Keeping users experiencing a suicidal crisis safe online: Current text-based practices of professional online mental health forum moderators

Introduction

Individuals experiencing suicidal thoughts and/or behaviors (STBs) often have difficulty accessing appropriate and/or timely mental health support (Lundström, 2018). The World Health Organization (WHO) recognizes access as an issue and actively promotes online mental health services to address unmet psychosocial support needs (WHO, 2020). The 24/7 accessibility and potential anonymity of online services enables support to be available when and where it is most needed and can play a critical role in preventing an individual from taking their own life (Prescott et al., 2020).

For those experiencing STBs, there are two broad types of online support that they can access. One is formal such as e-therapy services that are provided by mental health professionals (e.g., BetterHelp Online Therapy). The second is informal and is typically peer focused and led (e.g., “Side by Side” the online community of Mind, a UK based Mental Health Charity). Here peers interact anonymously with each other, giving and receiving support based on their shared lived experiences (Smith-Merry et al., 2019). This online peer interaction may be overseen by a moderator whose role it is to keep forum-users and the online space safe (Kendal et al., 2017). Online spaces are places where suicide and suicidal thoughts can be openly spoken about without fear of censure or shame. They are places where suicide is no longer a taboo subject. Moderators who oversee the safety of these spaces can be unpaid peer volunteers with lived experience or employed as moderators whereby they are often professionally qualified, holding tertiary level qualifications in mental health and/or having completed in-house training (Smedley & Coulson, 2017).

Despite the popularity of these online informal spaces for talking about suicide, there are concerns about whether these spaces and interacting within these spaces are safe for forum-users, regardless of whether a moderator is present or not (Nathan & Nathan, 2020). These concerns are often around whether talking openly about suicide, suicidal thoughts, plans and attempts may in fact be triggering, and therefore, exacerbate risk. Despite professional moderators being employed with the express intent of keeping forum-users safe (Kendal et al., 2017), a recent scoping review concluded that little is empirically known about how online professional moderators actually keep forum-users and online forum communities safe (Perry et al., 2021).

Perry et al. (2022) sought to address this gap by interviewing professional moderators about their perceptions of working online with individuals in a heightened state of suicidal desire. They found that professional online moderator work is perceived as complex yet constraining. The complexity is in dealing with the uncertainty that the online environment brings and managing at-risk users using only written communication whilst simultaneously dealing with other users and forum enforcement/monitoring tasks. The constraining is that managing risk is limited to risk identification and encouraging at-risk users to either take steps to keep themselves safe or to contact external crisis services. Moderators therefore have this multifaceted role in which they not only work with those in a heightened state of suicidal desire, but simultaneously with other users and the space to ensure the safety of all (Perry et al., 2022). This is unlike other suicide crisis response services where other potentially at-risk individuals typically do not need to be considered when responding nor does the service space itself need to be monitored. In this sense, keeping safe online is conceptualized as not only identifying at-risk users but also ensuring risks such as vicarious traumatization and suicide contagion are minimized through early risk identification and the “removal” of the at-risk user from the online space so that they, other users, and the space itself are kept safe (Perry et al., 2022).

Online safety for individuals experiencing STBs has traditionally been conceptualized through a user lens, with researchers juxtaposing user risk against forum-user benefits. Risks include the potential for forum-users to learn more about suicide methods (Mokkenstorm et al., 2020) and consequently engage in offline risky behaviors (Kvardova et al., 2021). Forum-user benefits include a reduction in suicidal thoughts (Dodemaide et al., 2019) and an increased sense of acceptance (Mokkenstorm et al., 2020). A potential drawback of juxtaposing risk against forum-user benefits is that it adopts a retrospective, user-focused approach, that draws upon the tension between potentially uncontained hazards and forum-user benefits that are linked to peer support. Unsurprisingly, what has emerged from the literature is conjecture about whether online spaces are in fact safe places for people to interact with others and talk about suicide. This has resulted in online spaces being seen on one hand, as places that “do not necessarily pose a risk to participants” (Mok et al., 2015, p. 703), and on the other hand, as places where there is “significant potential for harm from online behaviour” (Marchant et al., 2017, p. 2).

A different way of examining if online spaces are safe places to talk about suicide is to look at the intersubjectivity of safety. Intersubjectivity refers to the shared understanding between two or more interactants where this shared understanding is displayed through how interactants orient to what has been said/written (Raymond, 2019). Orient in this context refers to how interactants respond to what has been said/written (Heritage, 1984). Intersubjectivity enables researchers to consider how professional moderators understand a forum-user is at risk and how they interact with these users in these online spaces to accomplish safety. This is a pivot away from a focus on user risks or benefits and toward considering how professional moderators engage and interact with forum-users to keep forum-users safe online. This is important because people experiencing STBs who go online for suicide-related reasons have been found to have elevated levels of suicidality when compared to people who do not go online for such reasons (Bell et al., 2018; Harris et al., 2009; Mok et al., 2016; Niederkrotenthaler et al., 2017; Wong et al., 2021). Further, they are also more likely to make a future suicide attempt (Harris et al., 2009; Mok et al., 2016; Niederkrotenthaler et al., 2017). Given these individuals may already be in a heightened state of suicidal desire (Klonsky et al., 2021), it is critical that we understand how such forum-users are kept safe in that moment when they may be at most risk of taking their own life. In this sense safety becomes a process that is managed in-situ between professional moderators and forum-users.

To do this a safety guiding framework that focuses on how online spaces are made safe is needed. This necessitates the adoption of different theoretical and methodical approaches to researching suicidality (O’Connor, 2021), that includes approaches that are prospective and observational rather than retrospective and reflective. A safety guiding framework requires taking an in-situ interactional perspective that allows for the intersubjectivity of safety to be explicated. In this way, researchers can identify how safety is achieved in place through moderators interacting with forum-users during moments of heightened suicidal desire.

Moving toward understanding how safety is achieved online complements previous suicidology research. It does this by allowing us to better understand how the benefits of these spaces are realized and accomplished (Cherba et al., 2019), and conversely how risks are minimized (or increased) through online interactions. As a starting point in considering the intersubjectivity of online safety, we have focused on online spaces that proport to be safe spaces due to the presence of professional moderators. Further, we have looked at how professional moderators keep forum-users safe during moments of heightened suicidal desire.

We are not aware of any in-situ research that has explored the text-based working practices professional online moderators use to keep forum-users experiencing STBs safe online. There is research that has examined how online volunteer moderators work with individuals in a state of suicidal distress (e.g., Gilat et al., 2012) or what moderators do (e.g., remove inappropriate posts, Mokkenstorm et al., 2020). In terms of in-situ work, there is research that has explored how forum-users provide support to at-risk users (e.g., Bjärehed et al., 2023; Westerland, 2013). However, using forum-user and volunteer moderator research as a basis for professional moderator research is problematic. Forum-user findings may not easily translate to moderators given the different knowledge rights and responsibilities associated with both groups (Heritage, 2012). Volunteer moderators differ from professional moderators not just in terms of the absence and presence of a formal employment relationship respectively (Matias, 2019), but also in their role descriptions and therefore work online. Volunteer moderators focus more on content moderation and adherence to forum or online site rules. Professional moderators, whilst also having this focus, do more than this. They focus on identification of risk and the management of this risk as well (Perry et al., 2021). As noted by Perry et al. (2021), moderator research has also typically focused on what moderators do or the work tasks they perform to keep users safe. Given these differences, research on how volunteer moderators work may not fully capture the more nuanced work practices of the professional moderator.

Previous research has suggested that professional moderators wished that they could be “stepping in more” and using more of their counselling skills to assist forum-users who are in a heightened suicidal desire state (Perry et al., 2022). The present study extended on these findings by investigating how moderators work to keep forum-users safe online. That is, we sought to identify the text-based actions professional moderators undertake to keep forum-users safe through analyzing forum posts between professional moderators, at-risk users, and other forum-users and corresponding off-forum emails between professional moderators and at-risk users. The questions that guided the analysis were:

RQ1: When do professional moderators intervene?

RQ2: How do they intervene?

RQ3: What happens following the intervention?

Methods

The data in this study was collected from an online professionally moderated mental health forum run by a large mental health organization. Forum-user and post numbers are not reported by individual forum, however, over 38,000 members interact across the forums run by the organization with there being nearly one million posts each year. To post on the forum an individual needs to sign up and become a “member” of the organization, however an individual does not need to be a member to read forum posts. Posts are therefore able to be read by anyone. Membership is restricted to a geographical location and to individuals aged 18 years or older. Posts are not limited in terms of number of characters, with emoticons but not images being able to be included in posts.

University Ethics Committee approval was gained prior to data collection commencing and permission to use publicly available forum posts was granted by the organization who also provided additional moderator-user emails that were not publicly available. When professional moderators become aware of a forum-user who is in a heighted state of suicidal desire, they can intervene by posting on the forum, which is visible to all forum-users, and/or follow up by sending an off-forum (private) email to the at-risk user. Thus, forum-users are kept safe through public moderator posts and private moderator-to-user, user-to-moderator emails.

Initial email data consisted of over 10,000 emails dated June 2014 to December 2020. Emails were identified by the mental health organization using Microsoft 365 software to filter and export emails, as per the instructions provided by the authors (copy of the instructions is available upon request). The first author began by reading the most recent emails (December 2020) and worked backward until at least 100 pages were identified as this has been argued to be of sufficient volume for a text-based interactional study (Wiggins, 2017). The data corpus was constructed by first searching this reduced email data using keyword terms associated with suicidal behaviors (i.e., suicide, die, death, had enough, end it, hopeless, don’t want to be here, nothing left, the end, no point, accepted death, at peace with death) and then locating the forum thread associated with the identified email. This was done by using the date and time stamp of the emails, and the subject line of the email, which was the same as the name of the forum thread.

Once emails and corresponding forum threads were identified, they were copied and pasted in chronological order into 34 individual Word documents for analysis. Each individual Word document represented one forum thread and any email/s associated with the thread. Bringing the emails and associated forum posts together in this way allowed us to analyze what was occurring in each forum thread, and then what happened when an interaction between a professional moderator and a forum-user moved from the public thread space to the private email space, and at times, back to the public space. The final data corpus spanned a 12-month period and consisted of 178 pages of data comprising 617 posts from 34 forum threads and 56 related emails that featured 86 individual forum-users and 16 moderators.

The use of online forum interactions as data and research sites comes with specific ethical considerations. There is no agreement around whether online data is public or private or some combination of both or if consent should or in fact can be gained from “participants” (Ravn et al., 2020). Given the sensitive nature of interactions contained in the data corpus, all personal and any other potentially identifying information have been removed from post/email data included in this paper. Additionally, the forum from which data were collected and the mental health organization itself have not been identified. We acknowledge that this does not guarantee forum-user anonymity however it provides as much anonymity as possible whilst also maintaining the fidelity of the data. Membership to post on the forum entails agreeing to terms and conditions that includes an acknowledge that any forum-user data provided on forums or to the organization may be used for research and evaluation purposes. Whilst consent was not sought from members whose data was used in this study, as previously noted the mental health organization did give consent for forum data to be used in this study. Consent to use moderator forum posts and private emails was gained from the moderators themselves. In addition, we engaged in a backward search process by putting each excerpt in this paper into Google’s search engine with no forum post or forum-user being identified.

Through the data analysis process, we became aware of three types of forum-users present on the forum. There were new/occasional, regular contributors, and experienced forum-users. According to forum-user guidance notes located on the forum, experienced forum-users play a specific role which involves explicitly giving and receiving support and thus modelling the reciprocal nature of peer support on which the forum is founded. Some of the tasks that the experienced forum-users engage in are welcoming new forum-users to the forum, posting in quieter forums threads to promote interaction, and alerting the professional moderators to forum-users who may need more than peer support to be safe.

Data analysis drew upon Conversation Analysis (CA), as it is a method that has been specifically developed to study social interactions (Sidnell, 2009). It is interested in how talk is socially organized and how people make sense of one another to accomplish interactional tasks (Albert et al., 2018). It seeks to explicate “patterns of interaction that exist independently of any one instance of interaction” (Kempf-Leonard, 2004, p. 507). There are other methodological approaches that have been used to explore online suicidality interactions such as discursive psychology (e.g., Horne & Wiggins, 2009; Lundström, 2018; Wiggins et al., 2016), discourse analysis (e.g., Timm, 2018), critical discourse analysis (e.g., Westerlund, 2013) and thematic analysis (e.g., Bjärehed et al., 2023). Conversation analysis was chosen for this study because of its focus on understanding sense-making in naturalistic talk, the organization of this talk, and on how this sense-making is made visible by interactants (Sidnell, 2009). Thus, CA has an interactional practice focus that best suited the research question this study sought to answer.

Although CA has traditionally focused on naturalistic and spoken interaction (Wiggins, 2017), research attention has increasingly been given to synchronous and asynchronous digital/text-based interactions (Paulus et al., 2016). This has raised concerns about whether CA foundations and findings derived from spoken interactions have relevancy to such interactions (Meredith, 2019) and that perhaps a different form of CA should be used to analyze digital data, Digital CA (Giles et al., 2015). For the purposes of this study, following the recommendations of Jucker (2021), we adopted previously adapted, modified, and repurposed aspects of CA that kept true to the foundations of CA rather than adopt a pure Digital CA approach. These aspects are outlined below.

Recipient designed refers to how interactants design their turns at talk for each other (Sacks et al., 1974). In synchronous off-line interactions there is typically one intended recipient, thus the turn at talk is designed for them. In asynchronous online forum interactions, forum posts can be both specifically designed for a particular recipient (the original poster) and for other unknown recipients (other forum-users; Meredith & Potter, 2014). Thus, turns in online spaces are multi-recipient designed (Giles et al., 2015).

Sequential organization refers to not only how what is said is shaped by the previous turn but also how this then sets up the context for what will immediately follow (Schegloff, 2007). Given utterances or posts on online forums are asynchronous they can be re-visited and read in any order. A forum-user response to an at-risk post may not only be informed by that post but also by what other forum-users have posted previously. However, a forum-user response may also not be shaped by these in between turns. Whilst still context shaping, there may not be an immediate sequential ordering to forum responses (Herring, 1999). What this means is that how CA understands adjacency pairs (Schegloff & Sacks, 1973) needs to be reconsidered for in online interactions this adjacency is disrupted. Which reply is responding to which post is more difficult to identify in asynchronous online interactions (Herring, 1999).

Analysis begun by identifying critical moments in the data corpus. These were moments whereby others on the forum identified that a forum-user was in a state of heightened suicidal desire and at-risk based on what they had posted. This moment was usually identified by other forum-users rather than moderators. To identify these moments, the data corpus was read several times for the first occasion where risk was identified in a thread. The post/s prior to that occasion were then read along with proceeding posts to identify the social action of the interaction or what was the online interaction doing. Action was identified by looking at how other forum-users or moderators responded to the at-risk user’s post, how they responded to each other, and how the at-risk user in turn received these responses. These receipts, responses and the critical moment post where then examined to identify moderator practices (i.e., the phenomenon of interest). Analytic focus was on how moderators responded at these critical moments, be this on the forum itself or via private email to the at-risk user. This was done for each critical moment after which patterns of moderator text-based activities were identified by comparing across different critical moments. These patterns were accounted for by examining what was occurring within each instance. This accounting is reflected in the results below.

Results

Analysis identified that professional moderators transform the “risk-talk” in an at-risk user’s post into “safety-talk” that benefits both the at-risk user and other forum-users who may find the thread content triggering. Moderators do this by using a pattern of interaction that involves them entering the critical moment, aligning their response to the subjective risk, and transforming this subjective risk into intersubjective safety. Other forum-users, typically experienced forum-users, support the above actions by affiliating with the at-risk user, therefore, contributing to the transformation.

Whilst the total data corpus was analyzed, three interactions from the corpus have been selected that best illustrate the above analytic claims. These are now presented with each focusing on slightly different aspects of the pattern and other forum-user support. Interaction 1 demonstrates the key features of the pattern using the at-risk user’s own orientation to the safety transformation as evidence for the pattern as per CA’s next-turn proof procedure principle or how the participants make visible their understanding of what was written through how they respond (Hutchby & Wooffitt, 2008). Interactions 2 and 3 further illustrate the pattern but include affiliation work of other forum-users. These interactions display how other forum-users recognize and orient to the safety pattern and thus the transformation work that is being accomplished by the professional moderator.

Interaction 1

This interaction opens with a troubles-telling (Jefferson et al., 2015) post from a regularly contributing forum-user. Troubles-telling refers to interactions where there is a sequential order of actions in which the teller talks about their trouble and where the “listener”, here professional moderator or other forum-user, responds to this telling. The following interaction comprises two forum posts and two off-forum emails and comes from an ongoing forum thread consisting of approximately 6,906 forum posts. The participants in the interaction include a professional moderator, two experienced forum-users, and a regular contributor forum-user, who in this interaction, is oriented to as the at-risk user.

Excerpts 1, 2, 3, and 4 illustrate the identified safety interactional pattern. The first two excerpts are posts with the other two being emails. Excerpt 1 is from the at-risk user with Excerpt 2 being the first reply post to the Excerpt 1 forum post and was made by the professional moderator. Excerpt 3 is the off-forum email that the professional moderator sent to the at-risk user, and Excerpt 4 is the at-risk user’s off-forum email reply.

The extracts are presented as they were written but not as they physically display on the forum website. For text that has been redacted or changed to assist with anonymity, “{ }” brackets have been used. Unlike offline interaction transcription, the use of (word) does not indicate a lack of clarity around what was said or who said something. Instead, it reflects the text and bracket usage as posted by forum-users or moderators. Like standard CA transcription presentation, we have used line numbers for ease of data referencing however we have not used letters in the excerpts to indicate speaker identify given the way we have presented the data. Instead, we have placed the identity of the speaker in the excerpt heading.

Excerpt 1: User1 Post

1. Swallow me now please. Too many changes, too much lost. I am struggling to see a way
2. out of here.

In Excerpt 1, the regularly contributing forum-user troubles-telling indicates that the un-named changes and loss they are experiencing are overwhelming (line 1) and that there is little hope of their current situation changing (lines 1–2). This combined with the words “swallow me now” (line 1) that start the post, are indicative that the regularly contributing forum-user may be experiencing a heightened state of suicidal desire.

Excerpt 2: Professional Moderator Post

1. @{user1 in-crisis}, I’m going to drop you an email to check in.

In Excerpt 2, the professional moderator enters the interaction with the first reply to the original post. In these situations where a professional moderator has oriented to the forum-user as being at-risk there are two actions they can engage in. They can interact with the at-risk user on the forum or take the at-risk user, and thus the risk, offline. In this first reply, the professional moderator directly addresses the at-risk user via the “@” tag and advises that an off-forum email will be the next interactional activity. What is interesting here is that despite the perceived risk, the professional moderator does not attempt to relationally affiliate with the at-risk user. That is, they do not display empathy for, or endorse the stance of, the at-risk user. Instead, the professional moderator aligns (see line 3) with the risk that appears to be present in Excerpt 1. They do this by focusing on the safety action of checking in with this user off the forum. By aligning in this way, they take up the interactional role of professional moderator rather than that of other forum-user.

Alignment and affiliation are CA concepts that are indicative of how people in interactions cooperate with what has been said or written (Steensig, 2019). They are critical to the mutual understanding of any interaction and for progressing interactions. Alignment works at a structural level in that it projects understanding of the unfolding interaction with this projection being done through the acceptance and uptake of in-context roles. Affiliation is affective in nature working to support the stance of the speaker/writer through displays of empathy or agreement. Moreover, affiliation can be seen as an act of social solidarity (Lee & Tanaka, 2016).

While this post is addressed to the at-risk user, it is visible to all forum-users. For this reason, the professional moderator post serves a dual purpose. The first informs the at-risk user that their post has been read and that next interactional actions should occur offline. For this to occur the at-risk user needs to move from the forum to their email inbox as this is where the professional moderator wishes to interact. The second purpose is to inform other forum-users that the professional moderator is aware of the potential risk and is taking steps to address it. These dual-purpose posts contribute to the intersubjective of safety.

Excerpt 3: Professional Moderator Email

1. Hey {user1 in-crisis},
2. Sounds like you’re really struggling at the moment so I just wanted to let you know
3. you’re not alone. It really is a time of change and uncertainty, and so many people are
4. impacted quite heavily at the moment. Do you have someone you can speak to right
5. now? Are you worried about your immediate safety?
6. As you know we aren’t a crisis service but you can call any of the following services
7. for immediate support:
8. {Organization name and number redacted}
9. {Organization name and number redacted}
10. {Organization name and number redacted}
11. If in immediate danger: {number redacted}
12. Let us know if you’re safe for now, and I hope you keep posting on the forums tonight
13. and get some support. 
14. Warmest regards,
15. Moderator

Excerpt 3 is the professional moderator’s off-forum email that was signaled in Excerpt 2. Across the data corpus it is common for professional moderators to initially reply to at-risk users by posting on the forum, and then send a private email. Although in two instances, professional moderators reversed this process by first sending a private email to the at-risk user and then posting publicly in the forum. The sequence of forum post-private email or private email-forum post does not alter the interactional pattern.

We can see the professional moderator aligning with the subjective risk on line 2 where they project their understanding of the at-risk user’s current situation using the same word the at-risk user had previously used, “struggle”. Transformation occurs across lines 4–5 and 7–12 interspersed by alignment to the risk on line 6. The professional moderator moves from the topic of subjective risk to the topic of intersubjective safety on lines 4–5 by asking two polar questions of the at-risk user to ascertain their level of safety regarding whether they are currently alone and can they keep themself safe. In the data corpus, professional moderators rarely ask questions of forum-users in forum posts, and in the instances where they do, it is to gain more information about the level of risk that the at-risk user is experiencing thus allowing them to shape the transformative nature of their risk to safety message.

The professional moderator explicitly aligns to the presented risk on line 6 by projecting to the forum-user what in-context roles they cannot accept or take-up, here a crisis worker, as “we’re not a crisis service”. The transformation resumes on line 7 through suggesting a safety action the at-risk user can take and then provides information to facilitate this in the form of crisis organization phone numbers (see lines 8–11). The transformation to intersubjective safety does not stop there, it continues with the professional moderator asking the at-risk user to confirm their safety status and to go back to the forums as needed (see lines 12–13).

Excerpt 4: User 1 Email

1. Hi moderator, I am safe, my ex-husband has temporarily moved back in to monitor me

Using the next-turn proof procedure principle, the transformation from subjective risk to intersubjective safety is evidenced by the at-risk user’s email response. The response displays that they have moved from talking about risk to talking about safety having acted on the moderator’s safety talk presented in Excerpt 3, lines 4–5. As the at-risk user confirms their safety, this completes the interaction from the perspective of the professional moderator, with no further action being required on their part to reduce risk and maximize safety. We know this as the professional moderator sends a brief one-line email (not shown here) to the user acknowledging the safety reply and then there is no further interaction. Acknowledging forum-user safety was not always present across the data corpus.

What Excerpts 1–4 display is a safety transforming action. They illustrate a pattern of text-based interaction that professional moderators use to keep forum-users safe online. What underpins this pattern is alignment with the risk rather than affiliation with the user, and transformation of talking about risk to talking about safety. To make this clearer, professional moderators do not display empathy or emotional understanding, they do not recognize the affect the at-risk user has displayed. They focus on moving the at-risk user from risk to safety. In Interaction 1, this transformation was accomplished through a series of safety actions and safety requests in an off-forum email.

Interaction 2

This interaction is made up of 4 forum posts and comes from a thread containing 38 posts. The participants in this interaction are a professional moderator, an experienced forum-user, and a regularly contributing forum-user who is presenting at-risk. Excerpt 5 is from the at-risk user, with Excerpt 6 being the first reply post by the experienced forum-user. Excerpt 7 is the forum post from the professional moderator to the at-risk user. Excerpt 8 is the forum post reply of the at-risk user to the professional moderator. Excerpts 5, 6, 7, and 8 demonstrate the pattern for keeping users safe but this time analytic focus is on how other forum-users affiliate with the risk and the at-risk user.

Excerpt 5: User 2 Post

1. Seriously what’s the point

Excerpt 5 is the first post in the forum thread and may seem somewhat ambiguous in terms of risk given the absence of the at-risk user presenting or telling their troubles. This ambiguity or vagueness is a key feature of what Horne and Wiggins (2009) argue to be actively suicidal in immediate threat posts. As in the first interaction, it is in the next turn-proof procedure where the actual risk is identified and understood.

Excerpt 6: Experienced User Post

1. I’m happy to see you @{user2 in-crisis}.
2. I know you can’t see any point at the moment. But you are worthy of a good life (everyone
3. is)…. and things will get better if you can just hang on.
4. @Moderator

It is in this response that we see a forum-user orient to the poster as being at-risk through their affiliation on lines 2 to 4. The first affiliative action is to introduce an affective state into their reply clearly indicating social solidarity with the at-risk user using “I’m happy to see you” (see line 2). Line 3 starts with a statement of empathy, “I know” indicating to the at-risk user that they understand and importantly have shared what the at-risk user is currently experiencing. In this way, they signal that they know the poster is at-risk. Endorsing the stance of the at-risk user, using the same word “point” that the at-risk user used in their opening post, serves as a further affiliative action. By affiliating in this way, the forum-user is then able to offer support at the end of line 3 introducing the notion that this at-risk feeling is transient and that there is hope if the at-risk user can “hang on” (line 4). Whilst the forum-user has come in at a critical moment, unlike a professional moderator, they affiliate rather than align with the risk as there are no safety actions in their reply, only endorsement of stance and displays of empathic understanding.

This does not mean that the forum-user overlooks safety altogether, in fact they end their post by using the tag @moderator. This alerts a professional moderator that this interaction requires their attention and safety intervention (line 5). In this way the forum-user works alongside the professional moderator who is on shift when it comes to identifying and minimizing risk in the online space.

Excerpt 7: Professional Moderator Post

1. Hi @{user2 in-crisis} I can hear things are really tough right now. Do you think it might
2. be time to reach out to one of the following helplines?
3. {Organization name and number redacted}
4. {Organization name and number redacted}

In this excerpt, the professional moderator now enters the forum thread in response to the “@Moderator” tag. The professional moderator aligns to the subjective risk on line 6 projecting their understanding of the at-risk user’s current state reformulating the “seriously what is the point” (line 1) into “things are really tough right now” (line 6). The in-context role of professional moderator is signaled using “I hear” which may seem odd as this in a textual based interaction but is reflective of the paraphrasing that would occur in offline helping contexts.

The transformation from subjective risk to intersubjective safety occurs across lines 6–9. A polar question suggesting to the at-risk user’s that it is time to engage in safety actions occurs on lines 6 and 7 which is followed up with information that enables the at-risk user to take up these actions by providing them with crisis organization phone numbers (see lines 8 and 9). As this post was publicly visible, all forum-users could see it, and therefore, know that the professional moderator was working to ensure the safety of the at-risk user, relieving other forum-users of the responsibility of risk minimization. Like Interaction 1, the professional moderator did not attempt to relationally affiliate with the at-risk user, rather focus was on aligning with the risk and transforming talk from subjective risk to intersubjective safety.

Excerpt 8: User 2 Post

1. Eh...I’m in emergency getting medical attention. Really don’t see the point but eh.

Despite the professional moderator not requesting the at-risk user come back to the forum and report on their safety, the at-risk user does come back to the forum. Excerpt 8 is the at-risk user’s response to the professional moderator, advising that they are receiving medical attention and that they are moving toward being safe having acted on the professional moderator’s advice. Albeit that advice was to call a crisis line, they have physically presented for crisis intervention. What is interesting is that the at-risk user recognizes that they are still at risk finishing their response by recycling the “not seeing the point” reference from their original post (see line 1). This placed after the safety update and before qualifying the risk with the tag “but eh” signals to other forum-users and the professional moderator the importance of taking immediate steps toward being safe.

We note that this at-risk user continued to post in the forum thread advising other forum-users that they were continuing to move toward safety. Given the at-risk user had advised that they were seeking medical attention, combined with their further safety confirmation posts, the professional moderator did not engage further with the at-risk user. As the subjective risk had been transformed to intersubjective safety no further alignment from the professional moderator was required. This does not mean that affiliation also ceased. There can be affiliation in the absence of continued alignment (Steensig, 2019) and thus the intersubjective safety can been maintained through the continued social solidarity displayed by other forum-users to the at-risk user.

Extracts 5–8 display the affiliation actions that occur along with the professional moderator transforming safety action pattern. It illustrates how an ambiguous post from an at-risk user can be managed using affiliation from a forum-user along with the professional moderator’s safety pattern. Here, an at-risk user shares their troubles, albeit briefly and ambiguously, which is oriented to by a forum-user to infer that a heightened suicidal desire may be present, bringing about the tagging of a professional moderator into the interaction. This is not before they have affiliated through displays of empathy and affect and supporting the stance of the user. This allows the professional moderator, upon entering the interaction, to propose a safety action, resulting in the at-risk user moving away from risk and toward safety.

What is important to note is that there is no sequential order associated with professional moderator alignment, safety pattern transformation and user affiliation. Nor is there an order as to who must identify the risk first for the professional moderator to enter the critical moment. Other forum-users can tag moderators in so that they are entering a critical moment or professional moderators themselves can identify the risk and enter the interaction. Professional moderators can align first with other forum-users coming in afterward and affiliating. Conversely, other forum-users can affiliate first with the professional moderator aligning and engaging in the safety pattern afterwards.

Interaction 3

This interaction opens with a troubles-telling post from a new/occasional forum-user who is presenting at risk. The sequence of 6 posts comes from a forum thread of 9 posts and includes one email. Excerpts 9–15 are from this forum thread and show the intersubjectivity of safety. Excerpt 9 is the first post in the forum thread by a forum-user who is struggling with suicidal thoughts and desires. Excerpt 10 is the first reply to Excerpt 9 and is from an experienced forum-user who tags in a professional moderator. Excerpt 11 is the professional moderator post to forum and Excerpt 12 is the professional moderator email to the at-risk user. Excerpt 13 is the forum post response from the at-risk user. Excerpt 14 is a forum post response from a new/occasional forum-users who post providing support to the at-risk user.

Excerpt 9: User 3

1. Apart from the trauma it would cause my family, I really can’t see the point of even
2. trying anymore. I am not looking forward to my future, my past is a source of awful
3. trauma and anxiety to me, and I can’t even find a reason to get out of bed right now,
4. apart from the feed the dog. Letting the dog down is yet another source of anxiety. I
5. just want to curl up in a ball and die. I have everything planned and I think I’m just
6. waiting for one more bad thing to happen to justify it to myself.
7. I cannot bring myself to talk to someone about it face to face, and I don’t know why.

In Excerpt 9, a new/occasional forum-user presents themselves as in a heightened state of suicidal desire. Whilst references are made to family and a pet as protective factors (lines 1, 4), it is the past and ongoing pain (line 2–3) along with hopelessness (line 2–3), the explicit mention of death (line 5), and the signaling that they have both a plan and an increasing intent to end their life (lines 5–6) that is oriented to by a forum-user as indicative of the user being at-risk. This is the first post from this forum-user evident through the “welcome” response from the other forum-user. Despite being a first post, there is an absence of a greeting or self-introduction and the statement at the end of the post that they have no-one they can turn to, reads as if the at-risk user is experiencing a lack of social connectedness. This excerpt is presented as a deviant case example as across the data corpus it was uncommon for at-risk users to infer or explicitly state they had a suicide plan or indicate their readiness to activate their plan. In CA, deviant cases are used to check the validity and generality of specific instances of interest in conversational interactions (Lewis-Beck et al., 2004) and is being used as such here.

Excerpt 10: Experienced User 2

1. Hi, {user3 in-crisis}, and welcome. I’m glad you’ve reached out to the forums, as that
2. shows some spark of hope. Many people here can relate and understand what you’re
3. going through. I’m sorry to hear you’ve had such trauma in your life. I’m concerned
4. that you might not have enough professional support at the moment for your severe
5. depression...can I ask if you are on any medication for it, or seeing a counsellor?
6. You say you can’t talk about your SI (suicidal ideation) face-to-face, which makes it
7. hard. Is it easier to write about it, as you have done here? That way you could use the
8. {Names and contact details of crisis services redacted} would that be easier than face
9. to face? Thank you for posting here. Please stay safe, {@user3 in-crisis}. I’m also
10. going to call a @Moderator, as I’m concerned about you.

As in Interaction 2, it is a forum-user that orients to the poster as being at-risk. They begin their post by focusing on bringing back into the interaction a sense of social connectedness. This is evident through the words of welcome and acknowledgement of the at-risk user being new to the forum community (line 8). But it is the relational affiliation in this reply that signals to the at-risk user and others that a heightened state of suicidal desire has been identified. This is first evident through the expression of an affective state in line 8, “I’m glad” that works as an act of social solidarity. Affiliation continues on line 9, “many people here can relate and understand” and line 10 “I’m sorry” through displays of empathy. Stance endorsement as an affiliative action is evident through the recycled use of “trauma” (line 10) and “can’t talk about your SI face-to-face”.

It is only after affiliation has been established that the forum-user is able to expresses concern for the at-risk user and encourage them to stay safe (lines 10–12, line 16). This illustrates again the affiliation rather than alignment of risk position that forum-users take up, with alignment typically being the pejorative of the professional moderator. Whilst they point the at-risk user to an external service (lines 14–15), this is in the context of a possible solution to the at-risk user’s self-stated difficulties in talking about their issues face-to-face. Thus, like Interaction 1, the forum-user brings safety into the interaction through this reference and calling on a professional moderator (lines 16–17: “I’m also going to call a @Moderator, as I am concerned about you”). By tagging in a professional moderator at the end of the reply post, the forum-user signals that a different level of support is required. This indicates that whilst forum-users can share their own experiences, ask questions, and point at-risk users to both internal and external safety sources, they do so as affiliative rather than alignment actions. The services of professional moderators are required to transform the risk talk into safety talk, with other forum-users working alongside moderators in risk identification and minimization.

Excerpt 11: Professional Moderator 2

1. Hey there {user3 at-risk},
2. I’m the moderator on shift. I’m so sorry to hear that you are in so much pain and don’t
3. feel you have anything to look forward to. I can hear how hard it is to speak about this,
4. which is why it is so brave that you have reached out here. I am going to send you an
5. email to check in with you as I’m worried about you

The @moderator tag results in the professional moderator coming into the interaction to begin the risk to safety transformation. In this excerpt, the professional moderator begins their post by introducing themselves, indicating that they know the at-risk user is new to the forum community. What is different in this reply, compared to others we have seen, is that it also includes some affiliation activities (see line 19, “I’m so sorry”; line 20, “I can hear how hard”; line 21, “why it is so brave”). Whilst professional moderators tend to focus on alignment, with other forum-users focusing on affiliation, this does not mean professional moderators never affiliate or other forum-users never align. Each group can do both. In this excerpt, professional moderator affiliation appears to be reflective of the risk status indicated in Excerpt 9. The at-risk user has stated they have a painful and proactive event history (i.e., past, and continuing trauma and anxiety, lines 2–4) and they appear to be currently endorsing high levels of suicidal ideation (lines 4–5) and suicidal desire (lines 5–6). Given suicidal ideation is fluid and can rapidly escalate to a state of suicide readiness (Barredo et al., 2021), alignment along with affiliation would be appropriate professional moderator safety actions in such cases.

Like Excerpt 2, by notifying all forum-users that a professional moderator is working to ensure the safety of the at-risk user (lines 21–22): “I am going to send you an email to check in with you as I’m worried about you”), other forum-users are relieved of this responsibility. Moreover, it signals to the at-risk user what is to occur now in that the interaction is being moved to an off forum interactional space. This along with the affiliation activities work together to contribute to the intersubjective of safety among the at-risk user, other forum-users, and the forum itself.

Excerpt 12: Professional Moderator 2 Email

1. Hey there {user3},
2. I’m really sorry to hear that things are so difficult for you at the moment. You’ve shown
3. a lot of strength in coming to the forums and seeking help. It’s really important to talk
4. about how you’re feeling if you’re having thoughts of suicide like this.
5. Do you have someone you can speak to right now? Are you worried about your
6. immediate safety?
7. Unfortunately, the Forums aren’t a counselling or crisis service, however you can
8. contact any of the following services for immediate support. {Organization name and
9. number redacted} if it’s hard to speak face to face right now.
10. {Organization name and number redacted}
11. You don’t deserve to go through this alone. For lots of people who experience thoughts
12. of suicide, they can find it helpful to put together a safety plan. We really encourage
13. you to give it a go here {weblink redacted}. You might also like to have a read of this
14. resource on coping with thoughts of suicide {weblink redacted}. You’ve shown great
15. courage in reaching out for help & I hope you continue to do so.
16. You’re more than welcome to continue to post in the forums. However if you’re
17. concerned about your safety, it’s important to contact one of the numbers above or call
18. {number redacted} in an emergency.
19. Please take care of yourself and let us know if you’re safe for now,
20. Warmly,
21. Moderator 

Excerpt 12 is the professional moderator’s off-forum private email to the at-risk user. In this email, we see the continuation of affiliative actions recycling words posted previously (line 2) or re-worked words (brave into strong line 3). Transformation begins on lines 5–6 and then recommences on line 8 continuing until line 12. As previously demonstrated, subjective risk is moved to intersubjective safety through the professional moderator asking questions of the at-risk user to ascertain their level of safety through asking if there is someone they can reach out to and whether they are concerned about their own ability to keep themself safe (lines 5–6). Line 7 breaks the transformation sequence through alignment with risk bringing into the interaction the in-context roles the forum and by inference the professional moderator can and cannot take up. This is done by referencing what the forum is not, “a counselling or crisis service”. This alignment is followed by safety actions of connecting with specialist services and the giving of safety resource information. Transformation continues with the professional moderator encouraging the at-risk user to stay connected with the forum (line 16) and to confirm their safety status (line 19).

Excerpt 13: User 3 Post

1. Thanks all. It is serious. I’m not going to lie. I just feel low right now, instead of
2. suicidal, but I know the dip is going to come back again at some point soon, and that’s
3. when things get really scary. I have an appointment for a care plan on Thursday, but
4. worried I won’t make it that far. It’s weird, because I have felt much worse in the past,
5. but not so unsafe. I think it’s the quick swing to bad. That is new for me.

Excerpt 13 is the at-risk user’s forum response, and it is addressed to all. Transformation of subjective risk to intersubjective safety is evidenced by the at-risk user’s forum response. Thus, rather than emailing the professional moderator, the at-risk user has taken up the professional moderator’s safety advice and continued posting to the forum. By advising that the level of risk has shifted from suicidal to feeling low and indicating that they have an appointment for a care plan on Thursday (Excerpt 13: lines 28–30), the state of heighted suicidal desire is seen as having dissipated, and they are no longer unsafe for the moment. This talk about safety rather than risk means that there is no further action required by the professional moderator and they withdraw from the interaction. The absence of any ongoing professional moderator response further reflects the tendency of professional moderators to align rather than affiliate with risk. This occurs as the in-context role of professional moderators is to initiate a safety sequence of identifying risk and working to minimize it. Therefore, when risk is no longer present, professional moderators are not compelled to act or engage.

Excerpt 14: Occasional User Post

1. ((text omitted)) You are not alone. I’ve survived the same chaos, I know you can do the
2. same. You just need professional support ((text omitted))

Excerpt 14 is second forum-user response to Excerpt 13 and occurred 3 hours after Excerpt 13 was posted. This excerpt demonstrates that once a safety transformation activity is introduced by a professional moderator, other forum-users can assist with continued safety reinforcement (line 42: “You just need professional support”). But they do so by bringing back into the interaction, relational affiliation. Empathy is present through sharing that they and others have lived experience of what the at-risk user is facing (see line 41 “You’re not alone”) and that this has been overcome. By doing this it allows the other forum-user to engage in a safety reinforcement. Thus, other forum-users contribute to the intersubjective safety transformation by supporting these “risk” talk into “safety” talk actions through engaging in safety strengthening and empathy at the same time.

Like the previous two interactions a safety transforming action occurs across this final interaction. The excerpts in this interaction demonstrate that whilst the transformation pattern is one that is founded upon alignment with risk, to transform talk about risk to talk about safety, professional moderator affiliation was also present. This is most likely because of the concurrent suicidal ideation and suicidal desire on display that warrants high levels of attention and caution.

In summary, there is a pattern of professional moderator work that ensures the safety of online forums and those that interact in these spaces. This pattern seeks to transform the “risk talk” of at-risk users to “safety talk” for the benefit of those at-risk and the wider forum community. We have seen professional moderators taking up an in-context role that has a dual risk and safety focus that allows for individual forum-user safety needs to be balanced alongside online forum community needs. To achieve this, professional moderators enter forum threads at times of identified heightened suicidal desire, having either been called into a thread or having identified a risk themselves, to propose an immediate safety action/s. The safety transformation action is supported by other forum-users who engage in also moving the at-risk user from subjective risk to intersubjective safety.

Discussion

Four key professional moderator text-based activities have been identified. Professional moderators typically align rather than affiliate with at-risk users with the role of affiliation being taken up by other forum-users. This preference for alignment occurred in both professional moderator posts and their private emails. This is not to say alignment was an exclusive professional moderator action or affiliation an exclusive user action. Professional moderators did engage in affiliation and other forum-users did engage in alignment. Analysis indicates that alignment was more a professional moderator interactional activity, with affiliation being more a forum-user interactional activity. Once risk was identified by either professional moderator or other forum-user, the professional moderator engaged in sign posting activities. Asking the at-risk user to “look out” for the professional moderator’s email worked to inform other forum-users that the risk had been identified/confirmed and was being addressed. It also worked to signal to the at-risk user what next steps were needed from them. When interacting with at-risk users, professional moderators engaged in transformation text-based activities whereby subjective risk was transformed into intersubjective safety. This occurred both online for all to see as well as offline in private email interactions. Finally, there was a serial aspect to risk identification. Professional moderators typically posted to the forum first and then sent the at-risk user a private email once risk had been first identified.

The current study contributes to the broader online moderator research in the following way. This broader research has tended to focus on the role online moderators play (e.g., Skousen et al., 2020) rather than how moderators (professional or volunteer) enact that role. This is not to say research has overlooked this enactment or its importance. Deng et al. (2023) focused on this enactment by asking professional moderators to recall how they enhance peer support in online spaces. Professional moderators spoke of the various strategies they deployed to encourage forum-users to support each other and what they did as professional moderator to ensure the safety of the online space. They concluded that professional moderators are central to ensuring online spaces remain safe and supportive places. However, it needs to be noted that Deng et al. did not examine how these professional moderators performed this supportive and safety work. Despite their stated focus on how, in essence the researchers asked professional moderators what they did to keep the online space safe. Whilst they proport professional moderators can be a “shaping force” they did not analyze how they enact this “shaping force” in-situ. The current study has done this. The contribution of the current study has been to explicate professional moderator activities and practices as they occur in real-life. Based on our findings, it may not be the professional moderator “role” that keeps users safe do but rather how professional moderators interact with at-risk users that is critical for ensuring online forums are supportive, positive, and safe environments. Their “shaping force” can only understood through how they perform their work.

Based on the findings presented, we suggest that online mental health forums are elegant, complex, and imperfect public spaces for risk presentations to occur (Jefferson, 1988). Adding to this complexity, is the assertion that online mental health forums are not intended to be crisis services and are therefore not designed as suicide intervention or prevention spaces (Tucker & Lavis, 2019). However, as the current study demonstrates, in instances of heightened suicidal desire online mental health forums where professional moderators are present, such forums can move at-risk users closer to safety. In this way these forums do engage in suicide prevention work.

Professional moderators bring order to messy or in our case high risk situations (Kendal et al., 2017) and in doing so they maintain both a risk and safety focus. Our results have shown how professional moderators orient to the display of a heightened state of suicidal desire contained in a forum-user’s post, and how they engage in the initiation of a safety transformation pattern that is designed to keep the forum-user safe in and at that moment. This pattern consists of professional moderators entering interactions at critical moments (Perowne & Gutman, 2024) whereby they transform subjective risk into intersubjective safety.

In relation to who typically aligns and who affiliates online, we found instances when this typicality was breached. When an at-risk user with past trauma experiences presented on the forum with concurrent suicidal ideation and suicidal desire, professional moderators not only aligned but also engaged in affiliation through empathetic responses and/or endorsement of the stance of the at-risk user (Steensig, 2019). When this concurrence was not present, professional moderators demonstrated an understanding of the at-risk user’s experience and focused on the presenting risk by asking safety related questions and offering suggestions that transform risk into safety actions.

This professional moderator focus on risk means that other forum-users can attend to the relational affiliation needs of at-risk users (Jefferson, 1988). Simply put, when other forum-users relationally affiliate, professional moderators do not need to, leaving them to align with the presenting risk in a way that transforms it to safety. What we have demonstrated is the way in which professional moderators work enables each party to enact different in-context roles and attend to different yet complementary user needs. In doing so risk is minimized and the safety on the forum is maintained. This means that it is possible for both risk and safety to co-exist on forums; providing an example of the lived experience of people-helping and the lived experience of mental unwellness working alongside each other to keep forum-users safe. At-risk users can talk of their suicidality and other forum-users can respond with understanding and social solidarity that works to keep the at-risk user safe. An unexpected finding was the shared understanding of the different rights and responsibilities associated with each role as there is little research that has directly explored how professional moderators and forum-users interact about suicidal risk on online mental health forums.

Through understanding how the professional moderator role works in-situ, it becomes clearer how highly skilled the professional moderator role is (Perowne & Gutman, 2024). The skills of professional moderators is enacted for the moment of need. Professional moderators ensure the safety of the forum by undertaking the tasks that must be done (Webb et al., 2008). These actions include identifying and ascertaining risk and taking responsibility for the referral to the most appropriate external crisis services. Working alongside other forum-users is a challenging and unusual aspect of the professional moderator role, as forum-users are also engaged in a dual role of supporting their peers while also balancing their own mental health concerns. Therein lays the complexity for professional moderators when it comes to balancing the safety of all users. Much like the individuals who work as peer support workers alongside qualified professionals in face-to-face settings (White et al., 2020), there is scope for role confusion and a sense of undervaluing of the skills that both professionals and peers bring to the work (Shalaby & Agyapong, 2020). However, in the online space, forum-users and professional moderators appear to be able to work alongside each other in ways that may not be as easily achieved in person. This of course may be because of the different in-context roles that are being taken up in online spaces versus face-to-face settings. Even if this is the case, more research is needed to understand how professional moderators and forum-users work alongside each other to provide often life-saving support and how this may translate to other settings.

While previous research found that professional moderators wished they could use more of their counselling skills to help at-risk users and other forum-users keep safe (Perry et al., 2022), our findings suggest that by taking up a professional moderator role and using those skills, they are already keeping at-risk users, other forum-users, and ultimately the online space safe. It is possible that professional moderators are not cognizant of how “highly skilled” (Perowne & Gutman, 2024) and sophisticated their work in the online space is. For these reasons, study findings could be used to bring greater awareness to professional moderators that their current practices appear to be effective in keeping at-risk users safe. Furthermore, our findings suggest that professional moderators can be a little more confident that they do not need to be “everything to everyone” as not using more of their counselling skills, allows space for other forum-users to step forward and engage in peer support, the very support such forums are founded upon. There is research that demonstrates the benefits of peer support in mental health services, which include an improvement in self-reported recovery, and feelings of value, connection, and empowerment for individuals engaged in offering peer support (Gillard et al., 2022). Thus, by holding back on some of their skills, professional moderators enable forum-users to develop valuable life skills.

As online mental health forums have been in operation for some time, with individuals routinely turning to these online spaces to talk about their troubles and receive support (Pretorius et al., 2019), research is needed to more fully understand the richness of opportunity afforded by online forums. This includes identifying whether these research findings are unique to mental health forums where professional moderators are present, and whether moderators of these forums need to be qualified health professionals. Future research should include practice-based in-situ research that further investigates the interactions that occur between professional moderators and forum-users to achieve safety in these online spaces. More research is also needed to examine the text-based practices of professional moderators and how these are used by forum-users as they become more experienced in the online space, to move at-risk users to places of safety.

As an accompaniment to ecological momentary assessment studies, there are opportunities for longitudinal research that follows specific forum-users over time to examine if and how their engagement with other forum-users and professional moderators’ changes. Such research could capture through language, the dynamic and fluctuating nature of the suicide process, and how it unfolds in-situ. Thus, longitudinal research of this nature will provide an opportunity to “see” the rapid cycling of people in heightened states of suicidal desire as it occurs in real time, and how professional moderators and forum-users interact and respond in real-time.

Like all studies, this study has several strengths and limitations. A strength includes the large online data corpus that analysis has been based upon which consisted of 34 threads and 56 associated emails, which equated to 178 pages of written text. Moreover, what was captured in the corpus was interactions between 16 different moderators and 86 different forum-users, some of whom were at-risk. Furthermore, the access to the off-forum emails between at-risk users and professional moderators provided a rare glimpse into the ability of professional moderators to have both private and public interactions with at-risk users. It demonstrated how private and public interactions can be used in complementary ways to keep forum-users safe. The ability of professional moderators to move fluidly between the public and private spaces reflects a tension that they must manage through specific skill sets that are not typically needed in face-to-face contexts. A further strength was the data spanned a 12-month period, and therefore, could capture any potential peaks and troughs in suicidality (Hofstra et al., 2018). As this is the first study to offer an in-situ exploration of how professional moderators work to keep online forum-users safe, it makes an original contribution to the CA literature. It contributes to how risk and harm is made visible in online settings, and how moderators and forum-users mobilize safety responses. Furthermore, it contributes to the broader CA work that is being done on risk and harm in face-to-face contexts by explicating how online, risk and harm are accounted for as mundane aspects of everyday life.

Several limitations of this study must also be acknowledged. The generalizability of findings is somewhat limited due to the specific technology that underpins the forum website that was used in this study. The technology that underpins any digital platform or space affords particular types of interactions over others (Meredith, 2019). That is, how suicide risk is managed on Twitter may be different to how it is manage in other online space that do not have character limits or discourage emoji use. As the email data was sourced by the partnering forum organization we do not know if the organization deselected any emails, thus we are unsure if selection bias was present in this data. Finally, as the data corpus comes from an online mental health forum based in an English-speaking country, it is likely to reflect Westernized views of suicide. As such findings may not be applicable to forums of non-English speaking countries that do not hold Westernized views of suicide (Chu et al., 2017).

Conclusion

This study demonstrates that professional moderators appear to use a regular pattern of interaction to transform “risk-talk” into “safety-talk” to ensure the safety of at-risk users, the wider forum community, and the online space. This is achieved through professional moderators and forum-users both working to transform subjective risk into intersubjective safety. Additional research is needed to more fully understand how this intersubjectivity unfolds in real-time. Furthermore, while professional moderators may want to be able to use more of their skills to help at-risk users (Perry et al., 2022), we posit that they do not necessarily need to do more to keep users safe. That is not to say that additional benefit may not be gained by permitting professional moderators to use more of their skills online, however, before this is enacted we need to better understand what is occurring in these spaces, how these spaces are places where risk and safety can co-exist, and the roles that all on the forum play. Our findings also contribute to the growing body of literature that encourages practice-based research to further innovate online support for those in crisis (Perry et al., 2022).

Conflict of Interest

The authors have no conflicts of interest to declare.

Acknowledgement

Special thanks to the online mental health forum who helped to make this research possible by allowing the research team access to their forum and email data.

The primary investigator was a recipient of the Australian Government Research Training Scholarship and a UniSQ domestic PhD Stipend Scholarship.