suicide and self-harm discussions in terms of motivation, content and impact.
Whilst self-harm is a known risk-factor for suicide (Chan et al., 2016), suicide and self-harm are recognised to be distinct behaviours (Jarvi et al., 2013), each with complex aetiology. Yet cross-societal discussions of social media have often blurred their differences, tending to refer to self-harm and suicide content as synonymous and seeing both as conduits to suicide (BBC, 2019b;Lumley, 2019). Whilst there are overlaps, with hashtags sometimes placed side-by-side, these two bodies of content, and the communities that form around them, are distinct. Motivations for seeking self-harm content, and the peer support that forms around it, relate to an ongoing ebb and flow of distress, not necessarily to moments of crisis or immediate risk, or even to acts of self-harm. It is necessary to understand this in order to delineate the harms and benefits of such content.
A key component of self-harm content that has been largely considered in terms of suicide risk, but that urgently needs a contextual reconsideration is graphic imagery. We have shown that this cannot be dislocated from peer-support, with such images functioning as signifiers to alert others of a young person's need for help. That some of the photographs posted are found images rather than depictions of young people's own self-harm further underscores the need to recognise their interactional function as 'conversation starters'. This circulation of found photographs also challenges an assumption that sharing self-harm imagery is necessarily a competitive act. However, that the need to sustain peer-support can lead some young people to post increasingly graphic textual or visual content aligns with Brown et al.'s (2018) finding that, on Instagram, pictures depicting severer wounds received significantly more comments. As such, there is a competitive element but its underlying mechanisms are different from what might be assumed; crucially, it illustrates the desperate dynamics of need that are played out on social media.
The function of graphic imagery elucidates that moves to ban it, and self-harm content more broadly, without very careful consideration risk increasing the isolation and vulnerability of young people, rather than protecting them. If moving to eradicate this content, platforms and policy makers need to consider how to replicate its help-seeking function in an alternative form. Moreover, this study has demonstrated the limits to current moves to patrol self-harm content through blocking hashtags. That some of the more common self-harm related hashtags returned no results during our second round of data collection after Instagram's ban highlighted the loss not just of the imagery present the year before but also the peer-support. However, it remained possible to find the full spectrum of content, including graphic imagery, by searching for less obvious hashtags. Hashtags, then, are unreliable predictors of the positives or negatives of online content. This is also illustrated by the unexpected ramifications of peer-support, which warrant consideration by providers and policy makers in relation to how best to support young people.
In the context of mental health, peer-support has been defined as 'a system of mutual giving and receiving where individuals who have faced and endured the adversity of mental illness can offer hope, companionship, and encouragement to others facing similar conditions' (Naslund, Grande, Aschbrenner, & Elwyn, 2014; see also Mead, Hilton & Curtis, 2001). Reciprocity, in terms of the sharing of distressing experiences and mutual support, has previously been noted to be a key component of responsive peer-support (Naslund et al., 2016). Yet, providing support to others can pose difficulties to young people, particularly through the emotional impact of hearing another's distress. Young people posting and seeking self-harm images or discussions do not necessarily wish to communicate about selfharm itself. Rather, they want to find a place to, as one participant put it, 'talk about your life a bit'. Whilst this means that social media is replete with mundane stories of everyday life, these do mingle with much more emotive narratives of pain, abuse and bullying, for example.
It is in relation to distressing narratives as well as graphic imagery that trigger warnings are widely used across our data set. This shows how informal dynamics of moderation are part of peer-support and also that there is recognition that one's story may be distressing to another young person, even to the point of potentially triggering their self-harm. This challenges a narrower understanding of triggering as specifically linked to viewing self-harm imagery, posited by discussions of 'graphic' content as harmful. Whilst reciprocity is a clearly valued benefit of peer-support, it also has perhaps the most potential for harm. This challenges current underpinnings of policy and practice, raising urgent question regarding how to facilitate peer-support in a way that is safe for young people. That the triggering effect of others' narratives has also been noted in relation to moderated spaces, such as Mind's Elefriends (Tucker & Lavis, 2019) challenges moderation as a reliable means of ensuring the safety of social media users and points to the need for further research. That 'triggering' is framed as the re-inciting of existing self-harm rather than initiating it also elucidates a broader need to reconsider the potential for, and meaning of, contagion in relation to selfharm content, and its interplay with protective factors.
Underpinned by concerns of social contagion, the WHO (2008) suicide reporting guidelines suggest exercising caution in using photographs or video footage, avoiding repetition of stories about suicide or describing the method of a completed or attempted suicide. Yet, as this paper has elucidated, imagery, descriptions of self-harm methods and stories are all part of self-harm content. Their harms and benefits are, however, more nuanced than might be assumed at first glance. To not recognise this may serve to blind us to unexpected dangers and also impute harm where in fact content may be protective. Many types of content and forms of interaction are explicitly described across our data sets as 'supportive' and 'caring', even where this might seem surprising from an outside perspective. Notable here is the descriptions of methods that inhere in information regarding 'how to self-harm safely.' Social media platforms have reshaped how young people search for health-related information (Moorhead et al., 2013) with analyses of the 'medicalisation of cyberspace' (Miah & Rich, 2008) highlighting how online information exchange can run counter, and pose a challenge, to clinical discourses (Fox et al., 2005;Lavis, 2016). Exchanging information on 'how to self-harm safely' stems from the mutual recognition among social media participants of the meaning, and even ambivalent value, that self-harm may have to one another. Like the coping strategies also exchanged, this safety advice signifies an informal harm minimisation approach (Pengelly et al., 2008). Whilst it may normalise, exacerbate or entrench existing behaviours, it can also serve to mitigate these; what looks most dangerous may thus be protective. This counters calls to apply the WHO suicide reporting guidelines to self-harm content on social media (Carlyle et al., 2018), instead suggesting the need for policy makers and platforms to consider each element of this content in a contextual and sensitive way, with young people's own viewpoints placed at the heart of these reflections.

Limitations
Whilst we have attempted to offer a nuanced and upto-date analysis of self-harm content, the landscape of social media is dynamic and quickly changing. As such, it is clear that there is a need for ongoing research to keep apace as research papers will become out of date swiftly.
It is also the case that, whilst we did not impose a definition of self-harm on data collection, much of our ethnographic data relates to cutting. The largescale mapping of the field conducted before data collection almost uniformly unearthed narratives of cutting. This begs the question of whether we missed discussions of different methods such as burning or whether young people are utilising alternative forms of support. This warrants further research. We also did not examine cultural or geographical differences, as we looked at English language hashtags and content, and these aspects also warrant further investigation.
Although a core facet of data collection and analysis was the assessment of the age of the posters, it remains possible that some posts were from older social media users. As communities around selfharm date back approximately 20 years (Winter & Lavis, 2020a), participants to these spaces are socially varied. As such, future research would do well to explore differences in terms of harm, benefits, impact and vulnerabilities across social and cultural groups, with a focus on an older demographic warranted.
That all ten interview participants were female indicates our difficulty in recruiting young men, and it is a limitation of our study. Whilst this does to some extent reflect the fact that more young women than men self-harm (McManus et al., 2019), it is also clear from our ethnographic data collection that young men do engage with online self-harm content. However, the themes arising from the interviews closely mirror those of the online data set.

Conclusion
This study has offered an in-depth understanding of what is helpful and what is harmful in communications around self-harm online. It has aimed to build the evidence base for recommendations for policy makers, platforms and public health approaches to prevention and intervention.
It has been suggested that 'cutting has become a new moral panic about the dangers confronting today's youth' (Gilman, 2012(Gilman, : 1008, with social media identified as one of those dangers (Arendt, Scherr, & Romer, 2019). Framing online self-harm content as causal, however, locates 'blame' for selfharm in young people themselves through processes of contagion. It therefore allows us to avoid examining the wider social and structural contexts that give rise to both self-harm and social media use, and to ignore that social media is a mirror of society.
In seeking to, instead, take account of these wider contexts, this paper has placed the positive and negatives of peer support in the context of young people's lives and needs. It has highlighted the offline service gaps, as well as lack of societal understandings of self-harm that can lead young people to seek support from one another. It has also shown that giving this support can both benefit and detrimentally impact a young person's well-being. Understanding these complexities to how young people care for themselves and each other through social media is crucial to forging the evidence base for future technological or clinical interventions. Whether these seek to make social media safer or to harness the benefits from the existing peer-support dynamics, a more nuanced approach to online safety is imperative. Placing pressure on platforms to remove self-harm related content is inadequate and any moves to limit such content must be undertaken with caution in order not to cause unintentional harm.
In challenging a modelof contagion, the study also has implications for how we, across society, approach suicide discussions on social media. But it also strikes a note of caution, showing the need to consider self-harm and suicide content separately, each on its own terms. Further research is needed to understand their overlaps and differences in order to formulate policy and practice approaches to both. This imperative also extends to reflections on mental health content online more broadly. It is crucial to understand this is a less siloed way, reflecting how a variety of hashtags are used as conduits into a space of listening to the distress that underpins diagnostic labels. It is this distress that we need to be bear in mind when considering social media use, in policy and practice, and in society.